Abstract
This paper examines the relation between bank profit performance and business models, using a machine learning–based approach. The analysis contributes to the literature on this relation by considering the bank portfolio’s ability to yield profits as the identification criterion of strategic profiles and by including all the components of the business model simultaneously in the identification process. Our research strategy is applied to the European Union banking system from 1997 to 2021. The paper’s primary finding indicates that specialization seems to be a strategy that results in banks adopting business profiles with better profit performance, particularly if the banks specialize in the standard retail-oriented model.

Introduction

      What characteristics of bank business models enhance profitability? The search for answers has been among supervisory authorities’ priorities since the end of the Great Recession and is likely to maintain its relevance in the aftermath of the COVID-19 pandemic. The reasons for such interest lie in the fact that the relation between business models and profitability can reveal the viability of banks’ strategic plans and, hence, the viability of banks themselves. In this regard, the Anon (2016) links poorly adapted business models to low bank profitability in advanced economies, identifying the former as an eroding agent of bank resilience over time. However, concern about the profitability of banks’ business profiles has been especially strong in the European Union (EU), as reflected by the fact that the European Central Bank launched a thematic review in 2016 to assess the relation between profitability and business models (European Central Bank, 2018).

      Banking research has examined this relation in two broad and different ways, using regression analysis to explore the effects of the characteristics of the bank business profile on profitability or grouping banks into business models by means of clustering techniques.

Conclusion

     The ability of business models to generate profits is an ongoing concern for banking authorities, especially in the EU. Closely related to this concern, we propose a new methodological strategy to identify business models and to assess their relation to profit performance. Previous works on business models have assumed the business models as a given or have identified them using similarities in balance sheet components. In either of these cases, profit performance is a byproduct of the business model. By contrast, in our identification strategy, profitability is at the core of the definitions of banks’ strategic choices. In particular, business models are obtained using the contributions of the bank portfolio components to profitability as the instruments of clustering analysis.

      The empirical exercise focuses on the EU banking system in 1997–2021. According to the classification of business profiles based on their overall contributions to profit performance, the best profile is a standard retail-oriented model, the second-best one is a standard wholesale model, and the worst profile, which has only equity as a characteristic component, does not match any standard model. Therefore, specialization seems to be a strategy that results in banks adopting business profiles that perform better in terms of their contributions to profitability, particularly if the banks specialize in the standard retail-oriented model.

Abstract

In recent years, with the increasing demand for intelligent society, intelligent photonics has developed rapidly. Machine learning (ML), as a subset of artificial intelligence (AI), has played an important role in the intelligent evolution of optical fiber sensors. Its impact extends beyond enhancing sensor performance by introducing innovative problem-solving approaches. Specifically, ML algorithms have become instrumental in signal demodulation and elevating the efficacy of discrete and distributed sensors, and have also greatly promoted the development of optical fiber speckle pattern processing. This paper presents the latest advancements in ML-based optical fiber sensors, outlines the problems faced by conventional demodulation methods and the common ML algorithms applied in optical fiber sensors, and emphasizes key applications. Additionally, this paper delves into the challenges and future development of this emerging research direction.

Introduction

Due to advancements in powerful computing tools, hardware, widespread use of cloud data, and the evolution of the Internet of Things, the accessibility of large datasets has significantly improved. This has promoted the emergence of many efficient machine learning (ML) algorithms. These algorithms, aided by continuous advancements in data analysis technology, have demonstrated effectiveness in training models and solving specific problems. If ML algorithms are complex and deep enough in terms of layers, then ML can become so-called deep learning (DL) [1]. Regardless of the name, ML seems to have a wide range of applications in modern society, because the achievements of artificial intelligence (AI) often exceed those achieved by manual control design.

In 2021, Nature Photonics further advanced its exploration of the hot topic of optics/photonics through a focused issue called “Machine learning of Light”. As an important type of microphysics, optics seems to have similarities in complexity with some of the “black box” processes in ML. In the Q&A session, David Pille and Aydogan Ozcan discussed the two primary drivers behind the current application of ML in photonics [2]. One direction is to use AI methods on hardware to design optical structures and devices with specific performance. Another direction is to use optical systems as fully optical/hybrid statistical inference models to achieve AI calculations.

Optics and photonics applied to ML computing offer unique advantages, allowing for ultrafast calculations at extreme frame rates and low energy consumption [3], [4], [5], [6]. For instance, Feldmann et al. [7] proposed a fully optical neural synaptic system. They use wavelength division multiplexing technology to implement a scalable circuit architecture for photonic neural networks (NNs) and use optical pulse signals to regulate the phase change materials that encapsulate neurons. This phase change material can regulate transmittance and reflectivity through a controllable laser. This system can perform both supervised and unsupervised learning. The work successfully demonstrates direct pattern recognition in the full optical system. This photonic NN has the potential to leverage the inherent high-speed and high bandwidth of optical systems, thereby enabling the direct processing of full optical communication and visual data. However, logic gates, computation, and NNs are all linear (passive) systems that cannot address nonlinear (active) issues. The lack of effective nonlinear optical processes as activation functions for many nodes poses challenges for all-optical implementation [8].

Conclusion and future perspectives

This paper reviews the application of the ML algorithm in optical fiber sensors. In recent years, the advent of ML has greatly impacted optical development. Compared with static demodulation algorithms and complex spatial optical paths, the ML algorithm offers better signal processing mechanisms for optics. ML algorithms enable automated processing, analysis, and recognition of spectrums, thereby enhancing the efficiency and accuracy of spectrum analysis. Using ML algorithms to classify and identify the spectrum of different substances, enabling qualitative and quantitative analysis of substances. By learning from known corresponding spectrums and building models, ML algorithms can predict and analyze unknown spectrums. ML algorithms can extract and recognize features from the spectrum, analyze the content and proportion of various components, and reveal the intrinsic nature of substances. With the continuous development of spectroscopy and ML algorithms, the application prospects of ML in spectroscopy will become increasingly widespread. This advancement will contribute to more efficient and accurate analytical methods in chemical analysis, environmental monitoring, food safety, and other fields, promoting progress in scientific research and expanding application fields.

Abstract

Most of the companies’ sustainability and growth depend on how well its employees perform. However, the measurement of employees’ performance until now is inconclusive and inexhaustive. To accurately assess and predict an employee’s performance, numerous external factors (physical/environmental, social, and economic) related to an employee’s life have been taken into account in this work. The purpose of this research is to explore an unbiased AI algorithmic solution to predict future employee performance considering physical, social, and economic environmental factors that affect employee performance. We collected data of 1109 employees from the ‘For-Profit Organization’ in Bangladesh from both employers and employees to cover all the factors that justified the unbiased outcome. We utilized a few machine learning tools in this study including the Logistic Regression classifier, the Gaussian Naive Bayes, the Decision Tree classifier, the K-Nearest Neighbors (K-NN), the SVM classification, etc., in order to predict the employee performance evaluation. Then, we compared the effectiveness of those machine learning models by analyzing their precisions, recall, F1-score, and accuracy. This work can be utilized to obtain bias-free employee performance reviews. This fair employee performance assessment can aid decision-makers in making moral choices regarding employee promotions, career advancement, and training needs, among other things. The study also describes notes for future researchers.

Introduction

Performance evaluation, which includes assessing current performance, identifying good and poor performers, and providing feedback to staff, is one of the most challenging components of Human Resource Management (HRM) (Cherian et al., 2021, Fogoroș et al., 2020, Stone et al., 2015). Employee performance evaluations are not practiced systematically by numerous organizations. As a result, the evaluation method becomes erratic and ineffective. A systematic approach should be adopted in order to evaluate employees at the planning stage on a regular basis (Ahmed et al., 2013). Employees with these attributes—skills, dedication, attitudes, and knowledge—are valued as assets by the company (Al-Tit et al., 2022, Li et al., 2008, Yang and Lin, 2009). By creating new knowledge at firms’ level, an organization’s human resources can the firms to innovate (Terán-Bustamante et al., 2021). The accurate assessment of the employees’ performance contributes to the mission of the company with the maximum satisfaction of the employees (Pap et al., 2022). Since company progress depends on employee advancement, numerous executives search for efficient ways to improve performance drastically (Abbas and Yaqoob, 2009, Salam, 2021). To boost performance, employers first need to know the performance condition of the employee in any organization. An article in the Harvard Business Review (Antonio, 2018) claims that essential functions of the organizations, such as prediction, upselling, cross-selling, and performance management can be remarkably influenced by AI technologies (Ledro et al., 2023). In the future, businesses, communities, and nations will be significantly impacted by big data, automation, and machine learning (Lada et al., 2023, Tao et al., 2023). In this regard, AI has started to play a role in business, particularly in HRM, concerning the prediction and decision-making (Nilashi et al., 2023, Qureshi et al., 2023).

Conclusion

The performance of the company’s personnel determines a large portion of its sustainability and growth. Many external aspects (physical/environmental, social, and economic) relevant to an employee’s life have been included in this work in order to measure and anticipate an employee’s performance effectively. The goal of this project is to create an AI algorithmic-based ethical decision-making framework that takes into account the various environmental factors—physical, social, and economic – that have an impact on worker performance. Our results were impartial since we gathered information from a few “For-Profit Organizations” in Bangladesh, both objectively and subjectively. In this study, we used a variety of machine learning methods, and we were able to acquire a reasonable accuracy score. The Random Forest model obtained the highest accuracy score, and the Gaussian naïve Bayes model had the lowest. An impartial employee performance review can be obtained by using this work. Decision-makers can use this equitable employee performance evaluation to help them make moral decisions about training requirements, career advancement, and employee promotions, among other things.

Abstract

The popularity of cloud computing, with its incredible scalability and accessibility, has already welcomed a new era of innovation. Consumers who subscribe to a cloud-based service and use the associated pay-as-you-go features have unlimited access to the applications mentioned above and technologies. In addition to lowering prices, this notion also increased the reliability and accessibility of the offerings. One of the most crucial aspects of cloud technology is the on-demand viewing of personal services, which is also one of its most significant advantages. Apps that are cloud-based are available on demand from anywhere in the world at a reduced cost. Although it causes its users pain with safety concerns, cloud computing can thrive because of its fantastic instantaneous services. There are various violations, but they all accomplish something similar, taking the systems offline. Distributed denial of service attacks are among the most harmful forms of online assault. For fast and accurate DDoS (Distributed Denial of Service, distributed denial of service) attack detection. This research introduced the DDOS attack and a method to defend against it, making the system more resistant to such attacks. In this scenario, numerous hosts are used to carrying out a distributed denial of service assault against cloud-based web pages, sending possibly millions or even trillions of packets. It uses an OS like ParrotSec to pave the way for the attack and make it possible. In the last phase, the most effective algorithms, such as Naive Bayes and Random Forest, are used for detection and mitigation. Another major topic was studying the many cyber attacks that can be launched against cloud computing.

Introduction

DDos attack is a distributed type of attack mode in which an attacker controls a large number of attack machines and sends out DoS attack instructions to the machine. In the latest Internet security report, DDoS attacks remain one of the major cybersecurity threats. The inexpensive pricing and “pay-as-you-go” focused accessibility to computational features and amenities on demand make cloud-based services a formidable competitor to the conventional IT solutions available in prior eras. The use of cloud computing is gaining popularity rapidly. Whether entirely or largely governments and companies have moved their IT infrastructures onto the cloud. Cloud-based Infrastructure offers various advantages compared to traditional, on-site conventional infrastructures. The removal of expenses associated with operation and impairment, as well as the accessibility of materials on request, are only a few of the advantages. However, there are many concerns that cloud consumers have, and the research addresses these issues. The majority of these inquiries centre on safeguarding operational concepts and information. Many security-related attacks can be prevented in conventional IT systems that do not use cloud computing. Focused cloud-based crimes are already using their innovations. Many security vulnerabilities in cloud computing are unique compared to their predecessors in non-cloud computing environments because data and business logic are stored on an external cloud server that lacks accessible oversight. The denial-of-service (DoS) assault is one technique that has been in the spotlight recently. Denial-of-service incidents are directed at the server rather than the people it supports. DoS attackers attempt to flood live servers by masquerading genuine users to overload the service’s capacity to handle incoming inquiries [1]. Cloud computing is an Internet-based service that enables users to access configurable computing resource sharing pools (including server, storage, application software, services, networks, etc.) to achieve online access to computing resources on demand. As a mixture of emerging technologies and business models, cloud computing has developed rapidly in recent years due to its advantages of super-large scale, virtualization, high reliability, good scalability and on-demand services. To overcome this issue, multiple inquiries are sent to the server simultaneously. The term “distributed denial of service,” or DDoS, refers to a variation on the classic “denial of service” that uses numerous computers to attack and impair one service at a time simultaneously.

Conclusion

The key goals of this study are to learn how to recognize and prevent attacks involving distributed denial-of-service. The first and most crucial step is determining which ports can be exploited. Nevertheless, this approach is not risk-free because susceptible ports are more likely to be exploited. Given ParrotSec’s track record for stability and performance, we decided it would be the ideal choice for our company’s computer system. Since a DDoS attack involves sending one million separate packets toward the target, starting with an on-the-internet website would be best. The targeted website was taken offline after it became clear that an assault had happened. Machine learning is constructive in this detecting process as well. Using this data, the most popular and accessible tool, “weka,” is being trained. Employing pre-processing techniques and the “discretize” filter to achieve the desired effect. Therefore, the following phase is not only quite intriguing but also rather useful for both forecasting and detecting. We employed both methods and compared the findings on the same platform, and we found that the naive Bayes method provides the most trustworthy conclusions. PCA selected 21 features from the possible 42 features, while LVQ selected only 20 features. The results suggest that LVQ based feature selection in the DT model may be more accurate than other methods in identifying attacks. As mentioned earlier, the model also outperformed the previous models in terms of accuracy, recall, specificity, and f-score. It was shown that the naive Bayes model had significantly better predictive power than the random forest model. There is a chance that a false positive rate warning will be triggered for packet transmissions within a network. Moreover, when compared to the random forest, naive Bayes produces considerably more accurate forecasts. It was demonstrated that the Naive Bayes algorithm outperformed the random forest technique to identify the false and actual rate of transmissions. The result detection is not carried out in real time. Although attacks can be detected, real-time alarm cannot be realized in the environment of high cluster security, so the feasibility of real-time monitoring under Hadoop platform should be studied continuously.

Abstract

Decision-making in marketing has changed dramatically in the past decade. Companies increasingly use algorithms to generate predictions for marketing decisions, such as which consumers to target with which offers. Such algorithmic decision-making promises to make marketing more intelligent, efficient, consumer-friendly, and, ultimately, more effective. Not surprisingly, machine learning is a trending topic for marketing researchers and practitioners. However, machine learning also introduces important challenges to the marketing landscape. We discuss this development by outlining recent progress and future research directions of machine learning in marketing. Specifically, we provide an overview of typical machine learning applications in marketing and present a guiding framework. We position the articles in the Journal of Business Research’s Special Issue on “Machine Learning in Marketing” within this framework and conclude by putting forward a research agenda to further guide future research in this area.

Introduction

Marketing decision-makers today often struggle to adequately capture and transform (big) customer data into meaningful insights (e.g., De Luca et al., 2021, Sheth and Kellstadt, 2021). Recent research indicates that machine learning (ML)—a field of computer science dedicated to developing learning algorithms, often using big data, to generate predictions needed to make decisions (Agrawal, Gans, & Goldfarb, 2018)—can help companies manage the flood of data (e.g., Davenport et al., 2020, Hagen et al., 2020, Ma and Sun, 2020, Vermeer et al., 2019). ML has been a trending topic in many industries for quite a while now, the marketing industry will be no exception, and it is being used in various industries in the context of both B2C and B2B (e.g., Herhausen et al., 2020, Kumar et al., 2020, Luo et al., 2019, Rust, 2020). ML promises to make marketing more intelligent, efficient, consumer-friendly, and, ultimately, more effective (Huang & Rust, 2021). To put it more directly, proficiency in ML could become an essential skill for numerous marketing researchers and practitioners rather than just a desirable one.

Faced with this development, we aim to highlight recent progress and future research directions on ML in marketing. Such an endeavor is both important and timely, given the dramatic increase in publications on ML in marketing in recent years, as summarized in Fig. 1. Indeed, from 2012 to 2022, the number of publications has increased by almost 600 percent to more than 50,000 yearly publications. Rather than provide an exhaustive literature review, our purpose here is to discuss the most fundamental concepts and topics from past and present research that will drive future research on ML in marketing.1 In doing so, we position the articles of this special issue as an overarching framework that touches upon three key themes, namely employing ML in marketing, benchmarking ML in marketing, and managing ML in marketing.

Conclusion and acknowledgments

We hope this overview article and the papers in this Special Issue serve as an impetus for further research on the important topic of ML in marketing. In addition to summarizing the insights from the published papers, we also discussed important future research directions related to the themes of (1) automated machine learning, (2) data privacy and security, (3) model interpretability, (4) algorithm fairness, and (5) causal machine learning. Selected research questions for ML in marketing derived from these discussions are summarized in Table 2. In addition, broader ML themes might also benefit from a marketing perspective. For example, how can we ensure that ML models are producing ethical and socially responsible models and predictions (De Cremer, 2020; De Cremer & Kasparov, 2022)? Deep learning seems to change how prediction models are developed (Kaur & Sharma, 2023) – what will these techniques’ impact on marketing analytics be in the near future? And, importantly, how will ChatGPT and similar advanced AI technologies change marketing research (van Dis, Bollen, Zuidema, van Rooij, & Bockting, 2023)?

Abstract

Despite providing unparalleled connectivity and convenience, the exponential growth of the Internet of Things (IoT) ecosystem has triggered significant cybersecurity concerns. These concerns stem from various factors, including the heterogeneity of IoT devices, widespread deployment, and inherent computational limitations. Integrating emerging technologies to address these concerns becomes imperative as the dynamic IoT landscape evolves. Machine Learning (ML), a rapidly advancing technology, has shown considerable promise in addressing IoT security issues. It has significantly influenced and advanced research in cyber threat detection. This survey provides a comprehensive overview of current trends, methodologies, and challenges in applying machine learning for cyber threat detection in IoT environments. Specifically, we further perform a comparative analysis of state-of-the-art ML-based Intrusion Detection Systems (IDSs) in the landscape of IoT security. In addition, we shed light on the pressing unresolved issues and challenges within this dynamic field. We provide a future vision with Generative AI and large language models to enhance IoT security. The discussions present an in-depth understanding of different cyber threat detection methods, enhancing the knowledge base of researchers and practitioners alike. This paper is a valuable resource for those keen to delve into the evolving world of cyber threat detection leveraging ML and IoT security.

Introduction

Brendan O’Brien astutely observed, ”If you think the Internet has changed your life, think again. The Internet of Things is about to change it all over again!” [1]. This is indeed the case, as the Internet of Things (IoT) has heralded unprecedented connectivity. The advancements in sensor technology, wireless communication, and data analytics have spurred an exponential increase in connected devices. This influx of connectivity, brought about by integrating IoT into various industries, cities, and households, promotes unmatched efficiency and convenience. As the backbone of IoT, sensors and actuators acquire and convert data from the physical world into digital signals. These compact devices amass a diverse range of data, thereby enabling real-time monitoring and control of numerous systems and processes.

However, the rapid proliferation and extensive integration of IoT devices into everyday life have ushered in various security challenges. These issues must be robustly addressed to ensure the safety and reliability of this expanding ecosystem. The sheer volume and variety of IoT devices and their often inconsistent security features and protocols engender a fragmented environment teeming with potential attack vectors. IoT devices frequently prioritize low cost and user simplicity over security, making them susceptible to breaches and exploitation. As a result, these devices are at risk of various cyber threats, including data breaches, Distributed Denial-of-Service (DDoS) attacks, and malware infections. Any security breach in these devices could significantly compromise privacy and crucial infrastructure systems, given the sensitive nature of the data they handle. Moreover, IoT devices, potentially serving as entry points, might allow attackers to infiltrate broader networks, amplifying the potential impact of security breaches. Another primary concern is the security of communication routes between IoT devices and networks, as many IoT devices utilize wireless communication protocols susceptible to interception or manipulation. These vulnerabilities can be exacerbated by the resource constraints of specific IoT devices, which prevent them from adopting contemporary encryption and authentication techniques. Furthermore, the long lifespan and widespread deployment of IoT devices compound the difficulty of managing security upgrades and patches, as many devices may not receive regular updates or may be difficult to access for maintenance. This could lead to an increased number of outdated or vulnerable devices, further exacerbating security concerns [2].

Conclusion

IoT devices have become fundamental to everyday life, offering increased connectivity and convenience. However, this accelerated development in IoT devices also brought several security challenges that need to be addressed to guarantee the safety and reliability of these interconnected systems. The survey was organized into five sections detailing the role of machine learning in enhancing IoT security. We have provided an overview of the current trends in ML for cyber threat detection in IoT environments. Furthermore, we survey recent cyber detection methods, define them, highlight the approach, and detail the attack surface utilized along with their evaluations. The utilized ML techniques are also discussed, defined, and compared regarding advantages and drawbacks in the relevant use cases. Moreover, open issues were discussed briefly, concluding that further research and development is required as no current solution can address the issues related to IoT cyber threat detection. A holistic strategy combining diverse techniques and strategies is required as a final recommendation. We present this survey as a reference for the current advancements in the field and point out the direction being taken. Our goal is for the survey to create a more secure and resilient IoT environment by identifying the current issues, understanding their root causes and possible implications, and developing innovative solutions.

Abstract

Machine learning can be effectively used to generate models capable of representing the dynamic of production processes of small and medium-sized enterprises. These models enable the estimation of key performance indicators, and are often used for optimizing production processes. However, in most industrial applications, modeling and optimization of production processes are currently carried out as separate tasks, manually in a very costly and inefficient way. Automated machine learning tools and frameworks facilitate the path for deriving models, reducing modeling time and cost. However, optimization by exploiting production models is still in infancy. This work presents a methodology for integrating a fully automated procedure that embraces automated machine learning pipelines and a multi-objective optimization algorithm for improving the production processes, with special focus on small and medium-sized enterprises. This procedure is supported on embedding the generated models as objective functions of a reference point based non-dominated sorting genetic algorithm, resulting in preference-based Pareto-optimal parametrizations of the corresponding production processes. The methodology was implemented and validated using data from a manufacturing production process of a small manufacturing enterprise, generating highly accurate machine learning-based models for the analyzed indicators. Additionally, by applying the optimization step of the proposed methodology it was possible to increase the productivity of the manufacturing process by 3.19 % and reduce its defect rate by 2.15 %, outperforming the results obtained with traditional trial and error method focused on productivity alone.

Introduction

Nowadays, generalized adoption of the new manufacturing paradigms involves the assimilation of key technologies such as intelligent data analysis and machine learning (ML), among others, aiming at the digital transformation of enterprises [ 46 , 56 ]. This digital transformation is crucial to keep up with the competition, especially in the case of small and medium-sized manufacturing enterprises (SMEs). One contemporary target is to create a highly reconfigurable, decentralized, dynamic, self-organizing, and real-time (or near real-time) decision-making infrastructure enabling to analyze customer expectations and reach their targets [ 20 ]. By applying these transformations, SMEs should be capable of monitoring and improving their key performance indicators (KPIs) [ 67 ]. However, in practice, SMEs face several difficulties in applying these technologies, mostly related to the transition and maintenance costs, innovation complexity, and personnel training [ 50 ]. Additionally, due to limited human and computational resources, time constraints and complexity of the optimization processes, SMEs usually focus their efforts on a single productivity objective, despite being more desirable to consider and optimize multiple objectives, which leads not only to more efficient but also a more sustainable and environmentally friendly production. Due to these obstacles, a large number of SMEs don’t count yet with the necessary tools to continue with their digital transformation. In this context, the development of tools for generating useful information and smart recommendations of production systems in SMEs is almost mandatory in a high-competitive market and has a large number of potential adopters [ 12 , 42 ].

Conclusions

This work presents an automated machine learning methodology for optimizing manufacturing processes in SMEs by combining the standard tasks of AutoML tools, such as data preprocessing, feature selection, model training, and hyperparameter optimization, with preference-based multi-objective optimization. For this purpose, the basic AutoML workflow is used to generate models for each of the KPIs of the production process and, then, a new automated optimization step is introduced for using the generated models as objective functions, resulting in optimal parametrizations of the production process. By simplifying the way of interacting with this methodology, it is possible that manufacturing SMEs with low availability of highly-skilled personnel or limited computing power can benefit from advanced technologies making easier the digitalization and application of Industry 4.0 paradigm.

The methodology was implemented and validated in a production process where, firstly, the most relevant features for modeling each key performance indicator were automatically selected based on the Pearson’s correlation coefficient, allowing to reduce the dimensionality of the data. Then, models of key performance indicators were generated and their architecture/hyperparameters optimized. Generated models were compared to models obtained through other AutoML frameworks offering similar results, with values of MSE = 2.585 and R2 = 0.999, and MSE = 6 × 10−5 and R2 = 0.735, respectively. Finally, the models were used as objective functions in the R-NSGA-II algorithm for finding optimal parametrizations of the production process, yielding an improvement in both KPI, reducing scrap by 2.15 % and increasing throughput by 3.19 %, with regard to the baseline of conventional parametrization considering only a single productivity target. These improvements contribute to a higher production rate while, at the same time, the number of defective components is reduced, which underscore the potential of the proposed methodology to significantly boost overall efficiency and profitability for SMEs by optimizing their production processes more holistically.

Abstract

Purpose
This study examined the effect of human resource development practices on employee performance, with job satisfaction as a mediating variable among academic staff at Mizan Tepi University.

Methods
A standardized Likert-scale questionnaire was administered to academic staff members selected through stratified random sampling from eight colleges at Mizan Tepi University. Descriptive and inferential statistics, including a structural equation model and t-tests, were used for data analysis.

Results
The study found a significant positive correlation between human resource development practices and employee performance. Job satisfaction was identified as a mediator between HRM practices and employee performance. Specific dimensions of human resource development practices, such as training and development, academic career development, teamwork spirit, and counseling, had significant effects on employee performance, with both positive and negative impacts observed. Additionally, academic career development, succession planning, and counseling were found to have a statistically significant positive effect on the academic staff at Mizan Tepi University.

Conclusions
This study found that training and development, academic career development, teamwork, counseling, and job satisfaction significantly impact employee performance. The results of the regression analysis using the SEM model revealed that job satisfaction, training and development, and teamwork positively affect academic staff performance while negatively affecting academic career development and counseling. Academic career development, teamwork, counseling, and succession planning were found to have a positive effect on job satisfaction. The mediation analysis indicated that academic career development and counseling partially mediate employee performance, while succession planning fully mediates it. Therefore, focusing on training and development, performance appraisal, rewards, teamwork, academic career development, succession planning, and counseling is essential to enhancing job satisfaction among academic staff at Mizan Tepi University.

Introduction

A framework for increasing human capital inside an organization by developing the company and its employees to boost performance is known as human resource development [1]. Through planned learning activities, human resource development practices create essential competencies that people in companies need to accomplish their present and future roles. Consequently, the concept of HR development has emerged as a strategy to enhance the capacity of existing employees in organizations to perform their duties efficiently. Human resource development (HRD) practices apply to various types of organizations within the business sector that use HRD to initiate and manipulate exchanges that include knowledge sharing, skill development, information transfer, collaboration, feedback, or mentorship. HRD also guarantees a healthy balance between the desires of individual employees and those of the entire organization [1]. According to published studies [2], the foundation of human resource development is the enhancement of organizational performance through the work and skills of people. It seeks to enhance the accomplishment of the purposes and objectives of an organization. The main aim of HRD is to prepare employees to work in the current environment to the best of their abilities and to develop these abilities in preparation for upcoming challenges. Training and development programs create awareness to contribute to employee effectiveness, job performance, and overall organizational success among employees through the use of new technologies [3]. According to Ref. [4], one of the vital traits in the field of HRD is the focus on work-based training, or what may often be referred to as informal education techniques. To accomplish their objectives, organizations must improve the skills and capacities of their workforce by supporting individual training and creating settings where knowledge may be generated, shared, and used by a variety of stakeholders as well as by individual workers. According to Ref. [5], the relationship and impact of HRD practices on job satisfaction indicate that the HRD climate has a definite impact on job satisfaction, which in turn leads to increased organizational performance. They studied the impact of HRD on the job satisfaction of employees in selected public sector enterprises and subsequently studied the impact of HRD on employees and organizational performance. According to Ref. [6], there is a significant relationship between the HRD environment and job performance, and any positive change in the HRD environment will bring about positive changes in job performance. The HRD climate refers to the overall environment, culture, and practices within an organization that support and promote human resource development. HRD practices are the primary way organizations attempt to maintain the competency levels of their employees, thus increasing their performance levels. Ineffective HRD practices can result in various problems, such as a reduction in employees’ ability to develop and apply new skills, a decrease in staff productivity, an increase in employee turnover, and lower overall performance levels within organizations [7]. Problems relating to HRD result from HRD programs failing to address organizational and staff needs. When HRD initiatives are carried out effectively, they may boost productivity and save costs associated with personnel turnover, services, strikes, and losses. HRD plays a vital role in improving job performance and overall organizational effectiveness. Some studies have addressed current HRD practices as an HRM issue. For example, new-entry academic staff, especially those who receive training in SPSS, STATA, and SEM and who have skill gaps, will be more effective as a result of HRD practices. This means that they are better equipped with the necessary skills, knowledge, abilities, and competencies to perform their job tasks successfully and contribute to organizational effectiveness because the formal educational system does not adequately teach specific job skills for a position in a particular organization, so such employees will not have the requisite skills, knowledge, abilities, and competencies needed to perform a specific job. As a result, HRD activities are required to enhance their performance and to help them make substantive contributions to the organization’s growth. However, inefficient HRD practices can lead to many problems, such as reduced employee enthusiasm for learning and applying new skills, decreased employee productivity, low morale, and higher employee turnover rates [8]. One of the most prominent and well-respected jobs in the world is teaching.

Conclusions

The main objective of this study was to examine the effect of human resources development practices on employee performance and explore whether job satisfaction mediates the relationship between human resources development practices and employee performance. The study focused on academic staff at Mizan Tepi University as the sample population. The following conclusions have been drawn in light of the study’s objectives and findings:

Firstly, the analysis of demographic variables revealed that there were no significant mean differences between male and female respondents in terms of job satisfaction. However, gender affects employee performance. This suggests that gender diversity should be promoted in organizations, creating an inclusive environment and policies that eliminate gender-based discrimination.

Secondly, the results of the correlation analysis showed a positive and significant relationship between training and development, academic career development, teamwork, counseling, and job satisfaction. It implies that when the university effectively implements human development practices, such as rewarding employees, implementing succession planning, promoting academic career development, and fostering teamwork, there is a proportional improvement in employee performance. Furthermore, all the HRD practices showed significant and positive relationships with employee job satisfaction. This underscores the significance of focusing on training and development, performance appraisal, rewards, teamwork, academic career development, succession planning, and counseling to enhance job satisfaction among academic staff at Mizan Tepi University.

Abstract

Customer lifetime value is one of the most important tasks for enterprises to maintain customer relationships. However, due to the limitations of using a single data mining method, the measurement of customer lifetime value under the condition of noncontractual relationship has always been a research difficulty. This paper focuses on customer value measurement and customer segmentation based on customer lifecycle value theory, and carries out customer value measurement and customer segmentation research from the perspective of customer value, and constructs customer segmentation model. This paper first conducts feature engineering, such as data selection, data preprocessing, data transformation, and knowledge discovery, and then conducts customer value segmentation based on machine learning algorithms and customer relationship management analysis models and builds a customer value segmentation identification model under the condition of noncontractual relationship. Finally, empirical analysis is carried out with the real customer transaction data of the actual online shopping platform, which verifies the validity and applicability of the customer segmentation method and value calculation method proposed in this paper.

Introduction

Since the 1980s, customer relationship management (CRM) has increasingly become the focus of academia and industry. Facts have proven that maintaining good relationships with specific customers can increase corporate profits and increase the significant advantages of enterprises in market competition [1]. Enterprises can classify customer categories based on customer consumption behavior, customer lifetime value and other information, identify customer value, and apply it to CRM [2]. Scientific and effective research on customer lifetime value is the primary task of customer relationship management, which is of great significance to enterprise marketing plans and strategies. Scholars have confirmed that effective customer lifetime value segmentation research can bring more profits to enterprises [3]. However, due to the inadequacy of previous research on customer data processing and the limitation of single data mining methods, it is difficult to fully mine customer behavior information. The accurate measurement of customer life cycle value under non contractual relationships has always been a research difficulty. Through a literature review, the current calculation of customer lifetime value under noncontractual relationships mainly includes Pareto/NBD and other probability models [4,5], VAR, and other time series models, as well as machine learning algorithms [6]. Especially in recent years, with the wide application of Big Data in business intelligence and analytics, companies can collect a large amount of personal transaction data from customers at low cost. These can be analyzed using an array of methods including: neural network models, decision tree models, random forests, generalized additive models (GAMs), multiple adaptive regression splines (MARS), classification and regression trees (CART), support vector machines (SVM) and other machine learning algorithms. These methods are useful for the application of customer lifetime value calculation.

Conclusion and outlook

This paper studies the connotation of customer lifetime value under the noncontractual relationship from two aspects of customer segmentation and calculation methods, and constructs the customer segmentation mode under the noncontractual relationship and the customer value segmentation recognition model based on machine learning algorithm. This paper uses feature engineering to process the data, and comprehensively applies RFM model, machine learning algorithm and other methods to ensure the rationality and feasibility of the research results. The main innovation points of this paper are as follows: (1) Define the measurement method of customer value, build a customer segmentation model based on customer value under the noncontractual relationship, mine the characteristics of segmented customer groups, and enrich the existing research on customer value segmentation; (2) In the preprocessing stage of customer consumption behavior information, feature engineering is used to map the original data space to the new feature vector space through the data preprocessing and transformation process, so that in the new feature space, the machine learning model can better mine the feature vector information and improve the classification results of the model; (3) The traditional RFM model and machine learning algorithms are combined to expand the research on customer segmentation and customer lifetime value under the condition of noncontractual relationship.

Abstract

     Technology is pivotal in the rapid growth of services and intensifying the quality of life. Recent technology, like the Internet of Things (IoT), demonstrates an impressive performance in fast-forward development. Intrusion Detection System (IDS) is used as a lifeline to prevent attacks by classifying the activities as normal and suspicious. In this paper, we propose a two-phase IDS for IoT. In the first phase, we categorize data into four sections according to the data types (i.e., nominal, integer, binary, and float). We then classify them using different versions of the Naive Bayes classifier. After that, we use majority voting to choose the final result of the classification. In the second phase, we pass those data which behave normally or are benign in the first phase and classify them using an unsupervised elliptic envelope. We validated our work using the standard NSL-KDD, UNSW_NB15, and CIC-IDS2017 datasets. We found the proposed method more efficient than existing IDS techniques and achieved reasonable accuracy in the first phase. Furthermore, the benign data is sent to the second phase of the analysis. After the second phase, we achieved a 97% accuracy in the NSL-KDD dataset, 86.9% in the UNSW_NB15 dataset, and 98.59% accuracy in the CIC-IDS2017 dataset.

Introduction

     In the evolutionary era, the Internet has always been performing a most significant role. Globally, the total estimate of Internet users is projected to increase from 3.9 billion in 2018 to 5.3 billion by 2023, as stated by Cisco Annual Internet Report [1]. Furthermore, the Internet of Things (IoT) is becoming increasingly widespread. IoT integrates many heterogeneous objects (such as in a smart home: intelligent bulbs, refrigerators, fans, air conditioners, automated doors, and TVs.) with various connecting technologies such as Bluetooth Low Energy (BLE), WiFi, and ZigBee. There are also other domains and applications in which the IoT can play an important role and enhance our lives quality. These applications include smart transportation, industrial automation, agriculture, and healthcare [2].

     The IoT model [3] has been emerging towards formulating a cyber–physical environment where everything can be found, operated, investigated, and modernized. Because of being connected, the chances of attacks on the network increase. Many attacks and malicious incidents can affect different layers of the IoT architecture, creating security concerns. Makhdoom et al. [4] discussed the commonly known attacks on different layers, depending on the anatomy of the malware, and IoT-enabled cyber-attacks are also illustrated in a survey [5]. Similarly, Zarpel et al. [6] elaborated on intrusion detection systems in IoT. They have classified IDS based on placement strategies, detection methods, security threats, and validation strategies. Zargar et al. [7] explained in detail about Distributed Denial of Service (DDoS) attacks and also classified the countermeasures.

Conclusion

     This paper presents a machine learning-based two-phase IDS. Firstly, we categorize data into four sections according to the data types (e.g., nominal, integer, binary, and float). Then classify them using different versions of the Naive Bayes classifier. After that, with the help of majority voting, we choose the final result of the classification. In the second phase, we pass those data which behave like normal in the first stage, and these data are classified using an unsupervised elliptic envelope. It draws an imaginary envelope and assigns value 1, which lies inside the Envelope, and −1 outside the Envelope. Our proposed model is also performing very well in the imbalanced distribution of the data by providing the weight initialization to each class. Finally, we got an overall 97% accuracy with a meager false positive rate. The drawback of this model is that it does not work pretty in multiclass classification.

     In future work, we will improve the multiclass classification and feature engineering techniques model, expand this procedure in real-time for network traffic analysis, and evaluate performance. We will further attempt to capture network data by deploying IoT devices in the real world. Moreover, the IDS system will detect malicious incidents in real-time and immediately take appropriate action to prevent damage.

Abstract

Introduction
Child morbidity and mortality in resource-limited settings is a major public health problem. The previous studies were mainly concerned with determining the prevalence of child deaths and identifying associated factors. Extracting knowledge and discovering insights from hidden patterns in child data through supervised machine learning algorithms is limited. Therefore, this study aimed to predict the under-five death of children using a best performance-supervised machine learning algorithm.

Methods
A total of 1813 samples were used from the 2019 Ethiopian Demographic and Health Survey dataset. 70% and 30% of total instances were used for training the model and measuring the performance of each algorithm with 10-fold cross-validation techniques respectively. Five supervised machine learning algorithms were considered for model building and comparison. All the included algorithms were evaluated using confusion matrix elements. Information gain value was used to select important attributes to predict child deaths. The If/then logical association was used to generate rules based on relationships among attributes using Weka version 3.8.6 software.

Results
J48 is the second-best performance algorithm next to the random forest to predict child death, with 77.8% and 93.9% accuracy, respectively. Late initiation of breastfeeding, mothers with no formal education, short birth intervals, poor wealth status of the mother, and unexposed to media were the top five important attributes to predict child deaths. A total of six associated rules were generated that could determine the magnitude of child deaths. Of these, if children were rural residents, had a short birth interval, and if born as multiples (twins), then the probability of child death was 83.6%.

Conclusions
Five machine learning algorithms were included to predict child deaths and generate rules. Hence, the random forest algorithm was the best algorithm to predict child deaths. However, the study was limited since important attributes were not included in the data source, and irrelevant values were found. So, researchers are encouraged to use machine learning algorithms for future studies including important attributes that could predict child death. The current findings would be useful for stakeholders’ preparedness, and taking proactive childcare interventions. Encouraging women in education, media access, and economic development programs are essential interventions for child death reduction.

Introduction

Under-five mortality is the most important indicator to measure the health status of children, and it is a key marker for the development of countries [1]. The under-five mortality rate is the probability of children dying before their fifth birthday [2]. Globally, nearly 44% of all under-five deaths occurred before their first month of birth [3], and an estimated 4.1 million child deaths occurred in 2017 [4]. According to the Centers for Disease Control and Prevention, child mortality in the United States in 2020 was predicted to be 5.4 deaths per 1000 live births [5].

The risk of under-five mortality is highest in low-income countries. The under-five mortality rate in low-income countries was predicted to be 69 deaths per 1000 live births in 2017, which is almost 14 times the rate in high-income countries [6,7]. In Bangladesh, 522 under-five children died per 1000 live births [8]. In 2001, under-five mortality in Nepal was projected to be 91 deaths per 1000 live births [9]. Though under-five mortality shows a reduction from 166 to 67 per 1000 live births over a period of 16 years [10], Ethiopia appears to have the fifth-highest number of new-born deaths in the world [11]. Under-five mortality is projected to cause 472,000 children to die annually in Ethiopia before their fifth birthday, which places Ethiopia sixth in the world according to the number of under-five deaths [7,12]. According to WHO 2017, more than half of under-five deaths are due to infectious diseases that are easily preventable and treatable through simple and affordable interventions [13]. Under-five mortality is also caused by undernutrition, which further leads to stunting and wasting [14].

Conclusions and recommendations

This study aimed to identify the best-supervised machine learning algorithms to classify and select important attributes to predict the death of children before their fifth birthday. In line with the objectives, six supervised machine learning algorithms were considered that accurately predict the death of children before celebrating their fifth birthday. Different confusion matrix element was used to compare the candidate-supervised machine learning algorithms. Based on the result, the random forest algorithm was the best performance model to predict the death of children before celebrating their fifth birthday. Attributes such as late initiation of breastfeeding, mothers with no formal education, the short birth interval, poor wealth status of the mother, and being unexposed to media were the top important attributes to predict child deaths.

Generating associated rules for child death was another objective of the study. Accordingly, six rules were generated that were associated with the deaths of children before celebrating their fifth birthday. The findings of this study would have practical implications by supporting policymakers and stakeholders in developing childcare intervention mechanisms and preparing themselves to care for children as early as possible. Stakeholders are recommended to encourage mothers to initiate breastfeeding at the appropriate time. Improving mothers’ wealth status, closing the gap in media access, and creating awareness among mothers would be critical interventions to enhance the survival of children in Ethiopia. The generated rules would also have theoretical implications by extracting and representing knowledge. Moreover, researchers would use this study as a baseline and framework for further research studies, including important attributes that would predict child mortality in low-income countries.

Abstract

The use of computers and the internet has spread rapidly over the course of the past few decades. Every day, more and more people are coming to rely heavily on the internet. When it comes to the field of information security, the subject of security is one that is becoming an increasingly important focus. It is vital to design a powerful intrusion detection system in order to prevent computer hackers and other intruders from effectively getting into computer networks or systems. This can be accomplished by: (IDS). The danger and attack detection capabilities of the computer system are built into the intrusion detection system. Abuse has occurred and can be used to identify invasions when there is a deviation between a preset pattern of intrusion and an observed pattern of intrusion. An intrusion detection system (IDS) is a piece of hardware (or software) that is used to generate reports for a Management Station as well as monitor network and/or system activities for unethical behaviour or policy violations. In the current study, an approach known as machine learning is suggested as a possible paradigm for the development of a network intrusion detection system. The results of the experiment show that the strategy that was suggested improves the capability of intrusion detection.

Introduction

Over the past few years, there has been an increase in the usage of computer systems to make the lives of consumers easier and more convenient. When people try to takeadvantage of the amazing capabilities and processing capacity of computer systems, however, security has been one of the most significant problems in the field of computer science. This is because attackers frequently try to break into systems and act maliciously, such as stealing vital information from a corporation, rendering the systems useless, or even destroying the systems. Internal attacks, such as pharming, distributed denial-of-service (DDoS), eavesdropping, and spear-phishing attempts, are often among the most difficult to identify of all well-known attacks. This is due to the fact that firewalls and intrusion detection systems (also known as IDSs) often guard against attacks from the outside. At this time, the majority of systems authenticate users by analysing a login pattern consisting of the user ID and password. As a result of this, we have proposed in this study a security solution that we have dubbed the Internal Intrusion Detection and Protection System (IIDPS) [1]. This solution recognises hostile or malicious behaviour carried out against a system at the System call level. IIDPS uses data mining and forensic profiling techniques in order to mine system call patterns, also known as SC-patterns, which are the longest system call sequences (SC-sequences) that have repeatedly appeared numerous times in a user’s log file for the user. SC-patterns can be used to identify malicious activity. The user’s computer usage history is used to compile the user’s forensic features, which are then defined as a SCpattern that commonly appears in the user’s own submitted SCsequences but is rarely utilised by other users. This information is gleaned from the user’s computer.

Conclusion

As a consequence of this, within this work, we suggest asecurity system that we refer to as the Internal Intrusion Detection and Protection System (IIDPS). This system is capable of identifying hostile behaviour that is aimed towardsa system at the SC level. The IIDPS minessystem call patterns, also known as SC-patterns, which are defined as the longest system call sequences (System Call-sequences) that have repeatedly appeared numerous times in a user’s log file for the user. These are the operations carried out by the user, such as sending a file, updating a file, or viewing a file, and they are validated by an administrator. The user’s computer usage history is used to compile the user’s forensic features, which are then defined as a SC-pattern that commonly appearsin the user’s own submitted SC-sequences but is rarely utilised by other users. This information is gleaned from the user’s computer.

Abstract

Nowadays many researchers are developing various distributed and decentralized frameworks for federated learning algorithms. However, development of such a framework targeting smart Internet of Things in edge systems is still an open challenge. In this paper, we present our solution to that challenge called Python Testbed for Federated Learning Algorithms. The solution is written in pure Python, and it supports both centralized and decentralized algorithms. The usage of the presented solution is both validated and illustrated by three simple algorithm examples.

Introduction

Federated learning was introduced by McMahan et al. [1] as a decentralized approach to model learning that leaves the training data distributed on the mobile devices and learns a shared model by aggregating locally computed updates. They presented FedAvg, a practical method for the federated learning of deep networks based on iterative model averaging, see Algorithm 1 FederatedAveraging in [1] on page 5. The main advantages of federated learning are: (i) it preserves local data privacy, (ii) it is robust to the unbalanced and nonindependent and identically distributed (non-IID) data distributions, and (iii) it reduces required communication rounds by 10–100x as compared to synchronized stochastic gradient descent (FedSgd).

McMahan’s seminal paper [1] inspired many researchers’ papers and in this limited space we mention just few of them. Immediately after [1], Bonawitz et al. [2] introduced an efficient secure aggregation protocol for federated learning, and Konecny et al. [3] presented algorithms for further decreasing communication costs. More recent papers are focused on data privacy [4, 5].

Conclusion

In this paper, we developed the federated learning framework targeting smart IoTs in edge systems called Python Testbed for Federated Learning Algorithms (PTB-FLA), with the primary intention to be used as a framework for developing FLAs on a single computer. The solution is written in pure Python, and it supports both centralized and decentralized algorithms. The PTB-FLA usage is both validated and illustrated by three simple algorithm examples.

The main PTB-FLA advantages are the following: (i) it keeps the application footprint small so to fit to smart IoTs and (ii) it keeps the installation as simple as possible (with no external dependencies).