مقاله انگلیسی رایگان در مورد پیش بینی قیمت سهام با رگرسیون بردار پشتیبانی – الزویر ۲۰۱۸

Stock price prediction mechanisms are fundamental to the formation of investment strategies and the development of risk management models (Dash and Dash, 2016, p. 43). The Efficient Market Hypothesis (EMH), however, states that it is not possible to consistently obtain risk-adjusted returns above the prof itability of the market as a whole (Malkiel and Fama, 1970). Computational advances have led to several  machine learning algorithms used to anticipate market movements consistently and thus estimate future asset values such as company stock prices (Gerlein et al., 2016, pp. 193–۱۹۴). Models based on the Support Vector Machine (SVM) are among the most widely used techniques. Information is a valuable resource when building predictive models in the pursuit of profitable financial market transaction systems. Given the peculiarities of financial time series, various challenges must be faced when developing price forecasting systems (Araújo et al., 2015, p. 4081). From a theoretical point of view, under the EMH, relevant information would be widely available to all market participants and immediately reflected in price, according to Malkiel and Fama (1970, p. 383). Malkiel and Fama (1970)’s EMH claims that it is impossible, consistently and over the long term, to achieve above-market returns adjusted to the level of risk assumed. As summarised by Malkiel (2003), the EMH has been questioned since its introduction, especially with the development of Malkiel (2003), the EMH has been questioned since its introduction, especially with the development of predictive systems, as shown in studies based on SVM and other algorithms (for example, Ballings et al. (2015); Nayak et al. (2015); and Qu and Zhang (2016)) that can generate profit in the long term. Malkiel and Fama (1970, pp. 386–۳۸۷), however, argue that the market follows a random walk and that attempts to predict its movements in a consistent manner will be vain. Computational advances have led to the introduction of machine learning techniques for predictive systems in financial markets. In a review of articles on predictive systems, Hsu et al. (2016, p. 215) observed that it is common to use financial series to measure the efficiency of predictive algorithms and classifiers in machine learning. Classifiers are systems that can learn, through training, to recognise patterns and thus assign a class to new data. As an example, machine learning algorithms can be used to predict insolvency, as observed by Zhou et al. (2012) and Li et al. (2012). In such cases, the aim is to classify companies with the highest probability of insolvency, according to an automatic classifier algorithm. Other examples are credit risk measurement, as in Li et al. (2006), and asset price forecasting, as proposed by Kao et al. (2013) and Xiao et al. (2013)