Abstract

The prediction method plays crucial roles in accurate precipitation forecasts. Recently, machine learning has been widely used for forecasting precipitation, and the K-nearest neighbor (KNN) algorithm, one of machine learning techniques, showed good performance. In this paper, we propose an improved KNN algorithm, which offers robustness against different choices of the neighborhood size k, particularly in the case of the irregular class distribution of the precipitation dataset. Then, based our improved KNN algorithm, a new precipitation forecast approach is put forward. Extensive experimental results demonstrate that the effectiveness of our proposed precipitation forecast approach based on improved KNN algorithm.

Introduction

Precipitation plays crucial roles in climate because it not only is vital for agriculture, forestry and the energy industry but also provides stable habitats for great varieties of species [1–۴]. Nevertheless, heavy rains in a short time usually result in natural disasters such as flash floods, mud-rock flows, urban waterlogging and landslides, which causes tremendous losses in lives and properties of people [5,6]. And for this reason, reliable precipitation forecast is highly important and essentially needed. Unfortunately, the precipitation forecast is a major challenge in meteorology due to formation mechanism of precipitation, not completely understood so far, involves a rather complex physics [7–۹]. And currently the precipitation forecast is far from being satisfactory [10]. Accordingly, the interest in precipitation forecasts has grown significantly in recent years [11–۱۴].