Stock selection model based on advanced AdaBoost algorithm

Stock market is a complex non-linear dynamic system which is affected by many factors. Traditional analysis and forecasting methods are insufficient to accurately reveal the inherent pattern of the stock market, resulting in a big difference between expected and observed results. In recent years, machine learning analytical methods are applied to the stock selection model more often than before and have achieved good results so far. This paper introduces the application of machine learning in stock selection and conducts detailed research on AdaBoost algorithm. The aim is to establish a multi-factor stock selection model based on AdaBoost algorithm, by which we select stock through the analysis of various indicators of a stock. Furthermore, being aware of the flaws of the basic AdaBoost model, we optimize the stock selection model based on actual characteristics of stock selecting process. We also conduct an empirical analysis on Shanghai A-share Stock excluding the ones which were listed before 2010 and which are suspended. We compare the practicality and accuracy of the basic AdaBoost model and the advanced one. Based on experiment results, the advanced AdaBoost model outperforms the basic one by a substantial margin.