关键词: 核主成分分析, 遗传算法, KPCA-GA-SVR模型, 量化选股

Abstract: The problem of quantitative stock selection has always been the focus of study in the financial field. With the development of artificial intelligence techniques, great progress has been made in the quantitative stock study. We establish KPCA-GA-SVR artificial intelligence stock selection model based on stock fundamentals and trading data, and evaluate the predicted effect in short term and medium-and-long term stock. The main conclusions are as follows: firstly, genetic algorithm improves the prediction accuracy of SVR model and avoids over fitting compared with the traditional model; secondly, the predicted effect and robustness get better by extending the genetic algorithm with Kernel Principal Component feature extraction than PCA-GA-SVR model; thirdly, in medium-and-long term, the predicted error of the model runs down with the rise of sliding window length and the one-year prediction accuracy is best; in short term, the result of all kinds of sliding window performs well in one week. The study is useful for individual investors’ investment decisions and the government can make full use of it to monitor the smooth fluctuations in the stock market.

Key words: Kernel Principal Component Analysis, Genetic Algorithm, KPCA-GA-SVR Model, Quantitative Stock Selection