Разработка алгоритма количественного инвестирования на базе Random Forest

Abstract

In modern research of the stock market, specialists and scientists are improving algorithms and models, combining them with each other, with strategies and market conditions for stock selection. This paper presents an overview of stock selection models for quantitative investment, which was the basis for the proposed procedure and algorithm of quantitative investment, which allow modeling the investment process. The developed algorithm is based on the CART decision tree and Random Forest, which includes the bagging algorithm. The bagging algorithm divides the training set into several new training sets that build their own calculation models, and then their results are summed and integrated to obtain the final prediction. The randomness of Random Forest comes into play in the process of selecting samples from the training dataset and in selecting features to calculate the best split points. However, the proposed strategy is more stable than other stock selection strategies, is more suitable for building quantitative stock selection models, the proposed algorithm has an advantage over other algorithms, and is also more promising for further development.