Machine Learning based Stock Trading Algorithms

Abstract

Investment is a compulsory subject in modern society. One of the most popular and efficient investment approaches is stock trading. Stock trading can be used to make profits, or even get rich by accurate decisions. However, stock trading is a very profound and complex knowledge for most investors, no matter how much experience they have. That will cause unexpected loss for investors. In order to reduce the loss of investors as well as make decisions more accurate, a decision system using technical analysis is implemented that tries to help the investors to analysis the prospect of stocks and make correct decisions by given some indexes and suggestions. One of the systems applies Oscillation Box Theory and Support Vector Regression to perform the highest and lowest stock closing prices prediction. Another system utilizes Oscillation Box Theory and Deep Neural Networks to predict the similar box bounded by the highest and lowest closing prices. Then both further use Stock Trading Strategy to trade with various stocks based on the predicted Oscillation Box. That can compare which one is better to predict the stock prices or provide all those indexes and suggestions to investors and conclude a final decision.