Stock market prediction using artificial intelligence

Abstract

This work is divided into two sections, the first of which is a natural language processing module that analyzes sentiment for corresponding financial news, and the second of which is a stock prediction module that uses the output of the first module and past stock data to forecast future stock prices. This structure can more precisely forecast the stock's future trajectory. The natural language portion is built based on a recurrent neural network, which helps understand the link between nearby inputs better and is more similar to how humans interpret language. To increase the accuracy of the system, the stock prediction module uses the support vector machine method and mixes several kernel functions. This study builds a stock prediction network structure for a series of tests using the Python language on the Linux 16.4 environment. Although our experiment only employs a restricted number of stocks, it nevertheless yields competitive prediction results, which supports the conclusion analysis of this study.