A Hybrid GA-SVM and Sentiment Analysis for Forecasting Stock Market Movement Direction

Authors

  • Shashi Shekhar
  • Neeraj Varshney

Abstract

For a company, important indicator is stock price. This value is affected by various factors. Differently, public emotion and sentiment can be affected by various events. Price of stock market can be affected by this. Price of stock are not a static one due to dependency of various factors. They are nonlinear time series data, has high amount of noise and dynamic. To this area of research, machine learning algorithms may be applied to solve problems in nonlinear predictions because of its high learning ability.

For predicting stock price, learning based methods are used and its performance can be enhanced by various methods. Still it is challenging task to predict stock market. Psychology of investors are reflected by User-generated textual content provided by internet. On stock market, important role is played by sentiment of investor and it is used for prediction of stock price. Support vector machine based on Genetic algorithm is integrated with machine learning based sentiment analysis.

An improvement about 18.6% in accuracy is obtained by combing sentiment variable. Final accuracy is about 89.93%. Risk of investors can be reduced and they are allowed to make wide range of decisions by combining proposed method with stop-loss order strategy. Asset fundamental value information is contained by sentiment. Stock market can be indicated by this in an effective way. Time interval can be expanded in future for gathering huge amount of textual documents.

Downloads

Published

2020-01-01

Issue

Section

Articles