Author : Vinayak Kusumkar and Prof. K. C. Waghmare

Time series data is quite different from the traditional machine learning dataset as it presents order dependency between observations. The main aim of time series forecasting is to understand the behavior of observed series and predict future values of that series based on the previous pattern of the series. Stock market movement is one of the ideal and the most volatile examples of time series. Forecasting stock volatility can give rise to better trading strategies which can limit the risks and enhance the return. Apart from the historic data, information from different news, discussion boards, and social media can be used to predict the future movement or volatility of the stocks. Regression, classification, deep learning, etc. are some approaches that can be applied individually or ensemble of these techniques on the stock market data. The ensemble of techniques gives better performance as compared to techniques applied individually. The proposed approach uses historic data and news content to forecast stock market volatility using an ensemble of machine learning models.

Keywords: Time Series Forecasting, Machine Learning, Ensemble Methods, Stock Market Prediction