Stock Market Forecasting Using AI Techniques

Abstract

The Stock Market has been a center of attraction for over a long period of time. The amount of return percentage that it encourages to offer is highly motivating for investors and activists. However, with the immense profit it offers, it comes with a wide number of challenges and risks and an uncertain factor which depends on currently dynamic changing scenarios. The stock market prediction has an assumption of relating the prices to previous stock prices patterns and trends. This assumption has been effectively explored by analysts and various studies have been proposed over it to implement them. Machine learning models, time series predictive models and deep learning techniques have been improved to predict the actual price of a stock for over a period of time. The report proposes a methodology to utilize the machine learning models with improved prediction scores adjusted with sentiments and aims to reduce the root mean square error between the already predicted price and the actual price. Sentiments score and polarity metric reduces the error due to present dynamic situations that can not be incorporated via only previous year trends. Since Intra day Sentiments are more prone to stochasticity with more trading and lesser variations in final closing price, an ensemble technique has been used to predict the sentiment score. These models give us a good accuracy score to the actual price however, still can not be the sole reason for an investor for his investments.