Robotic Process Automation for Stock Selection Process and Price Prediction Model using Machine Learning Techniques

Abstract

The usage of AI in the financial industry has increased the reliance on stochastic models for forecasting market behaviour. Quantitative analysts are constantly working to increase the precision with which machine learning models predict stock returns. Regression models using Support Vector Machine (SVM) and Random Forest are well renowned for their ability to predict closing prices with high accuracy. This study suggests a method for analysing and forecasting stock prices using an ensemble of these algorithms.Using basic market price data from India's National Stock Exchange (NSE), datasets are used that have been preprocessed to add common technical indicators as features. The size of the training dataset is decreased by using feature selection techniques to rank the features according to their impact on the final closing price. Sentiment analysis is also used in the study to examine the effect of investor sentiment on stock prices. Twitter postings with a specific corporate hashtag are rated as good or bad using a trained Word2Vec model. The ensemble model is then trained on a fresh dataset made up of counts of both positive and negative tweets across time as well as technical indicator data.This paper makes a contribution to the area by offering an ensemble model for stock price prediction that blends SVM and Random Forest regression models. The study illustrates the significance of feature selection in lowering dataset size as well as the limited influence of aggregated sentiment analysis from Twitter data on the performance of the model as a whole. For researchers and quantitative analysts looking to improve the precision of stock price prediction models in the financial industry, these findings offer invaluable insights.