Novel Approach For Analyzing Intraday Stock Market Behavior Using Stream Data Analytics

Vishal Shah

Abstract: Stock price prediction is an emerging domain of machine learning applications that largely depends on the continuous monitoring and processing of information. One of the open challenges in this domain is to find the best possible machine learning regression algorithm to accurately predict the closing price of any stock in real time using stream data. This research assesses the possibility of solving this challenge - whether a single ML algorithm can be superior for such a task. Further, an incremental learning approach is proposed to improve the performance of the stock market prediction system. Moreover, a prediction system has been implemented that gives buy/sell suggestions and predicts how much the stock price can vary in the future. The experiments are based on the historical data collected from Google Finance API and realtime data fetched from the yfinance python library. For analyzing data and building prediction models, the scikit-learn library has been used to analyze data and build predictions. This paper covers different algorithms that perform well on different stocks and reflects that adding an incremental learning approach boosts prediction accuracy.