An Integrated Machine Learning Approach Predicting Stock Values Using Order Book Details

Abstract

Machine learning, Artificial Intelligence and deep learning can be applied in numerous fields like medical diagnosis, computer networks, travel and tourism, banking, stock market, etc. The proposed approach emphasizes the stock market to forecast the stock values, average prices, turnover, etc. This integrated model is built on recent methods such as linear regression, multiple linear regression, classification, clustering, long short term memory (LSTM), convolutional neural network (CNN), etc. Model processes the sample data downloaded from the National Stock Exchange (NSE) which consists of the previous stock value, open price, high price, low price, average price, turnover, total traded quantity, etc. The model applies the linear regression, multiple linear regression, K-means clustering, Bayesian correlation and LSTM to forecast the desired values such as future turnover, probable total traded quantity if the average price and previous closing prices are increases. It executes the multiple linear regression and found as the previous closing price of the stock and average price on a daily basis increases it directly affects the total traded quantity. K-means and Bayesian correlation helps in estimating the stock values, turnover and total traded quantity. Bayesian correlation helps in correlating the variables and estimates the evidence in terms of ‘very strong’, ‘strong’, ‘moderate’, etc., and generates charts/plots accordingly. The advantage of using such model is less human intervention, minimal time to process the data and accurately generate the results. This approach consists of both practical and theoretical aspects to explore the stock market with different features.