Stock Price Prediction: An Incremental Learning Approach Model of Multiple Linear Regression

Abstract

The endeavour of predicting stock prices
using different mathematical and technological methods
and tools is not new. But the recent advancements and
curiosity regarding big data and machine learning have
added a new dimension to it. In this research study, we
investigated the feasibility and performance of the
multiple linear regression method in the prediction of
stock prices. Here, multiple regression was used on the
basis of the incremental machine learning setting. The
study conducted an experiment to predict the closing
price of stocks of six different organizations enlisted in the
Dhaka Stock Exchange (DSE). Three years of historical
stock market data (2017-2019) of these organizations
have been used. Here, the Multiple Regression, Squared
Loss Function, and Stochastic Gradient Descent (SGD)
algorithms are used as a predictor, loss function, and
optimizer respectively. The model incrementally learned
from the data of several stock-related attributes and
predicted the closing price of the next day. The
performance of prediction was then analysed and
assessed on the basis of the rolling Mean Absolute Error
(MAE) metric. The rolling MAE scores found in the
experiment are quite promising.