Title: Evaluation of factors involved in predicting Indian stock price using machine learning algorithms

Authors: Archit A. Vohra; Paresh J. Tanna

Addresses: R K University, Bhavnagar Road, Kasturbadham, Rajkot – 360020, Gujarat, India ' School of Engineering, R K University, Rajkot, India

Abstract: This study evaluates the effect of training dataset size, dimensionality and rolling dataset on the prediction accuracy of decision tree regression (DTR), support vector regression (SVR), long short-term memory (LSTM) and neural network multi-layer perceptron (NNMLP). Data of ten stocks from different sectors of National Stock Exchange Fifty (NIFTY 50) was considered. Execution time for each model is calculated to find out the fastest algorithm. Finally, correlation between prediction accuracy and performance measures is established. The results clearly show that increasing the training dataset size does not always increase the prediction accuracy. A characteristic of the dataset is one major factor that is responsible for predicting accuracy. DTR and SVR have very low average execution time compared to LSTM and NNMLP. Very strong negative correlation was found between mean absolute percentage error (MAPE) and prediction accuracy.

Keywords: prediction accuracy; training dataset size; rolling dataset; performance measures; regression; neural network; execution time; stock price.

DOI: 10.1504/IJBIDM.2023.133147

International Journal of Business Intelligence and Data Mining, 2023 Vol.23 No.3, pp.201 - 263

Received: 10 Dec 2021
Accepted: 11 Mar 2022
Published online: 01 Sep 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article