Title: An optimised CNN-stacked LSTM neural network model for predicting stock market time-series data

Authors: Kalva Sudhakar; Satuluri Naganjaneyulu

Addresses: Department of Computer Science and Engineering, Jawaharlal Nehru Technological University Kakinada, Kakinada, India; CSE Department, G Pulla Reddy Engineering College (Autonomous), Kurnool, A.P, India ' Department of Information Technology, LakiReddy Bali Reddy College of Engineering (Autonomous), NTR District, Mylavaram AP‑521230, India

Abstract: Stock market analysis and prediction are crucial for understanding business ownership and financial performance, this study proposes an optimised CNN-stacked LSTM neural network model for accurate stock market trend prediction. The initial challenge lies in designing a customised CNN-stacked LSTM model for stock data prediction due to the abundance of non-optimised algorithms. To address this, we conducted training and testing using diverse datasets, including NYSE, NASDAQ, and NIFTY-50, observing variations in model accuracy based on the dataset. Remarkably, our model demonstrated exceptional performance with the NIFTY-50 dataset, accurately predicting up to 99% of stocks even during the testing phase. Throughout training and validation, we measured mean squared error (MSE) values ranging from 0.001 to 0.05 and 0.002 to 0.1, depending on the dataset. Our proposed CNN-stacked LSTM model presents a promising solution for accurate prediction of stock market trends, addressing the limitations of previous methods.

Keywords: stock market prediction; CNN-stacked LSTM model; time-series data; NYSE; NASDAQ; NIFTY; mean squared error; MSE; mean absolute error; MAE.

DOI: 10.1504/IJCEE.2025.145022

International Journal of Computational Economics and Econometrics, 2025 Vol.15 No.1/2, pp.196 - 224

Received: 06 Dec 2023
Accepted: 25 Nov 2024
Published online: 17 Mar 2025 *