Enhancing CNN-LSTM Models for Financial Time-Series Forecasting through Hyperparameter Optimization and Regularization

Saudagar, Rishabh Deepak (2025) Enhancing CNN-LSTM Models for Financial Time-Series Forecasting through Hyperparameter Optimization and Regularization. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (855kB) | Preview
Preview	PDF (Configuration Manual) Download (952kB) | Preview

Abstract

The work in this paper explores the enhancements of the CNN-LSTM hybrid model of predicting the closing prices of the following days of three financial indices, the NSEI of India, the ISEQ of Ireland, and the S&P 500 of America. The focus is on the improvement of the prediction accuracy and generalization through the combination of hyperparameters optimization and over-fitting alleviation approaches. A decade of OHLCV data was used to construct technical indicators such as moving averages, RSI, MACD and volatility on each having a lookback period of 60 days. Each market RMSE and MAE were the main evaluations used to train and test the baseline CNN-LSTM model separately per market. The later hyperparameter optimization with the Keras Tuner showed quite modest results. Comparative performance in terms of markets was explained through visualizations, such as plots of real vs anticipated and bar charts of RMSE/MAE. The USA model performed better and this shows the relevance of the approach in mature markets. Forecasts and results were exported to be analysed within Power BI. The project offers the replicable model of financial forecasting with deep learning.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Khan, Sallar UNSPECIFIED
Uncontrolled Keywords:	CNN-LSTM; Financial Forecasting; Hyperparameter Optimization; Overfitting; Deep Learning
Subjects:	H Social Sciences > HG Finance Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	03 Jul 2026 10:12
Last Modified:	03 Jul 2026 10:12
URI:	https://norma.ncirl.ie/id/eprint/9458

Actions (login required)

View Item