A Hybrid Data Mining Framework for Financial Time-Series Prediction

Research Square 2026 Preprint Open access

Read full text
Authors: Ujjwal Pradhan, Sourav Mahapatra, Manjusha Pandey, G V S Narayana
DOI: 10.21203/rs.3.rs-8933577/v1
Cited by: 0
References: 4

Abstract

Abstract Nowadays stock price forecasting is a challenging task because of inherent unpredictability, noise, and non-linear dynamic stock time-series data. This study proposes a hybrid data mining model that integrates wavelet decomposition, LSTM model, and Light gradient Boosting Machine (LightGBM) to increase the accuracy and stability. Wavelet transformation works to disintegrate the original price series into trend and residual components, that enables effective reduction of noises and separation of features.

Fields of Study & Subjects

Stock Market Forecasting Methods Time Series Analysis and Forecasting Forecasting Techniques and Applications

Publication details

Journal / source
Research Square
Publication date
24 February 2026
Language
EN
Document type
Preprint