Dynamic Analysis of Predicted Stock Price Fluctuations Using Machine Learning
Abstract
Abstract This research presents a deep learning–based time series forecasting methodology for short-term stock price prediction, focusing on a 7-day forecast horizon. The study adopts a quantitative, data-driven approach using a supervised learning framework, with the Long Short-Term Memory (LSTM) network as the core predictive model. LSTM is selected for its proven capability to capture long-term temporal dependencies and non-linear patterns in sequential financial data while effectively addressing the vanishing gradient problem associated with traditional Recurrent Neural Networks (RNNs). The forecasting strategy is based on Recursive Multi-Step Prediction, where each predicted time step is iteratively appended to the input sequence to generate subsequent forecasts. The model utilizes a 60-day lookback window and applies Min–Max normalization to one year of historical closing price data to ensure numerical stability. The final 7-day forecast is transformed back to original price values through inverse scaling and aligned with the NSE trading calendar to reflect actual market days. Experimental results demonstrate strong predictive performance, with the proposed model achieving an R² score of 0.95, indicating its effectiveness in capturing short-term stock price dynamics. This study provides a robust and practical framework for short-term stock market forecasting using deep learning techniques. Keywords: Deep Learning; LSTM; Stock Price Forecasting; Time Series Analysis; Recursive Prediction; Financial Data Modeling
Fields of Study & Subjects
Stock Market Forecasting Methods
Forecasting Techniques and Applications
Hydrological Forecasting Using AI
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 31 January 2026
- Language
- EN