Market Trend Prediction and Analysis Using Improved Signal Decomposition and Time Series Algorithm
INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT · 10.04.2026 · 10 (04) · 1-9
DOI: 10.55041/ijsrem59744 ↗Kurz gefasst
Fachartikel in der Zeitschrift INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT, online erschienen am 10.04.2026. Verfasst von Kanish T und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Abstract- Prediction of stock price is an important issue that needs to be solved in financial markets because stock time-series data is very volatile, noisy and non-linear. Conventional forecasting methods are not always good at capturing the complexity in the market, which restricts their usefulness in real-time financial decision making. In response to this issue, this paper will propose an intelligent stock price prediction system that will combine both hybrid signal decomposition and deep learning with the environment of a web application based on Streamlit. System uses the Yahoo Finance Application Programming Interface (api) to retrieve the historical stock market data using the close price values to estimate the forecasting activities. Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) approach is implemented to break down the stock time series into several intrinsic mode functions (IMFs) to enhance the signal quality and minimize noise in the financial data. Chosen intrinsic features are reassembled and input them into a Long Short-Term Memory (LSTM) neural network model. Min-Max normalization is used to perform the preprocessing of the data and a sliding window mechanism is used to discover temporal dependencies in sequence financial data. The LSTM model is a pre-trained model that predicts the history as well as the future price of the stock using an auto-regressive forecasting technique. The combination of CEEMDAN and LSTM is experimentally tested and exhibits superior predictive status with about 93-95 percent forecasting results with an F1-score of 0.93, precisions of 0.92, and recall rates of 0.91 along with low levels of prediction error (RMSE [?] 0.015]. The system enables a dynamic user entry such as symbol of stock ticker, date range, lookback window, and prediction horizon and the results are displayed in interactive visualisation of real price, forecast of future trends and predictions. The suggested structure is a web-enabled efficient algorithm that integrates signal processing with deep-learning algorithms to enable the accuracy and real-time prediction of the stock market. Keywords: Stock Price Prediction, LSTM, ICEEMDAN, CEEMDAN, Time Series Forecasting, Deep Learning, Streamlit Application, Yahoo Finance, MinMaxScaler, Financial Data Analysis
Zitieren
APA
T, K., R, F., & Ramya, D. R. (2026). Market Trend Prediction and Analysis Using Improved Signal Decomposition and Time Series Algorithm. INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT, 10(04), 1-9. https://doi.org/10.55041/ijsrem59744
Vancouver
T K, R F, Ramya DR. Market Trend Prediction and Analysis Using Improved Signal Decomposition and Time Series Algorithm. INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT. 2026;10(04):1-9. doi:10.55041/ijsrem59744
BibTeX
@article{T2026,
author = {T, Kanish and R, Fazil and Ramya, Dr. R.},
title = {Market Trend Prediction and Analysis Using Improved Signal Decomposition and Time Series Algorithm},
journal = {INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT},
year = {2026},
volume = {10},
number = {04},
pages = {1--9},
doi = {10.55041/ijsrem59744}
}
EUVIMED