Market Trend Prediction and Analysis Using Improved Signal Decomposition and Time Series Algorithm

It is a real-time stock price prediction model that combines the signal decomposition method based on the ICEEMDAN with an existing Long Short-Memory (LSTM) model, which is trained and deployed in a Streamlit web-based application. The Yahoo Finance API is accessed dynamically to obtain historical stock data and CEEMDAN decomposition is applied to the nonlinear stock time series to remove noise and obtain meaningful trend elements. The reconstructed signal is normalized and converted to sequential learning samples to be successfully modeled as time-series. The LSTM model obtains temporal dependencies of the data processed and makes multi-day predictions and historical forecasting based on an auto-regressive prediction method. Experimental analysis shows that the suggested ICEEMDAN-LSTM hybrid model provides high accuracy of prediction (about 94% and low forecasting error (RMSE [?] 0.015), and the hybrid model can also remain stable to trend- following under varying market environments. The visualization module is also integrated such that it allows real-time intuition of real prices, projected prices and forecast trends. All in all, the suggested framework offers a convenient, precise and accessible platform to short-time forecasting of the stock market, providing useful tools in terms of analytical support to the investors, analysts and financial researchers.

[1]  Mutinda, J. K., & Geletu, A. (2025). Stock market index prediction using CEEMDAN‐LSTM‐BPNN decomposition ensemble model. Journal of Applied Mathematics, 2025(1), 7706431. [2]  Lu, Z., Cao, Z., Xiang, Z., Li, J., & Li, M. (2025). Carbon market price prediction in the Yangtze River Basin based on improved deep learning ensemble model with CEEMDAN and Attention‐RNN. Carbon Balance and Management, 20(1), 59. [3]   Liu, Z., & Li, X. (2025). Coal price forecasting using CEEMDAN decomposition and IFOA‐ optimized LSTM model. International Journal of Computational Intelligence Systems, 18(1), 1–20. [4]    Li, Y. (2025). Recent methods in stock price prediction: A review. [5]  Dou, S. Financial time series forecasting based on novel CEEMDAN‐LSTM‐BN network. [6]  Agarwal, S., Sharma, S., Faisal, K. N., & Sharma, R. R. (2025). Time‐series forecasting using SVMD‐ LSTM: A hybrid approach for stock market prediction. Journal of Probability and Statistics, 2025(1), 9464938. [7]  Zhang, Z., Liu, Q., Hu, Y., & Liu, H. (2025). Multi‐ feature stock price prediction by LSTM networks based on VMD and TMFG. Journal of Big Data, 12(1), 74. [8]   Guo, J. (2025). Integration of wavelet transform convolution and channel attention with LSTM for stock price prediction‐based portfolio allocation. arXiv preprint arXiv:2507.01973. [9]  Huang, H., & Cui, K. Exchange rates forecasting and trend analysis from machine learning. [10]  Han, H., Peng, J., Ma, J., Liu, S. L., & Liu, H. (2025). Research on load forecasting based on CEEMDAN, SE, VMD and self‐attention TCN fusion model. Scientific Reports, 15(1), 14530.