Stock market forecasting demands models that balance high accuracy with interpretability, particularly when handling highly volatile and uncertain data.This study introduces a novel interpretable forecasting framework that integrates the Fuzzy Time Series (FTS) model with the Linear Fuzzy Information Granule (LFIG) method.The proposed model addresses two major limitations: the inability of conventional FTS models to effectively capture trend dynamics, and the limited capacity of the LFIG beetroot birkenstock method to account for the influence of recent data.
The core contributions of this work are threefold: 1) a variable-sized interval partitioning technique optimized via fuzzy C-means clustering and the principle of justifiable granularity, achieving adaptive data segmentation that balances coverage and specificity; 2) a trend extraction mechanism based on LFIG approach, which applies time-dependent linear functions within sliding windows to quantify short-term trends and associated uncertainties; and 3) a fusion of FTS and LFIG outputs via the ordered weighted averaging operator, which emphasizes trend-consistent predictions to enhance forecasting accuracy.Experimental evaluation on five benchmark datasets from Yahoo Finance demonstrates that the proposed model outperforms eight state-of-the-art forecasting methods in taylor te400 terms of predictive performance.Furthermore, it maintains interpretability through transparent fuzzy rules and explicit trend representations, providing a robust and explainable framework for stock market forecasting.