COMPARATIVE ANALYSIS OF TRIANGULAR FUZZY HIDDEN MARKOV MODELS AND TRADITIONAL HIDDEN MARKOV MODELS

Abstract

This study optimizes Traditional Hidden Markov Models (THMMs) using Triangular Fuzzy Membership Functions, resulting in Triangular Fuzzy Hidden Markov Models (TFHMMs) that tolerate ambiguous observations and gradual state transitions in agricultural data prediction. Using oilseed area data from 1992 to 2022, we compare TFHMMs to traditional HMMs, focusing on predicting accuracy using the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Corrected Akaike Information Criterion (AICc), and Hannan-Quinn Information Criterion (HQIC). Our research includes stationary parameters to improve model stability and uses the Viterbi technique to find optimal state sequences, which improves forecast interpretability. The results show that THMMs outperform TFHMMs at capturing the complicated patterns of agricultural data, with lower prediction errors and more reliability. This work emphasizes the potential of fuzzy logic in not improving probabilistic models for agricultural forecasting, providing a more delicate and accurate approach to analyzing and predicting agricultural trends.