ADVANCED HYBRID FUZZY–SOFT SET BASED RISK ASSESSMENT MODEL

Abstract

This study proposes a hybrid Fuzzy-Soft Set model that combines Mamdani–Sugeno fuzzy inference systems with Random Forest machine learning and the PSOGA metaheuristic optimisation to manage the inherent uncertainty in disease predictions for gastric and prostate cancer risk assessments. The model presents fuzzy-soft hybrids tailored for cancer parameter approximations, and PSOGA optimised rule bases to enhance inference accuracy, as well as extensions through fuzzy-soft-TOPSIS for the multi-criteria ranking of tumour risk. These methods demonstrate superior performance over classical Mamdani-Sugeno baselines with accuracy boosts of up to 15%, enhanced interpretability and robustness. Validated over multi-hospital oncology data sets (n > 1000 cases), the framework involves advanced preprocessing of heterogeneous clinical parameters, hyperparameter tuning of uncertain scenarios and robust metrics including sensitivity analysis, AUC-ROC metrics and decision boundary visualisation. Theoretical proofs show the completeness and stability of the framework for oncology applications. It also gives the missing links in the literature on hybrid fuzzy-ML cancer prediction models.