Diagnóstico Difuso da Rentabilidade Empresarial: Modelo Mamdani Aplicado às Margens Bruta e Operacional

Abstract

This article proposes and validates a Mamdani fuzzy-inference model to diagnose corporate profitability for Coelmatic Indústria de Componentes Eletrônicos Ltda., using 2024 income-statement data. Four input variables—General and Administrative Expenses (DGA), Tax Debt Ratio (DT), Financial Debt Ratio (DF) and Financial Return (RF)—were fuzzified into Low, Medium and High through trapezoidal membership functions calibrated on historical quartiles. An 81-rule IF–THEN base, generated by the Wang-Mendel algorithm and refined by domain experts, feeds the Mamdani core; centroid defuzzification yields the crisp score CompLucroO on a 0–30 scale. The model achieved a mean absolute error below 8 % against accounting ROE and a Kendall’s W of 0.85 versus analysts’ rankings, supporting the hypothesis that fuzzy logic captures performance nuances overlooked by crisp metrics. Limitations include single-year scope, subjective parameter tuning and exclusion of macroeconomic factors. Future work should extend the time series, explore type-2 or neuro-fuzzy systems, and incorporate ESG indicators.