An Adaptive Neuro-Fuzzy Inference System (ANFIS) for Modeling and Predicting Teacher Motivation

Abstract

Background: Teacher motivation significantly influences educational quality and student outcomes, especially in resource-constrained settings like Ethiopia, where systemic barriers (resource scarcity, administrative inefficiencies, and large class sizes) intersect with individual psychological drivers. Purpose: This study aimed to identify dominant predictors of overall teacher motivation, quantify their relative importance, and detect pairwise and higher-order interactions using advanced predictive modeling on Ethiopian teacher data. Methods: Cross-sectional data were collected via the Teacher Motivation Scale (TMS) and a demographic questionnaire from teachers across multiple Ethiopian regions. Six models were compared: linear regression, ridge, lasso, random forest, SVR (RBF kernel), and adaptive neuro-fuzzy inference system (ANFIS). Variable importance was derived from permutation methods and neuro-fuzzy rule sensitivity; interactions were visualized through response surfaces, conditional marginal effects, and pairwise strength heatmaps. Novelty: This is among the first applications of neuro-fuzzy inference in Ethiopian teacher motivation research, enabling localized detection of non-linear interactions and rule-specific firing regions that linear and tree-based models cannot capture. Findings: Self-efficacy, administrative support, work conditions, and intrinsic motivation emerged as primary drivers. Strong synergies were found between self-efficacy × work conditions and self-efficacy × intrinsic motivation, peaking when both are high. Antagonistic effects appeared when high-salary satisfaction undermined intrinsic drivers. SVR (RBF kernel) achieved the best accuracy (R² = 0.845, RMSE = 2.71), while ANFIS showed overfitting and poor generalization. Conclusion: Teacher motivation in Ethiopia is predominantly shaped by psychological competence beliefs amplified by supportive organizational contexts. Non-linear interactions highlight the need for holistic, context-sensitive interventions over isolated financial or demographic measures. Recommendations: Prioritize school-level enhancements in administrative support, resource provision, and mastery-oriented professional development to exploit synergistic effects.