Empirical Analysis of the Determinants of Mobile Money Adoption in Rural North West Nigeria Using Ensemble Methods

Mobile money adoption in rural northern Nigeria lags significantly behind urban rates, and identifying the specific determinants using advanced machine learning ensemble methods fills a gap in the financial inclusion analytics literature for the region. This study empirically examined the determinants of mobile money account ownership among 1,800 rural households in Jigawa, Sokoto, and Zamfara States using survey data collected through multistage random sampling. Predictor variables including literacy level, distance to agent, network coverage quality, trust in mobile money providers, social network influence, perceived benefit, and gender were assembled. Random forest, gradient boosting, and logistic regression models were trained on an 80-20 split and compared for predictive accuracy and interpretability. Random forest achieved an AUC of 0.87, outperforming logistic regression by 0.11 units. SHAP analysis identified agent proximity within five kilometres, primary school completion, and social network peer adoption as the three strongest positive determinants. Gender showed a significant interaction with trust: low-trust conditions reduced female adoption probability by 53 percent but only 21 percent for males. Model performance was robust across the three states, confirming transferability of findings within the zone. The study fills an empirical gap on gender-trust interaction in northern Nigerian mobile money adoption and recommends geographically targeted agent network expansion, female-specific trust-building communication strategies, and integration of community peer adoption messaging into mobile money onboarding campaigns in the three covered states.

Keywords: mobile money adoption, rural North West Nigeria, ensemble methods, SHAP analysis, financial inclusion