From Digital Grievance to Actionable Insights: Leveraging Machine Learning Sentiment Classification to Enhance Operational Efficiency among Lagos State Power Consumers

ABSTRACT: Nigeria's electricity distribution companies (DisCos), particularly those operating within Lagos State, are confronted with an unprecedented and chaotic influx of customer grievances channelled through social media platforms, mobile applications, and web-based customer portals. Manual triage of these unstructured complaints — spanning billing discrepancies, estimated billing, power outages, and prepaid meter faults — has produced severe processing backlogs, protracted resolution latency, and eroded consumer trust. This study examines the extent to which Machine Learning Sentiment Classification, Automated Text Analytics, and Real-Time AI Emotion Detection Systems (independent variables) influence Operational Efficiency, Consumer Dissatisfaction Polarity, Customer Trust, and Complaint Resolution Latency (dependent variables) among Lagos State power consumers. Employing a mixed-methods, longitudinal panel design anchored on secondary and primary data spanning 2013 to 2025 and drawn from Nigerian Electricity Regulatory Commission (NERC) records, DisCo customer service logs, and social media analytics, alongside supervised machine learning classifiers (Support Vector Machines, Naïve Bayes, and transformer-based BERT models), the study finds that sentiment-driven triage systems are associated with a reduction in average resolution latency of over 40 percent and a statistically significant improvement in Service Level Agreement (SLA) compliance. The findings offer actionable, evidence-based recommendations for DisCo management, regulators, and technology partners seeking to transition from reactive complaint management toward proactive, data-driven customer experience governance.