ML-Driven Multifaceted Classification Framework for Transport Service Evaluation

This study presents a multifaceted classification framework using machine learning (ML) to evaluate transport services. The framework integrates feedback analysis, sensor-based monitoring, and travel time evaluation to provide a comprehensive assessment. The system employs LSTM (Long Short-Term Memory) for sentiment analysis, categorizing feedback into Positive, Neutral, and Negative sentiments. Simultaneously, sensor data from GP2YI0I0AU0F (dust), ADXL345 (acceleration), and MQ135 (odor) undergo anomaly detection using autoencoders, classifying conditions from Excellent to Very Poor based on predefined thresholds. Travel time deviations are evaluated using autoencoders, categorizing travel experiences from Excellent (minimal deviation) to Very Poor (significant deviation). These strategies are integrated through a unified classification function, combining LSTM sentiment analysis, Random Forest rating prediction, and autoencoder-based categorizations. This approach enables actionable insights to enhance service quality through comprehensive feedback analysis, sensor monitoring, and evaluation of travel time. This ML-based framework leverages advanced techniques to improve transport service assessment, supporting informed decision-making and operational enhancements.