Vehicle Speed Profile Classification Using Machine Learning

Efficient identification of traffic states is a critical requirement for modern transport monitoring systems.Instead of relying on direct speed sensor measurements, this study investigates whether structured vehiclecount observations can be used to infer traffic speed profiles and classify congestion levels.A complete supervised learning pipeline is designed using vehicle category counts as predictive variables.Data preparation includes attribute filtering, statistical imputation, encoding of categorical labels, normalization, and controlled dataset partitioning.Six classification algorithms are evaluated under identical preprocessing conditions to ensure fairness in comparison.Experimental findings demonstrate that ensemble-based learners provide superior discrimination capability compared to linear and proximity-based models.The proposed framework is lightweight, reproducible, and suitable for integration into intelligent traffic monitoring systems where only structured count data is available.