Traffic Management Enhancement: A Competitive Machine Learning System for Traffic Condition Classification

In big cities, traffic congestion is a prevalent issue.In order to decide how to manipulate traffic in order to alleviate congestion, traffic regulators, who supervise traffic flow, must conduct an analysis of present conditions.Classifying traffic conditions from road information is a critical step that impacts these decisions.Traffic conditions can be categorized using a variety of techniques, each with benefits and drawbacks of its own.Recently, the rapid development of machine learning techniques has accelerated their use in a variety of sectors, including intelligent transportation systems (ITS).In this study, a competitive machine learning system is introduced to support the decision-making process in ITS, specifically in traffic condition classification.The proposed system operates in two stages: first, identifying the best model configuration from various machine learning methods, and second, deciding through a voting system based on the selected models.The proposed system employs six machine learning methods, each with 4-5 variations in model configurations.The methods tested include Neural Networks, k-Nearest Neighbor, Logistic Regression, Bayesian Networks, Decision Trees, and Random Forests, with individual accuracy rates of 66.2%, 70.5%, 44.4%, 46.1%, 72.2%, and 72.6%, respectively.The models that achieved the highest performance for each method proceed to a voting system, both non-weighted and weighted.The experimental results indicate that the nonweighted system achieved an accuracy of 68.6% to 69.3%, while the weighted system reached 71.9% to 72.5%.The findings show that the proposed competitive machine learning system offers a viable solution for classifying traffic conditions with promising results, especially for implementation in Bandung City, Indonesia.