Anomaly Detection in Traffic: A Data Analysis Approach

This study explores the performance of multiple machine learning models in detecting unusual events within traffic datasets. The methods tested include Isolation Forest, One-Class SVM, DBSCAN, Autoencoder, and K-Means. Among these, DBSCAN achieved the highest accuracy at 98.8%, with K-Means closely following at 95.7% and the Autoencoder at 95.0%. The other two models, The Isolation Forest achieved an accuracy of 90%, while the One-Class SVM slightly outperformed it with 92.5%, showing consistent reliability in anomaly detection. To provide a clear comparison, the outcomes were illustrated using visual plots and a summary bar chart. Overall, the results demonstrate that while each technique has its strengths, the choice of algorithm should be guided by the type of anomaly being analyzed in traffic management systems.