Perpustakaan Digital ITB

Traffic management is often operated to solve traffic jam which caused by the vehicle volume and the road capacity. But, currently, the traffic management still rely on a subjective information from specific stakeholder who manages day by day. The observation process is done by first-person view. By this condition, the management will be a subjective model. Besides, the world has envolved day by day and gives us chances to apply a good technology for our problems. Because of the opportunity and threats, there are chances to get an objective information by applying machine learning. In this final project, the algorithm that being used is Artificial Neural Network – Multilayer Perceptron and Random Forest. The applied methodology in the final project is Cross-Industry Standard Process for Data Mining which consists of business understanding, data understanding, data preparation, modelling, and evaluation. In data preparation process, oversampling is being applied to data collection. In parameter adjusment and modelling process, grid-search and k-fold cross validation is being applied to the process. In evaluation process, there are 3 metrics for measurement. They are consists of recall, precision, and F1-score. The data collection which being examined in this project is Australia data in 2014-2017. The project shows that Multilayer Perceptron is the best model with 0.997232 for recall score, 0.998974 for precision score, and 0.998100 for F1-Score.