A Hybrid Machine Learning and Optimization Model to Minimize the Total Cost of BRT Brake Components

Public transport is amongst critical infrastructures in modern cities, especially megacities, home to millions of people. The reliability of these systems is highly crucial for both citizens and service providers. If service providers overlook system reliability, a considerable amount of expenses will be wasted. Several factors such as vehicle failure, accident, lack of budget weather factors, and traffic congestion cause unreliability, among which vehicle failure plays a prominent role. The brake system is the most vulnerable and vital component of a public transportation bus. Brake reliability depends on driver’s expertise, component quality, passenger loading, line situation, etc. Driver’s expertise and components’ quality are the most important factors for brake system reliability. This study aims to implement a hybrid machine learning and optimization model to minimize the total investment and reliability-related costs in a bus rapid transit (BRT) system. A regression analysis method is proposed to capture the main attributes of a joint brake system, including the level of education, training, and drivers’ experience. The failure rate is modeled as a linear function of ETE and the quality of brake system subcomponents using a Lasso regression model. MILP optimization is then provided for optimizing the total expected costs for a bus rapid transit (BRT) system. Furthermore, a practical case is studied to investigate whether this optimization can reduce costs. The results confirm the efficiency of the hybrid optimization approach.