Traditional fleet maintenance follows a scheduled servicing with consequent unnecessary repairs or un-detected failures. In this research, the Rental Wheelz fleet management system is enhanced with the implementation of a pre-captured data predictive model for maintenance. For the lack of real-time sensor data, history of these vehicle parameters such as engine temperature, fuel efficiency, brake life, and transmission operation was utilized. Preprocessing and cleaning of the dataset were carried out prior to training and testing the model. Among all the algorithms experimented with, a random forest model performed 88% accurately, forecasting failures up to 17 days ahead of time. The model was then incorporated into the system so that manual input of data would lead to maintenance suggestions. The system offers daily risk ratings, making it possible for fleet managers to schedule high-priority vehicle checks. Despite its drawbacks of relying on human input and the absence of live sensor readings, the solution has changed significant cost savings and improved customer satisfaction through the reduction of unexpected breakdowns. The process can be expanded to rental and logistics sectors for greater efficiency.