Abstract: Implementations include providing assistance services, and more specifically for selecting a roadside assistance provider from a plurality of providers using one or more predictive models of aspects of a roadside assistance request. Selection of the roadside assistance provider may be based on ranking of roadside assistance providers associated with a service area, the ranking based on one or more predicted output values from one or more predictive models of aspects of a roadside assistance service, such as estimated arrival time and an estimated probability of acceptance of the request by a roadside assistance provider. One or more of the predicted output values may be adjusted based on a boost value as determined boost value predictive model to increase a likelihood that a selected roadside assistance provider accepts a request to provide the assistance.

Abstract: Implementations include providing assistance services, and more specifically for selecting a roadside assistance provider from a plurality of providers using one or more predictive models of aspects of a roadside assistance request. Selection of the roadside assistance provider may be based on ranking of roadside assistance providers associated with a service area, the ranking based on one or more predicted output values from one or more predictive models of aspects of a roadside assistance service, such as estimated arrival time and an estimated probability of acceptance of the request by a roadside assistance provider. The predictive models or other components of the roadside assistance system may be machine-learning and adaptable based on historical data and feedback information of responses and rankings to prior roadside assistance requests.