Using a Machine Learning Approach to Implement and Evaluate Product Line Features
This work addresses the need for predictive features in bike-sharing systems to improve urban mobility, but it is incremental as it applies existing machine learning methods to a specific domain.
The paper tackled the problem of predicting bike-sharing system states by using machine learning to analyze usage patterns and learn computational models from system logs, resulting in a principled way to assess the cost-performance trade-off of features before deployment.
Bike-sharing systems are a means of smart transportation in urban environments with the benefit of a positive impact on urban mobility. In this paper we are interested in studying and modeling the behavior of features that permit the end user to access, with her/his web browser, the status of the Bike-Sharing system. In particular, we address features able to make a prediction on the system state. We propose to use a machine learning approach to analyze usage patterns and learn computational models of such features from logs of system usage. On the one hand, machine learning methodologies provide a powerful and general means to implement a wide choice of predictive features. On the other hand, trained machine learning models are provided with a measure of predictive performance that can be used as a metric to assess the cost-performance trade-off of the feature. This provides a principled way to assess the runtime behavior of different components before putting them into operation.