Bike sharing systems, as one of the complementary modes for public transit networks, are designed to help travelers in traversing the first/last mile of their trips. Different factors such as accessibility, availability, and fares influence these systems. The availability of bikes at certain times and locations is studied under rebalancing problem. The paper proposes a bottom-up cluster-based model to solve the static rebalancing problem in bike sharing systems. First, the spatial and temporal patterns of bike sharing trips in the network are investigated. Second, a similarity measure based on the trips between stations is defined to discover groups of correlated stations, using a hierarchical agglomerative clustering method. Third, two levels for rebalancing are assumed as intra-clusters and inter-clusters with the aim of keeping the balance of the network at the beginning of days. The intra-cluster level keeps the balance of bike distribution inside each cluster, and the inter-cluster level connects different clusters in order to keep the balance between the clusters. Finally, rebalancing tours are optimized according to the positive or negative balance at both levels of the intra-clusters and inter-clusters using a single objective genetic algorithm. The rebalancing problem is modeled as an optimization problem, which aims to minimize the tour length. The proposed model is implemented in one week of bike sharing trip data set in Chicago, USA. Outcomes of the model are validated for two subsequent weekdays. Analyses show that the proposed model can reduce the length of the rebalancing tour by 30%.
