Predicting Bike Usage and Optimizing Operations at Repair Shops in Bike Sharing Systems

Supply chain responsiveness and big data analytics (BDA) have garnered considerable interest in academia and among practitioners. BDA helps researchers understand the current challenges in data management, including the high volume, velocity, and variety of data. This study is concerned with improving the responsiveness of supply chain networks to bike-sharing systems (BSS), which exhibit BDA characteristics. To address the challenges of forecasting bike usage and accordingly optimizing repair shop operations, we analyze multi-factor BSS data (Data from Washington D.C. BSS available to public), wherein attributes, such as weather conditions, registration, humidity, date, and time, are present. We use machine learning algorithms, such as neural networks, decision-tree-based regression, K-nearest neighbor, support vectors, and ensemble random forest, to predict bike usage and repair. This work contests the results and demonstrates the effectiveness of combining machine learning with supply chain network design. Supply chain networks model bike repairs by means of capacity extensions, which entails a nonlinear problem. In this study, we utilize a gradient search to solve a nonlinear supply chain network model. By enabling capacity extension, bike repair shops within the BSS exhibit a promising 50 % reduction in lead repair time. Furthermore, a 25 % overall throughput increase in BSS is achieved. Ultimately, this study demonstrates the importance of operational flexibility in responding to big data challenges.

Automated summary

This study focuses on enhancing the responsiveness of supply chain networks to bike-sharing systems by analyzing multi-factor bike-sharing data and using machine learning algorithms to predict bike usage and repair. The results demonstrate the effectiveness of combining machine learning with supply chain network design in improving operational efficiency.