Research on Optimization of Shared Bicycle Scheduling Based on Genetic Algorithm and LSTM

In this paper, a dynamic scheduling algorithm based on multi-objective optimization is designed and illustrated with the example of shared bicycle scheduling optimization. Firstly, this research conducted a spatio-temporal analysis based on the publicly available dataset of Citi Bike in New York, and by applying the K-means clustering algorithm, user behaviors were classified into three major categories with significant differences in the frequency of use and riding paths. On this basis, this paper proposes and constructs a demand prediction model based on LSTM (Long Short-Term Memory Network) introducing the attention mechanism. Finally, this paper innovatively combines the global search ability of genetic algorithm and the local optimal search advantage of dynamic programming to design a dynamic scheduling algorithm based on multi-objective optimization. By introducing an adaptive cross-variance strategy, the algorithm not only significantly improves the convergence speed, but also enhances the adaptability to complex demand patterns. The results show that the dynamic scheduling scheme proposed in this paper reduces the average waiting time of users by about 35% and improves the turnover rate of a single vehicle by about 30% compared with the traditional static scheduling strategy. Especially during peak hours, the scheduling scheme greatly alleviates the bicycle shortage problem and significantly improves the overall efficiency of the system and user satisfaction.