Research And Application Of Demand Forecasting And Scheduling Optimization Algorithm For Shared Bicycles

As the third major way of public transportation,shared bikes provide travelers with more transportation options.Its emergence not only solves the pain point of users' short-distance travel but also conforms to the new urban development concept of innovation,coordination,green,open,and sharing.However,due to the large-scale growth of bike-sharing enterprises,the lack of management experience leads to a series of difficulties in the development of bike-sharing.To this end,this thesis designs and implements a prototype system of intelligent dispatching for the management department of shared bikes.By referring to the research theories of traditional public bike dispatching management,the thesis aims at minimizing the dispatching cost and forms a solution idea from demand analysis to demand prediction,and from region division to region scheduling.The main work of this thesis is as follows:(1)A Bike-sharing Demand Forecasting Algorithm Based On Improved Random Forests(IMRF-DF)is proposed to solve the prediction problem of the management department of bike-sharing on the number of vehicles launched at the station.By analyzing the influence of travel characteristics such as time,weather,temperature,and windspeed on the demand,a bike-sharing demand prediction model was established.In the process of solving the model,the weighted idea is introduced,and Pearson Correlation Coefficient Algorithm is used to improve the feature selection method,increase the feature selection diversity,and improve the prediction accuracy of the algorithm.The principle of feature interval division is adopted to enhance the optimum of feature subset and avoid the algorithm falling into local optimum.Experimental comparisons on CitiBike data sets show that the IMRF-DF algorithm improves the prediction accuracy by about 5.13% compared with the RF algorithm,which proves that the improved algorithm has a great improvement in the prediction of bike-sharing demand.(2)A Bike-sharing Scheduling Path Optimization Algorithm Based On Improved Ant Colony Optimization Algorithm(IMACO-SP)is proposed to solve the optimization problem of scheduling path selection.Considering the scheduling path conflict caused by traditional manual vehicle scheduling and the problem of solving the optimal scheduling path,a bike-sharing scheduling optimization model with the goal of minimizing the total scheduling cost is constructed.In the process of solving the model,the regional division principle of K-means and associated attributes is adopted to transform the scheduling problem from global optimum to local optimum.The ACO algorithm's pheromone update principle is improved,and the net scheduling demand is referred to as the pheromone parameter,so as to avoid the scheduling paths falling into local optimum due to the rapid volatilization of pheromone.Experimental results on CitiBike data sets indicate that the IMACO-SP algorithm has a significant improvement in solution accuracy compared with the ACO algorithm.The global optimal solution is optimized by about 3.69%,the global average solution is optimized by about 4.60%,the mean deviation of the solution is reduced by about 0.9%,and the maximum deviation is reduced by about 0.49%.The improved algorithm not only improves the solving rate of the algorithm but also improves the robustness of the algorithm.(3)Finally,the prototype system of intelligent dispatching of shared bikes is designed and implemented.This prototype system is based on B/S architecture,using Java as the main development language,MySQL as the database.The prototype system can realize such functions as scheduling demand forecasting,scheduling area division,and scheduling path generation for shared bicycles,and visually display the results.