基于机器学习的共享单车需求预测

Shared Bicycles Demand Prediction Based on Machine Learning

随着中国城市的发展,共享经济的概念已经逐渐融入人们的生活。共享单车、共享汽车、共享充电宝等新事物正在不断地改变着大家的生活习惯。摩拜单车、美团单车等知名共享单车企业成立以来,已经成为绿色出行、健康环保的代表。然而,共享单车的过度投放、安全隐患、管理混乱等问题也引起了共享单车管理者的关注,同时影响了各共享公司的运营和盈利。本研究根据实时收集到的时间、季节、天气、温度、湿度、风速等数据;构造更加符合共享单车使用场景的离散型变量,并将数据输入到不同的机器学习和深度学习模型,达到准确预测城市中的共享单车需求的目的。通过比较Rsquare、MSE、RMSE等指标来评估所有模型的预测效果。基于模型得出的结果,共享单车管理者能够合理投放相应数量的共享单车,达到减少浪费的目的。

With the development of Chinese cities, the concept of sharing economy has gradually been integrated into people’s lives. New concepts such as shared bicycles, shared cars, and shared power banks are constantly changing everyone’s living habits. Mobike, Meituan and other well-known sharing bicycles companies have become representatives of green travel, health and environmental protection since their establishment. However, problems such as excessive delivery of bicycles, potential safety hazards, and management confusion have aroused the concerns of bicycle-sharing managers, which also affects the operation and profitability of each sharing company. After collecting the data in real time, such as time of day, season, weather, temperature, humidity, and wind speed, this research constructs some new discrete variables which are more in line with the use of bicycle sharing scenarios. After feeding the data into different machine learning and deep learning models, they achieve accurate prediction of bike-sharing demand in cities;also R square, MSE, RMSE and other metrics are used to evaluate prediction effectiveness of all the models. Based on the prediction results, bicycle sharing companies’ managers can deliver the appropriate number of bicycles reasonably and reduce waste.