The number of shared bicycles is large and the parking is not standardized. Accurate position prediction of the bicycle can effectively prevent the collision between the number of vehicles and the parking position. The traditional machine learning algorithm is slowly to process and can not meet the accuracy requirements. Therefore, the Boosting algorithm which promote the weak learner to a strong learner is adopted to enhance running speed and accuracy. At the same time, the way of constructing feature groups is proposed, and the importance of features is evaluated. The comparison experiment compares the prediction capabilities of the three Boosting algorithms, XGboost, LightGBM and Catboost. The example verification shows that the LightGBM algorithm has the highest test accuracy, and then this algorithm is used to predict the dynamic distribution of the Mobike in Beijing throughout the day. And show the heat distribution map of the prediction results of Mobike in different time periods.
[1]
Tie-Yan Liu,et al.
LightGBM: A Highly Efficient Gradient Boosting Decision Tree
,
2017,
NIPS.
[2]
Andy Liaw,et al.
Extreme Gradient Boosting as a Method for Quantitative Structure-Activity Relationships
,
2016,
J. Chem. Inf. Model..
[3]
Zhenzhong Li,et al.
The application of factorization machines in user behavior prediction
,
2016,
2016 IEEE/ACIS 15th International Conference on Computer and Information Science (ICIS).
[4]
Gaël Varoquaux,et al.
Scikit-learn: Machine Learning in Python
,
2011,
J. Mach. Learn. Res..
[5]
Yu JIANG,et al.
Short-term wind power forecasting using hybrid method based on enhanced boosting algorithm
,
2017
.
[6]
Cho-Jui Hsieh,et al.
GPU-acceleration for Large-scale Tree Boosting
,
2017,
ArXiv.