地铁交通流量预测在天池全球城市计算AI挑战赛-A榜222319中获得222319的成绩。

简介：
Subway traffic forecasting using Tianchi—a resource for newcomers and experienced practitioners alike. The Tianchi Global City Computing AI Challenge—Metro Passenger Flow Prediction, achieved Honorable Mention in the A-Ranking for 2022/2023, and this code represents a solution that earned that recognition. If you find this helpful, please give it a star and consider sharing it with others. The team includes buger, taoberica, selina雪, and we extend our gratitude to fish佬 baseline for their valuable contributions. Notably, a portion of the A-Ranking code was inspired by fish佬’s open-source code; however, due to not advancing to the final round, the elimination tournament code has not been released. The dataset can be downloaded via the provided link; the extraction code is ‘arse’. Furthermore, several promising ideas are currently under investigation and are open for experimentation by interested individuals. Specifically, exploring adjustments such as modifying the interval from every ten minutes to every five minutes would significantly increase the volume of data. Additionally, removing the first three days following a ‘shift’ operation could mitigate issues stemming from inaccurate data introduced during that period. Beyond applying the shift strategy to the most recent three days, experimenting with shifting two days prior plus corresponding weekly data could yield further improvements. Initially, we also investigated utilizing LightGBM (lgb) models but found their performance to be inferior compared to XGBoost (xgb), suggesting a potential benefit from employing a blending approach with both models.