多特征下LSTM模型的实现
    在长短时记忆网络中,会添加各时段的特征信息,通过构建多变量因素条件下的预测模型,可以预测客流分布的时空相关性,在模型的输入特征中使用时间序列的特征,如前15分钟的客流量,前30分钟的客流,前一天该时段的客流量等信息,通过不同时间粒度的采集,在深度学习中会有更好的表现效果。
    时间序列特征:
    前15分钟客流量,前30分钟客流量,前45分钟客流量,前60分钟客流量,前一天客流量,前一周同时间段客流量,前两周同时间段客流量。
    空间序列特征:
    天气类型,最高气温,最低气温,站点信息,站点类型,日期类型,节假日类型。

    http://course.sdu.edu.cn/G2S/Template/View.aspx?courseId=2593&topMenuId=183267&action=view&type=&name=&menuType=1

    3D25C270-A4BA-4276-8834-EED49493E657.png

    https://tianchi.aliyun.com/competition/entrance/231641/introduction

    https://zhuanlan.zhihu.com/p/67832773

    https://www.datacamp.com/tutorial/tutorial-time-series-forecasting

    Time Series Forecasting Methods
    Time series forecasting can broadly be categorized into the following categories:

    Classical / Statistical Models — Moving Averages, Exponential Smoothing, ARIMA, SARIMA
    Machine Learning — Linear Regression, XGBoost, Random Forest, or any ML model with reduction methods
    Deep Learning — RNN, LSTM

    https://www.kaggle.com/learn/time-series 传统的
    https://www.kaggle.com/code/thebrownviking20/everything-you-can-do-with-a-time-series/notebook

    https://blog.csdn.net/weixin_46649052/article/details/115406977

    https://www.math.pku.edu.cn/teachers/lidf/course/fts/ftsnotes/html/_ftsnotes/ftsnotes.pdf

    https://zhuanlan.zhihu.com/p/22113312

    https://uqer.datayes.com/v3/community/share/5790a091228e5b90cda2e2ea

    https://otexts.com/fppcn/holt-winters.html

    https://zhuanlan.zhihu.com/p/421710621

    image.png

    deepAR
    https://zhuanlan.zhihu.com/p/348889806

    https://www.kaggle.com/code/satishgunjal/tutorial-time-series-analysis-and-forecasting/notebook

    https://github.com/jiwidi/time-series-forecasting-with-python

    https://github.com/ChuanyuXue/The-Purchase-and-Redemption-Forecast-Challenge-baseline

    金融时间序列数据预测_文献回顾与展望.pdf