# times_series

**Repository Path**: xiwanggit/times_series

## Basic Information

- **Project Name**: times_series
- **Description**: 时序建模：对铁路货运量/客运量进行时序建模预测
- **Primary Language**: Python
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 12
- **Forks**: 5
- **Created**: 2020-07-01
- **Last Updated**: 2025-04-16

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

*  目的：对铁路货运量进行时序建模预测
*  code
   *  Python: 3.7
      *  run.py：程序入口
      *  model_evalute.py：模型评估
      *  algorithm
         *  line_model.py：简单线性模型
         *  stl_model.py：STL模型
         *  holt_winters_damp_model.py：holt-winter 阻尼季节性模型
         *  arma_arima_model.py：ARMA & ARIMA模型
   *  R: 4.0.2
      *  run.R：程序入口
*  package prophet(by facebook): support R & Python. 
   *  [Github](https://github.com/facebook/prophet)
   *  [API](https://facebook.github.io/prophet/docs/quick_start.html)
*  理论
   *  [王燕 应用时间序列分析 第三版](https://gitee.com/xiwanggit/book-math/blob/master/note/%E5%BA%94%E7%94%A8%E6%97%B6%E9%97%B4%E5%BA%8F%E5%88%97%E5%88%86%E6%9E%90.md)
   *  [预测：方法与实战](https://gitee.com/xiwanggit/book-math/blob/master/note/%E9%A2%84%E6%B5%8B%EF%BC%9A%E6%96%B9%E6%B3%95%E4%B8%8E%E5%AE%9E%E6%88%98.md)
   *  [note](https://gitee.com/xiwanggit/times_series/blob/master/note.md)

*  过程
   *  数据探索：铁路货运量数据

![输入图片说明](https://images.gitee.com/uploads/images/2020/0720/171744_b415285e_379161.png "数据探索.png")

   *  简单线性回归

![输入图片说明](https://images.gitee.com/uploads/images/2020/0720/171812_44e4cae5_379161.png "简单线性回归.png")

   *  时序分解：STL模型:看起来还行，但残差并不是白噪音

![输入图片说明](https://images.gitee.com/uploads/images/2020/0720/171846_086d8301_379161.png "STL模型.png")

   *  指数平滑：Holt-Winters阻尼趋势模型:同STL

![输入图片说明](https://images.gitee.com/uploads/images/2020/0720/172000_e6714340_379161.png "Holt-Winters阻尼季节性模型.png")

   *  ARMA模型
      *  二次差分

![输入图片说明](https://images.gitee.com/uploads/images/2020/0720/172120_7645f1d5_379161.png "二次差分ARMA模型.png")


      *  月度差分（diff(12)）


![输入图片说明](https://images.gitee.com/uploads/images/2020/0720/172130_3f4daa1d_379161.png "月度差分ARMA模型.png")

   *  非季节性ARIMA模型与二次差分模型极度相似
   *  季节性ARIMA模型

![输入图片说明](https://images.gitee.com/uploads/images/2020/0729/133501_96d25f8f_379161.png "SARIMA模型.png")

   *  一层LSTM

![输入图片说明](https://images.gitee.com/uploads/images/2020/1006/075903_41d5197b_379161.png "一层LSTM.png")

   *  一层GRU

![输入图片说明](https://images.gitee.com/uploads/images/2020/1006/075915_a501bc91_379161.png "一层GRU.png")
   
*  评估指标

| 序号  | 模型  | mae  |  mape | rmse  |
|---|---|---|---|---|
| 1  | STL  | 1708.587890625  |  4.75034999847412 | 1843.4534912109375  |
| 2  | Holt-Winters阻尼趋势  | 1760.7532958984375  |  4.894587993621826 | 1995.978759765625  |
| 3  | ARMA-二次差分  | 2923.4296485428586  |   | 3265.973788764896  |
| 4  | ARMA-月度差分  | 1398.5633840690107  |   | 1656.7750122665723  |
| 5  | 季节性ARIMA  | 1172.88818359375  | 3.175144910812378  | 1496.1046142578125  |
| 6  | LSTM  | 2685.266845703125  | 7.491035461425781  | 3051.925537109375  |
| 7  | GRU  | 2669.312255859375  | 7.446711540222168  | 3031.5517578125  |