文件列表:

.DS_Store (6148, 2020-10-06)
ARIMA.ipynb (9462, 2020-10-06)
Bilstm_multivar.ipynb (99539, 2020-10-06)
LICENSE (1064, 2020-10-06)
MD_pic (0, 2020-10-06)
MD_pic\1hour_nox.png (45296, 2020-10-06)
MD_pic\96hours_nox.png (21811, 2020-10-06)
MD_pic\location.jpg (185381, 2020-10-06)
MD_pic\results_nox.png (103786, 2020-10-06)
MD_pic\time_window.jpg (344390, 2020-10-06)
convlstm_multivar_sites.ipynb (506158, 2020-10-06)
data (0, 2020-10-06)
data\Bloomsbury.csv (2143228, 2020-10-06)
data\Bloomsbury_clean.csv (3330544, 2020-10-06)
data\Eltham.csv (1712714, 2020-10-06)
data\Eltham_clean.csv (3265074, 2020-10-06)
data\Harlington.csv (1781386, 2020-10-06)
data\Harlington_clean.csv (3310156, 2020-10-06)
data\Marylebone_Road.csv (2424629, 2020-10-06)
data\Marylebone_Road_clean.csv (3404958, 2020-10-06)
data\N_Kensington.csv (2113631, 2020-10-06)
data\N_Kensington_clean.csv (3368373, 2020-10-06)
data_process.ipynb (1926912, 2020-10-06)
lightGBM_multivar_single_sites.ipynb (482004, 2020-10-06)
lstmWithAttention_multivar_sites.ipynb (403364, 2020-10-06)
lstm_multivar.ipynb (478924, 2020-10-06)
lstm_multivar_sites.ipynb (567486, 2020-10-06)
lstm_singvar.ipynb (761111, 2020-10-06)
pic (0, 2020-10-06)
pic\Bloomsbury_windRose.png (121443, 2020-10-06)
pic\Eltham_windRose.png (122762, 2020-10-06)
pic\Harlington_windRose.png (121334, 2020-10-06)
pic\Marylebone_Road_windRose.png (121858, 2020-10-06)
pic\N_Kensington_windRose.png (127530, 2020-10-06)
pic\map.jpg (123092, 2020-10-06)
... ...

### English version README can be seen here [English version](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/README_en.md) # air_pollutants_prediction_lstm 这个项目是用于预测伦敦空气质量的状况。其中有五个监测站的数据被选用。这五个监测站分别是:Harlington, North Kensington, Marylebone, Bloomsbury and Eltham. 该数据来源为openair开源库。数据的时间跨度为2018-2019。数据的属性有:NOX, NO2, NO, O3, PM2.5, 风速, 风向和空气温度。因为是用于水毕业的论文,所以整个实验并没有采用面向对象的方式进行封装,所以比较杂乱,见谅。但这样也有好处,就是任何单独的一个Jupter notebook文件 都可以独立的运行实验,并且完成实验的结果。  # 运行环境 - google colab (建议) or - pytorch - numpy - matplotlib - seaborn # 使用算法 在使用算法之前,先要对数据进行[预处理](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/data_process.ipynb)并且理解他们。 - [lstm (单变量)](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/lstm_singvar.ipynb) - [lstm (单站点, 多变量, 共8变量)](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/lstm_multivar.ipynb) - [lstm (多站点, 多变量, 共40变量)](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/lstm_multivar_sites.ipynb) - [BiLSTM](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/Bilstm_multivar.ipynb) - [ConvLSTM](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/convlstm_multivar_sites.ipynb) - [LSTM + Attention](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/lstmWithAttention_multivar_sites.ipynb) - [LightGBM](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/lightGBM_multivar_single_sites.ipynb) - [ARIMA](https://github.com/RobinLuoNanjing/air_pollutants_prediction_lstm/blob/master/ARIMA.ipynb) > 注意: 除了BiLSTM以外,其余算法全都采用双层的全连接层。 # 介绍 ## 数据集划分 80% 训练集, 10% 验证集,10%测试集。 ## 评估方法 注意!该实验的评估方法有两种。一种是用10%的测试集去评测。这意味着模型只需要预测未来1小时的情况。然后预测10%个相应的1小时。预测出来的值是不会重新放回time window里的。 另外一种,就是用上面筛选出来的模型去预测未来96小时的情况。每预测1小时,预测出来的新的值,将会塞回时间窗口。所以这个过程,偏移将会越来越大。  ## 结果 下面是预测Bloomsbury未来96小时的NOX含量的结果  ### 致谢 - [Openair](https://davidcarslaw.github.io/openair/) - [ndrplz](https://github.com/ndrplz/ConvLSTM_pytorch)

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：