组合学习模型

1.使用lasso回归进行特征选择

《基于Ｌａｓｓｏ和ＢＰ神经网络的组合预测及其应用———以居民消费支出预测为例》

*为了消除各变量之间的量纲的影响，且比较容易得到平稳序列，需要对部分数据进行对数处理。

*单变量神经网络，滚动预测法疑问：神经网络（机器学习算法）在怎么利用多变量数据预测未来值？

Ensemble learning

组合预测

2.将时间序列预测转化成有监督学习（非常好的几篇博客）

https://machinelearningmastery.com/start-here/#process

https://machinelearningmastery.com/time-series-forecasting-supervised-learning/

https://machinelearningmastery.com/convert-time-series-supervised-learning-problem-python/

https://machinelearningmastery.com/backtest-machine-learning-models-time-series-forecasting/

https://machinelearningmastery.com/multivariate-time-series-forecasting-lstms-keras/

https://en.wikipedia.org/wiki/Multicollinearity

https://machinelearningmastery.com/understand-machine-learning-data-descriptive-statistics-python/

https://machinelearningmastery.com/gentle-introduction-autocorrelation-partial-autocorrelation/

窗口法和时间步方法的比较

https://machinelearningmastery.com/time-series-prediction-lstm-recurrent-neural-networks-python-keras/

1)LSTMs for Univariate Time Series Forecasting：https://machinelearningmastery.com/time-series-forecasting-long-short-term-memory-network-python/

2)LSTMs for Multivariate Time Series Forecasting：https://machinelearningmastery.com/multivariate-time-series-forecasting-lstms-keras/

3)LSTMs for Multi-Step Time Series Forecasting：https://machinelearningmastery.com/multi-step-time-series-forecasting-long-short-term-memory-networks-python

How to Update LSTM Networks During Training for Time Series Forecasting

https://machinelearningmastery.com/update-lstm-networks-training-time-series-forecasting/

独热码

https://machinelearningmastery.com/how-to-one-hot-encode-sequence-data-in-python/

2.学习率怎么调

将整个training set的10-20%作为validation set，每个epoch之后计算val set的loss，将val loss和train loss比较，当train loss持续下降而val loss不再一同下降或不降反升的时候以某个系数削减learning rate。我喜欢0.3左右的值，直到val loss趋于平稳，也就是将early stoping和learning rate的调整结合…

作者：Nutastray
链接：https://www.zhihu.com/question/56152826/answer/147994162
来源：知乎

3.优化器怎么选

https://www.jianshu.com/p/d99b83f4c1a6

4.在keras中如何对参数进行调优

http://baijiahao.baidu.com/s?id=1594071159327391746&wfr=spider&for=pc

5.如何在时间序列预测训练中更新LSTM网络?

https://machinelearningmastery.com/update-lstm-networks-training-time-series-forecasting/

6.详细讲解lstm数学原理

https://zybuluo.com/hanbingtao/note/581764

7.keras：4)LSTM函数详解

https://blog.csdn.net/jiangpeng59/article/details/77646186/

https://blog.csdn.net/luoganttcc/article/details/78981815

8.提升深度学习模型的表现，你需要这20个技巧

https://blog.csdn.net/shingle_/article/details/52653588

9.使用深度学习LSTM时间序列预测

http://www.jakob-aungiers.com/articles/a/LSTM-Neural-Network-for-Time-Series-Prediction

10.LSTM超参数调试注意事项

https://blog.csdn.net/chenzhi1992/article/details/77005876

11.lstm股价预测

Stock Market Predictions with LSTM in Python：https://www.datacamp.com/community/tutorials/lstm-python-stock-market

12.scikit-learn线性回归算法库小结

https://www.cnblogs.com/pinard/p/6026343.html

13.用深度学习每次得到的结果都不一样，怎么办？

https://machinelearningmastery.com/reproducible-results-neural-networks-keras/

https://www.leiphone.com/news/201706/zt4Dm491Ol58C8Mc.html

14.Python时间序列数据的基本特征工程

https://machinelearningmastery.com/basic-feature-engineering-time-series-data-python/