标签： python neural-network time-series lstm forecasting
我对时间序列预测还很陌生，但我有些困惑。我有一个看起来如下的数据集：
Data forecasting. There are 100 columns and 22 rows. I need to forecast the 22nd row based on the previous 21 rows. The values seem to range between [-50,50]
这些行可以看作是每月的观察。我使用了简单的MA ARCH方法以及LSTM。但是我没有得到令人满意的结果。有什么建议对这种数据集最适合哪种预测方法？
![Data forecasting. There are 100 columns and 22 rows. I need to forecast the 22nd row based on the previous 21 rows. The values seem to range between [-50,50]](https://i.stack.imgur.com/TKJmJ.png){kind=link}