I have 1-hour resolution temperature data with the following sample input-1 format. I use the LSTM to find anomalies with this format successfully. But I am not sure if LSTM’s learn any temporal dependencies with this type of format. Do I need to use the format as shown in input-2 where I create subsequences for each day?
input1
2020-01-01 01:00:00,100
2020-01-01 02:00:00,200
2020-01-01 03:00:00,900
2020-01-01 04:00:00,300
2020-01-02 01:00:00,100
2020-01-02 02:00:00,200
2020-01-02 03:00:00 100
2020-01-02 04:00:00 300
input2
,1,2,3,4
2020-01-01,100,200,900,300
2020-01-02,100,200,100,300