I have sensor dataset. I have already classified these data with LSTMs.I have a dataframe with 2 features and a class column. Assume that I take every two rows(inputs) respectively and make the dataframe as 4 features which means I merge two inputs together even they have same features.
Converting from Table 1
|Time||Feature 1||Feature 2||class|
to Table 2
|Time||Feature 1||Feature 2||Feature 3||Feature 4||class|
Is this configuration may increase accuracy? Is this configuration may lighten LSTMs? Maybe the provided data can be classificated with classical dense networks? Or, Is it not necessary?