I first got the concept of LSTM on how it works word to word prediction etc. However, I want to know how it work with the time-series classification. For example I have the follwing data (see image below).
Typically, we divide data into segments and feed them to the classifier. However, for the sake of simplicity, here, we will assume only one instance as a segment (window) to the LSTM. So now, my issue is, how can I apply LSTM to this type of data set? In my opinion, I have the following options:
One way One way is a many-to-one relationship ship. I should feed one instance to the LSTM at a time and learn the temporal sequence between Features 1, 2, and 3. However, in this case, I believe that I would need to utilize three LSTM cells. Is it going to be appropriated? To be honest, I think NO, based on my uneducated guess.
Another possibility is a many-to-many relationship. That is, I should feed several instances of LSTM simultaneously. For example, if I use three LSTM cells, I will need to input three instances to the LSTM simultaneously. 1 instance to the first cell, 2 instances to the second cell, and 3 instances to the third cell. Am I right?
I truly have no idea how LSTM will benefit this type of data. Researchers have previously demonstrated that LSTM works exceptionally well with this type of data set. I read several articles but could not correctly grasp how this type of data will feed to LSTM.