I have a dataset where I am recording temperature after every 4milliseconds till 500 and another feature "conductivity value". The length of the dataset is around a 1000 rows. I need to find the conductivity value based on the temperature pattern.

t1 t2 t3 .... t5 conductivity
90 91 93 .... 96 0.34
92 91 93 .... 95 0.36

I am bit confused on how to use the dataset in a time series model such as LSTM because I have all the time sequence in columns and I don't know the conductivity values in between as in t2,t3,t4.

I think the dataset becomes a classification problem with the current format.

Can you guys help me out?

  • 1
    $\begingroup$ I think that what You could do is to treat your 't' data as input vector and conductivity value as y and, as You said, train LSTM-based network on this. If You want to make it work in real time, You can collect last 5 't' values into a list and then put it into network for prediction. After that erase the oldest value from the list and add new one on the other side. If You'll loop that You'll have real-time conductivity predictions. $\endgroup$ Dec 21, 2021 at 13:02

1 Answer 1


I don't think time series model necessarily makes sense if you have one conductivity value to predict for each time series.

A regression like setup makes more sense here: you could model this by letting the vector of time points represent the input. So you'd end up with a $n \mbox{ x } t$ matrix as input to predict the conductivity value.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .