I have a dataset where I have multiple entries for the same timestamp and I want to use LSTM to forecast the next timestamp given the previous 5 timesteps. From https://machinelearningmastery.com/multivariate-time-series-forecasting-lstms-keras/ i understand how to deal with multivariate time series and in the blog he reshaped the time series such that we forecast the next step based on the previous time steps. But in my case i have multiple entries for the same time step when i do the reshape it creates the wrong relationship since it assumes the time it's sequential but in fact they are the same time step. In this scenario how should I structure my data for training?

The data is a COVID dataset with header below:

Accurate_Episode_Date, Age_Group, Client_Gender, Case_AcquisitionInfo, Reporting_PHU_City, Outbreak_Related, Outcome1

2020-03-30, 70s, MALE, OB, Stratford, Yes, Fatal
  • $\begingroup$ Like this, there might be several possibilities. Can you share an example of your data? $\endgroup$ – Oscar Apr 11 at 4:20
  • $\begingroup$ @Oscar i updated the summary. $\endgroup$ – AlwaysNull Apr 12 at 0:45
  • $\begingroup$ So you wanna predict the outcome of a covid infection out of a time series of covid episodes? $\endgroup$ – Oscar Apr 12 at 10:24
  • $\begingroup$ yes that's correct! $\endgroup$ – AlwaysNull Apr 12 at 23:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.