How to understand the outcome of a prediction (keras, R, tutorial code)

[Bettina10]

Hello everyone,

I try to follow the tutorial from "Deep learning with R" and use it for binary classification with time series, here is the link (and thanks a lot for jjallaire´s tutorial):

• [https://jjallaire.github.io/deep-learning-with-r-notebooks/notebooks/6.3-advanced-usage-of-recurrent-neural-networks.nb.html ]

  I found that:

  1. Total number of rows in predict_generator > total number of rows in prediction outcome + 2*lookback
  2. The relationship seams: Rows_Nr. of Prediction = Rows_Nr. of Predict_generator + lookback

And the question is, how could I check the result of the prediction? I mean:

1. How could I get the table for true positive, false positive, etc
2. How should I understand the relationship of the row_numbers between predict_generator and prediction outcome?

Thanks in advance for your time to read this and any answer for this question!

[Bettina10]

I´ve tried another model. The model has a lookback of 864, a batch_size of 512. The dataset for prediction (the input of test_generator) has 20000 lines. The output of predict_generator has 18981 lines. Does somebody knows how to interpret it?