How can I align test data to predicted data with keras.prediction?

Question

I have a trained model and I am running prediction with keras via:

model = pets.get_model(input_size=input_units)
model.compile(loss='categorical_crossentropy',
              optimizer='adam', metrics=['accuracy'])
model.load_weights('models/2019-03-01-02-03-53.h5')

prediction = model.predict(X)

This gives me a list that looks like [0.323 0.43 .099] and so on. How can I map that to rows in my X (which is a pandas DataFrame) so that I have an easy representation of input to outputs?

Victor Valente · Accepted Answer · 2019-03-01T16:04:34.330

2

By default, model.predict(X) and X itself will already be in the same order (first value of prediction corresponds to X's first row and so on).

For ease of visualization, you could try something like X['predicitons'] = prediction but that would add a column to X.

edited Mar 01 '19 at 16:04

answered Mar 01 '19 at 15:57

Victor Valente

761
9
24

Oh snap - so simple! And it'll be perfectly aligned? – Shamoon Mar 01 '19 at 15:59
Yes. I updated the answer to add a little bit of context. – Victor Valente Mar 01 '19 at 16:05
1

Wait - slight issue. My prediction has 3 columns. So I need to somehow add 3 columns? – Shamoon Mar 02 '19 at 22:51
1

Does your dataset have 3 different classes? If that's the case, then you can turn `prediction` into a DataFrame and join with X. Take a look at https://stackoverflow.com/questions/20602947/append-column-to-pandas-dataframe – Victor Valente Mar 04 '19 at 12:52

How can I align test data to predicted data with keras.prediction?

1 Answers1