Issue with custom stacking pipeline about lime HOT 5 CLOSED

Are you sure pipe_stacking.predict_proba(test_data) works perfectly fine? In this line:
self.prob_result = [self.classifiers[0].predict_proba(X)[0][0], self.classifiers[1].predict_proba(X)[0][0], self.classifiers[2].predict_proba(X)[0][0], np.asarray(list( self.classifiers[3].predict_proba(X)))[0][0], np.asarray(list( self.classifiers[4].predict_proba(X)))[0][0]]
you seem to be taking the first row and first column ([0][0]) of the prediction probability for each classifier, regardless of the size of the input. That is, I think you'll always output one row in predict_proba, even if the input is 10 rows.
LimeTabular assumes the X in predict_proba can be a 2d array, not only a 1d array.

from lime.

This is a use case specific implementation, wherein the pipeline is used only for prediction against a single input vector. Henceforth the size of the input is one here. X in predict_proba comes after PCA, which is a 2d array. I have used the same training input matrix for initialization & used the same input data point against lime explainer built on Random forest earlier, it did work fine. It doesnt seem to work for the custom stacking implementation.

The error log has marked this an error;

--> 531     check_consistent_length(X, y)
ValueError: Found input variables with inconsistent numbers of samples: [44, 1] 

from lime.

can you share a notebook with the error?

from lime.

Sure. I have shared the same to your inbox.

from lime.

Thanks for the right direction. Your previous correction was perfectly right, the stacking predict_proba function was returning only the first output. I had changed it to work on an array of any given size. Since the sample in my implementation was initialized with a size 44, the ideal output from the predict_proba from my stacking implementation should be (44x2).

from lime.