[FIXED] Making sense of predicted probabilities for multi-labeled problems

Issue

I'm currently working on a multi-label classification problem using scikit-learn, and I'm having a bit of trouble learning how to get the predicted probabilities for each class/label - similar to what scikit-learn's .predict_proba() method does on binary classification tasks.

My y is a 100x10 2d array, having 10 unique classes, and I'm using scikit-learn's ensemble.RandomForestClassifier() as my classifier.

What I would like to do is just predict the probability that a set of given features belongs to one of the classes (noted below as cl_) in y. So basically I'm imagining an output similar to this:

cl_1 | cl_2 | cl_3 | cl_4 | cl_5 | cl_6 | cl_7 | cl_8 | cl_9 | cl_10
---------------------------------------------------------------------
 0.0 |  0.0 |  0.0 | 0.1  |  0.3 | 0.0  | 0.0  |  0.0 | 0.6  | 0.0

NOTE: I have fit the Random Forest classifier to the dataset and the .predict() method works as expected. However, the .predict_proba() method returns probabilities in the shape of C x M, where C is the number of unique classes, and M is the number of instances in the dataset.

Clarification/Update: Here's an actual y instance from my dataset:

[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 0]

Using the .predict_proba() on X_test returns a 2d-array of shape 39 x 848,048 (where 39 is the number of unique classes and 848,048 is the number of instances). So in layman's terms, what I'm asking is: How do I make since of what .predict_proba() returns (insofar as the little chart above is concerned).

NOTE: This:

y_hat_proba = clf.predict_proba(X_test)
print(y_hat_proba[0][1])

returns [ 1. 0.]

Solution

You can use .predict_proba(X_test) to get the probability per class on a new set of points. If you need a prediction for a single point, just make sure X_test has a single row.

The shape of the output should be in the shape [n_samples, n_classes] (in the test set). You can see the function's documentation.

Answered By - Dimosthenis