How to get original values after using factorize() in Python?

Question

I'm a beginner trying to create a predictive model using Random Forest in Python, using train and test datasets. train["ALLOW/BLOCK"] can take 1 out of 4 expected values (all strings). test["ALLOW/BLOCK"] is what needs to be predicted.

y,_ = pd.factorize(train["ALLOW/BLOCK"])

y
Out[293]: array([0, 1, 0, ..., 1, 0, 2], dtype=int64)

I used predict for the prediction.

clf.predict(test[features])

clf.predict(test[features])[0:10]
Out[294]: array([0, 0, 0, 0, 0, 2, 2, 0, 0, 0], dtype=int64)

How can I get the original values instead of the numeric ones? Is the following code actually comparing the actual and predicted values?

z,_= pd.factorize(test["AUDIT/BLOCK"])

z==clf.predict(test[features])
Out[296]: array([ True, False, False, ..., False, False, False], dtype=bool) 

Answer 1

First, you need to save the label returned by pd.factorize as follows:

y, label = pd.factorize(train["ALLOW/BLOCK"])

And then after you got the numeric predictions, you can extract the corresponding labels by label[pred]:

pred = clf.predict(test[features])
pred_label = label[pred]

pred_label contains predictions with the original values.

---

No you should not re factorize the test predictions, since very likely the label would be different. Consider the following example:

pd.factorize(['a', 'b', 'c'])
# (array([0, 1, 2]), array(['a', 'b', 'c'], dtype=object))

pd.factorize(['c', 'a', 'b'])
# (array([0, 1, 2]), array(['c', 'a', 'b'], dtype=object))

So the label depends on the order of the elements.