how to get probability of classification reliability

[Sandy4321]

how to get probability of classification reliability
like
predict_proba(self, X)
in
LogisticRegression
https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html

[kliegr]

In principle, this is easy. Proba can be estimated (though imprecisely) as the confidence of the firing rule.

[jirifilip]

Hello, Sandy, sorry for the late reply.

I've just implemented a new method for getting the probability of a prediction.

cba = CBA(algorithm="m1")

test_dataframe = pd.read_csv(dataset_file, sep=";")
        
transactions = TransactionDB.from_DataFrame(test_dataframe)
transactions_test = TransactionDB.from_DataFrame(test_dataframe[:2])

cba.fit(transactions)

cba.predict_probability(transactions_test)

I've also added some new tests into the /test folder to test the new functionality. Everything should work correctly, but if you encounter any errors, please let me know.

[Sandy4321]

great thanks , start to test it
by the way
what is difference between predict_probability_all and predict_probability
for example from your test file?
header1 = ["A", "B", "Y"]
rows1 = [
[1, 1, 0],
[1, 1, 0],
[1, 1, 1],
[0, 0, 0],
[0, 0, 1],
[0, 0, 1]
]

transactions = TransactionDB(rows1, header1)

cba = CBA()

cba.fit(transactions)

probs1 = cba.clf.predict_probability_all(transactions)
probs2 = cba.predict_probability(transactions)
q=0

probs1
[0.5, 0.5, 0.5, 0.6666666666666666, 0.6666666666666666, 0.6666666666666666]
probs2
[0.5, 0.5, 0.5, 0.6666666666666666, 0.6666666666666666, 0.6666666666666666]

when
cba.predict(transactions)
['0', '0', '0', '1', '1', '1']

---

by the way how you deal with unbalanced data?

for example in
https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html
they do have class_weight parameter
so in your case , do you calculate predictions for balanced case or just optimize global precision
for example for this target
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1
if you use global precision
then you even may not run prediction
but give answer
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

meaning all predictions are 0s, then by doing nothing you have good precision ( probability of mistake)
P = 3/30 = 0.1 ??

class_weight dict or ‘balanced’, default=None
Weights associated with classes in the form {class_label: weight}. If not given, all classes are supposed to have weight one.

The “balanced” mode uses the values of y to automatically adjust weights inversely proportional to class frequencies in the input data as n_samples / (n_classes * np.bincount(y)).

Note that these weights will be multiplied with sample_weight (passed through the fit method) if sample_weight is specified.

New in version 0.17: class_weight=’balanced’

[jirifilip]

cba.clf.predict_probability_all and cba.predict_probability are the same method

I added cba.predict_probability so you don't have to access the classifier object with cba.clf