How to use predict_proba at SVM Sentiment Analysis

Question

I applied predict_proba on classfication problem. I have some experience with building classification models in R but it's my first time with Python's sklearn.

So the the problem is: In sklearn after fitting I can't find a way to access probabilities. Is it possible? There is a method predict_proba(), but...as the name suggests, it is prediction. This is my code:

import pandas as pd 
from sklearn.svm import SVC
from sklearn.svm import LinearSVC
import pickle
from nltk.tokenize import sent_tokenize
from Model import SkripsiPipeline

def konten(kata, model):
    item = []
    loaded_model = pickle.load(open(model, 'rb'))
    for v in kata.itertuples(index = False):
        sentiment = [] 
        variabel1 = v[0]
        variabel2 = v[1]

        kalimat = variabel1 + variabel2

        hasil_tokenize = sent_tokenize(kalimat)

        preds = loaded_model.predict(hasil_tokenize)

        if preds == 1:
            proba = loaded_model.predict_proba(hasil_tokenize)
            proba = proba.reshape(-1, 1).tolist()
            sentiment.append('Positif')
            sentiment.append(proba[0])

        elif preds == 0:
            proba = loaded_model.predict_proba(hasil_tokenize)
            proba = proba.reshape(-1, 1).tolist()
            sentiment.append('Netral')
            sentiment.append(proba[1])

        elif preds == -1:
            proba = loaded_model.predict_proba(hasil_tokenize)
            proba = proba.reshape(-1,1).tolist()
            sentiment.append('Negatif')
            sentiment.append(proba[2])

        item.append(sentiment)

    return item

but I got this error :

AttributeError: 'SkripsiPipeline' object has no attribute 'predict_proba'

and this is the SkripsiPipeline code :

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer
from sklearn.metrics import accuracy_score, confusion_matrix
from sklearn.feature_extraction.text import TfidfTransformer
from sklearn.model_selection import KFold
from sklearn.svm import SVC
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import classification_report, confusion_matrix, accuracy_score
import pickle

class SkripsiPipeline():
    def __init__ (self, predictor):
        self.predictor = predictor
    def fit(self,X,y):
        vectorizer = CountVectorizer()
        tfidf_transformer = TfidfTransformer()
        svm_predictor = self.predictor
        X = vectorizer.fit_transform(X)
        X = tfidf_transformer.fit_transform(X)
        svm_predictor.fit(X,y)

        self.vectorizer = vectorizer
        self.tfidf_transformer = tfidf_transformer
        self.svm_predictor = svm_predictor

    def predict (self, X):
        X = self.vectorizer.transform(X)
        X = self.tfidf_transformer.transform(X)
        prediction = self.svm_predictor.predict(X)
        return prediction

I am new to Python Sklearn package. Can anyone tell me what is wrong with my Python code. I've googled it, but could not understand properly.

score 0 · Answer 1 · answered Sep 11 '19 at 04:49

You call a predict_proba method that doesn't exist in the SkripsiPipeline class. You should implement one, similar to the predict method, but that calls predict_proba from your svm_predictor instead of its predict method.

it should look like this:

def predict_proba(self, X):
    X = self.vectorizer.transform(X)
    X = self.tfidf_transformer.transform(X)
    proba = self.svm_predictor.predict_proba(X)
    return proba

score 0 · Answer 2 · answered Oct 08 '21 at 16:00

I know that is late, but those who may encounter this issue, according to sklearn documentation , the method 'predict_proba' is not defined for 'LinearSVC' so you ahve 2 options:

1)

LinearSVC_classifier = SklearnClassifier(SVC(kernel='linear',probability=True))

Use SVC with linear kernel, with probability argument set to True. Just as explained in here .

2) You can use _predict_proba_lr() instead predict_proba. Something like this:

from sklearn import svm
clf=svm.LinearSVC()

clf.fit(X_train,Y_train)

res= clf._predict_proba_lr(X_test,Y_test)

res would be a 2d array of probabilities of each classes against samples.

score 0 · Answer 3 · answered Jul 25 '22 at 12:01

I workaround is to wrap LinearSVC in a CalibratedClassifierCV:

linear_svc = LinearSVC()
calibrated_clf = CalibratedClassifierCV(
           base_estimator=linear_svc,
           cv="prefit")

# assume xtrain, ytrain is your train data
calibrated_clf.fit(xtrain, ytrain)
yhat = calibrated_clf.predict_proba(xtest)

yhat[:5]

0s
yhat[:5]
array([[0.01336953, 0.98663047],
       [0.82758885, 0.17241115],
       [0.99592713, 0.00407287],
       [0.99374837, 0.00625163],
       [0.08030293, 0.91969707]])

How to use predict_proba at SVM Sentiment Analysis

3 Answers3