from sklearn.naive_bayes import MultinomialNB nb_classifier = MultinomialNB() #train the naive bayes model on features present in count_train nb_classifier.train(count_train, y_train) #predict pred = nb_classifier.predict(count_test) #check accuracy score = metrics.accuracy_score(y_test, pred) print(score) #confusion matrix cm = metrics.confusionmatrix(y_test, pred, labels=['FAKE', 'REAL']) print(cm) ##naive bayes classifier using features created by tdidf vectorizer nb_classifier2 = MultinomialNB() nb_classifier2.train(tfidf_train, y_train) pred2 = nb_classifier2.predict(tfidf_test) #check accuracy score2 = metrics.accuracy_score(y_test, pred2) print(score2) #confusion matrix cm = metrics.confusionmatrix(y_test, pred2, labels=['FAKE', 'REAL'])
employment = {'Salaried': 0, 'Self employed': 1, ' ': 1}
#importing dataset
dataset = pd.readcsv('CreditCardData.csv')
dataset = dataset.dropna()
#creating dummy variable for column EmploymentType
dataset.EmploymentType = [
employment[number] for number in dataset.EmploymentType
]
loanDefaulter = dataset[['loandefault']]
factors = dataset[[
'disbursedamount', 'assetcost', 'EmploymentType', 'PRI.CURRENT.BALANCE',
'PRI.SANCTIONED.AMOUNT', 'PRIMARY.INSTAL.AMT'
]]
#creating training and testing set and scaling the data
factorsTrain, factorsTest, loanTrain, loanTest = sl.traintestsplit(
factors, loanDefaulter, testsize=0.2, shuffle=True)
scaler = StandardScaler()
factorsTrain = scaler.fittransform(factorsTrain)
factorsTest = scaler.fittransform(factorsTest)
#creating regression object
LogisticRegressor = LogisticRegression()
LogisticRegressor.fit(factorsTrain, loanTrain)
predictedResult = LogisticRegressor.predict(factorsTest)
print('Confusion matrix using solver: ')
print(metrics.confusionmatrix(loanTest, predictedResult))
print('Accuracy using solver: ')
print('Accuracy:', metrics.accuracy_score(loanTest, predictedResult))
rf = RandomForestClassifier(n_estimators=100, max_depth=10, min_samples_split=2, min_samples_leaf=1)
rf.fit(fps_train, act_train)
# write the model
cPickle.dump(rf, outfile, 2)
i += 1
outfile.close()
print "done"
# make predictions for test compounds
test_SVMpredictions = rf.predict(fps_test)
# cross-validate
scores = cross_validation.cross_val_score(rf, fps_train, act_train, cv=5)
# scores = cross_validation.cross_val_score(clf_RF, X_train,y_train, cv=cv_counter,score_func=metrics.zero_one_score)
# output confusion matrix and percentage accuracy on test sets
print metrics.confusionmatrix(act_test, test_SVMpredictions)
accuracy = rf.score(fps_test, act_test)
print accuracy
# calculate probabilities for each test molecules
test_SVMprobabilities = rf.predict_proba(fps_test)
# compute AUC metric for this CV fold
fpr, tpr, thresholds = metrics.roc_curve(act_test, test_SVMprobabilities)
roc_auc = metrics.auc(fpr, tpr)
print "AUC (fold %d/%d): %f" % (i + 1, n, roc_auc)
mean_auc += roc_auc
print "Mean AUC: %f" % (mean_auc/n)
plt.plot(fpr, tpr, label="Model#%d (AUC=%.2f)" % (model_id + 1, roc_auc))
for k,v in testmols_dic.items():
test_fp = AllChem.GetMorganFingerprintAsBitVect(k, 2)
test_fps.append(test_fp)
test_fps_array = np.array(test_fps)
test_acts.append(testmols_dic [k])
test_acts_array = np.array(test_acts)
# make predictions for test compounds
test_SVMpredictions = rf.predict(test_fps_array)
# cross-validate
scores = cross_validation.cross_val_score(rf, fps_array, acts_array, cv=5)
# scores = cross_validation.cross_val_score(rf, X_train,y_train, cv=cv_counter,score_func=metrics.zero_one_score)
# output confusion matrix and percentage accuracy on test sets
print metrics.confusionmatrix(test_acts_array, test_SVMpredictions)
accuracy = rf.score(test_fps_array, test_acts_array)
print accuracy
# calculate probabilities for each test molecules
test_SVMprobabilities = rf.predict_proba(test_fps_array)
# compute AUC metric for this CV fold
fpr, tpr, thresholds = metrics.roc_curve(test_acts_array, test_SVMprobabilities)
roc_auc = metrics.auc(fpr, tpr)
print "AUC (fold %d/%d): %f" % (i + 1, n, roc_auc)
mean_auc += roc_auc
print "Mean AUC: %f" % (mean_auc/n)
plt.plot(fpr, tpr, label="Model#%d (AUC=%.2f)" % (model_id + 1, roc_auc))
