def train_test_split_evaluation_procedure(X_train ,Y_train, X_test, Y_test):
print 'Starting with train_test_split procedure','\n'
accuracy_score =[]
model = KNeighborsClassifier(n_neighbors=5)
model.fit(X_train, Y_train)
predicted_labels = model.predict(X_test)
accuracy_score.append(metrics.accuracy_score(Y_test, predicted_labels))
print("Confusion Matrix")
print(confusion_matrix(Y_test, predicted_labels))
print("Precision")
print(precision_score(Y_test, predicted_labels, average=None))
print("Recall")
print(recall_score(Y_test, predicted_labels, average=None))
print("F1 score")
print(f1_score(Y_test, predicted_labels, average=None))
filename = 'pickle_modified/KNN_5.pkl'
pickle.dump(model, open(filename, 'wb'))
return accuracy_score
def crossvalidation(): neural_model = MLPClassifier(solver='sgd',activation='logistic',alpha=0.00000001,learning_rate_init=0.2,learning_rate='adaptive') cv = KFold(Number_of_samples,n_folds=3) Total_TrainData=df[['X Coordinate','Y Coordinate','Year','MONTH','DAY','Time_in_Mintues']] Total_TrainData_Target=df['Primary Type'] newtraindata=Total_TrainData.as_matrix() newtraindata_target=Total_TrainData_Target.as_matrix() accuracy_score = [] accuracy=[] for train_cv, test_cv in cv: model = neural_model.fit(newtraindata[train_cv],newtraindata_target[train_cv]) neuralnetwork_predictions=model.predict(newtraindata[test_cv]) accuracy_score.append(model.score(newtraindata[test_cv],newtraindata_target[test_cv])) #accuracy.append(accuracy_score(neuralnetwork_predictions,newtraindata_target[test_cv])) mean_accuracy= np.array(accuracy_score).mean() print accuracy_score #print accuracy return mean_accuracy
def main():
model = Model()
university_names = []
accuracy_score = []
# list of training csv files
training_List = ['asu.csv', 'clemson.csv', 'iitc.csv', 'mtu.csv']
# list of testing csv files
testing_list = [
'asu_test.csv', 'clemson_test.csv', 'iitc_test.csv', 'mtu_test.csv'
]
# get results for each university data
for trainD, testD in zip(training_List, testing_list):
result, uni_name = model.loadData(trainD, testD)
accuracy_score.append(result * 100)
university_names.append(uni_name)
# end for
print('University predictions for student:')
for uni, accuracy in zip(university_names, accuracy_score):
print(uni + ' -> ' + str(accuracy) + '%\n')
macro_recall = report['macro avg']['recall']
macro_f1 = report['macro avg']['f1-score']
accuracy = report['accuracy']
print('jjjjj',macro_precision)
#f1_score1 = (f1_score(y, y_pred))
f1_score.append(macro_f1)
#precision_score1 = (precision_score(y, y_pred))
precision_score.append(macro_precision)
#recall_score1 = (recall_score(y, y_pred))
recall_score.append(macro_recall)
#acc = accuracy_score(y, y_pred)
accuracy_score.append(accuracy)
sensitivity1 = cm_train[1,1]/(cm_train[1,0]+cm_train[1,1])
#print(sensitivity)
#0.6666666666666666
sensitivity.append(sensitivity1)
specificity1 = cm_train[0,0]/(cm_train[0,0]+cm_train[0,1])
#print(specificity)
#0.89171974522293
specificity.append( specificity1)
PPV1 = cm_train[1,1]/(cm_train[1,1]+cm_train[0,1])
#print(PPV)
#0.5
# Libraries
from sklearn.svm import SVC
from sklearn.metrics import jaccard_similarity_score
from sklearn.metrics import f1_score
import matplotlib.pyplot as plt
get_ipython().run_line_magic('matplotlib', 'inline')
kernel_func = ['sigmoid', 'poly', 'rbf', 'linear']
accuracy_score = []
for k in kernel_func:
svc_model = SVC(C=0.01, gamma='auto', kernel=k)
svc_model.fit(X_train, y_train)
svc_yhat = svc_model.predict(X_test)
accuracy_score.append(f1_score(y_test, svc_yhat, average='weighted'))
# In[31]:
# Visualization of best kernel
y_pos = np.arange(len(kernel_func))
plt.bar(y_pos, accuracy_score, align='center', alpha=0.8)
plt.xticks(y_pos, kernel_func)
plt.xlabel('Kernel Functions')
plt.ylabel('Accuracy')
# In[32]:
# Final Support Vector Machine Model
svc_model = SVC(C=0.01, gamma='auto', kernel='poly')
svc_model.fit(X_train, y_train)
X_train, X_test, y_train, y_test = train_test_split(X_sm,
Y_sm,
random_state=10,
test_size=0.3)
from sklearn.feature_selection import RFE
accuracy_score = []
for i in range(1, 39):
X_train_rfe = X_train
X_test_rfe = X_test
logreg = LogisticRegression()
rfe_model = RFE(estimator=logreg, n_features_to_select=i)
rfe_model = rfe_model.fit(X_train_rfe, y_train)
feat_index = pd.Series(data=rfe_model.ranking_, index=X_train_rfe.columns)
signi_feat_rfe = feat_index[feat_index == 1].index
accuracy_score.append(rfe_model.score(X_train_rfe, y_train))
lis_acc = {(i + 1, np.round(accuracy_score[i], 4)) for i in range(0, 38)}
lis_acc
# # Logistic regression (RFE)
# In[47]:
logreg = LogisticRegression()
rfe_model = RFE(estimator=logreg, n_features_to_select=33)
rfe_model = rfe_model.fit(X_train, y_train)
feat_index = pd.Series(data=rfe_model.ranking_, index=X_train.columns)
signi_feat_rfe = feat_index[feat_index == 1].index
print(signi_feat_rfe)