LR_solver.py 文件源码

def generate_LR_model(file_name):
    train_df = read_from_file(file_name)
    selected_train_df = train_df.filter(regex='label|connectionType_.*|telecomsOperator_.*|sitesetID_.*|positionType_.*|gender_.*|haveBaby_.*|age_scaled')
    train_np = selected_train_df.as_matrix()
    y = train_np[:,0]
    X = train_np[:,1:]
    print 'Train Logistic Regression Model...'
    start_time  = datetime.datetime.now()
    clf = linear_model.LogisticRegression(penalty='l2',C=1.0,solver='sag',n_jobs=-1, tol=1e-6, max_iter=200)#, class_weight='balanced')
    clf.fit(X,y)
    end_time = datetime.datetime.now()
    print 'Training Done..., Time Cost: '
    print (end_time-start_time).seconds

    print 'Save Model...'
    joblib.dump(clf, 'LR.model')
    return clf