train_multiclass.py

import sys
import os
from tqdm import tqdm
import numpy as np
import pandas as pd
import argparse
from src import predict_multiclass, preprocessing
from collections import Counter
from sklearn.model_selection import train_test_split
from imblearn import over_sampling
from imblearn import combine
from sklearn import svm
from sklearn.tree import DecisionTreeClassifier
from sklearn.neighbors import KNeighborsClassifier

# set save path
def set_path(basename):
    os.makedirs('./output', exist_ok=True)
    name = os.path.splitext(basename)
    save_path = 'output/{}.csv'.format(name[0]) 
    return save_path, name[0]

# train data + mndo data
def append_mndo(X_train, y_train, df):
    X_mndo = df.drop('Label', axis=1)
    y_mndo = df.Label
    X_mndo = np.concatenate((X_mndo, X_train), axis=0)
    y_mndo = np.concatenate((y_mndo, y_train), axis=0)

    return X_mndo, y_mndo


if __name__ == '__main__':
    # Load dataset
    parser = argparse.ArgumentParser()
    parser.add_argument('data', help='dataset')
    parser.add_argument('generated', help='generated data')
    args = parser.parse_args()
    
    try:
        data = pd.read_csv(args.data)
        mndo_generated = pd.read_csv(args.generated)
        save_path, file_name = set_path(os.path.basename(args.data))
    except IndexError:
        sys.exit('error: Must specify dataset file')
    except FileNotFoundError:
        sys.exit('error: No such file or directory')

    # one-hot encoding and sort columns
    data = pd.get_dummies(data)   
    mndo_generated = mndo_generated.ix[:, data.columns]

    # split the data
    X = data.drop('Label', axis=1)
    y = data.Label
    X = np.array(X)
    y = np.array(y)

    # calc number of  samples to synthesize
    RANDOM_STATE=42
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4,
            shuffle=True, random_state=RANDOM_STATE)

    print('y_train: {}'.format(Counter(y_train)))
    print('y_test: {}'.format(Counter(y_test)))


    #-----------------
    # Preprocessing
    #-----------------
    # Multivariate over-sampling
    X_mndo, y_mndo = append_mndo(X_train, y_train, mndo_generated)
    print('y_mndo: {}'.format(Counter(y_mndo)))

    os_list = [[X_mndo, y_mndo]]
       
    # scaling 
    os_list, X_test_scaled = preprocessing.normalization(os_list, X_test)
    #os_list, X_test_scaled = preprocessing.standardization(os_list, X_test)

    #-------------
    # Learning
    #-------------
    svm_clf = []
    pred_tmp = []

    #svm
    for i in range(len(os_list)):
        svm_clf.append(svm.SVC(kernel='rbf', gamma='auto',
            random_state=RANDOM_STATE, probability=True).fit(os_list[i][0], os_list[i][1]))
        
    for i in range(len(svm_clf)):
        prob = svm_clf[i].predict_proba(X_test_scaled[i])[:,1]
        pred_tmp.append(predict_multiclass.calc_metrics(y_test, svm_clf[i].predict(X_test_scaled[i]), i))
    
    # tree
    tree_clf = []
    for i in range(len(os_list)):
        tree_clf.append(DecisionTreeClassifier(random_state=RANDOM_STATE).fit(os_list[i][0], os_list[i][1]))
        
    for i in range(len(tree_clf)):
        prob = tree_clf[i].predict_proba(X_test_scaled[i])[:,1]
        pred_tmp.append(predict_multiclass.calc_metrics(y_test, tree_clf[i].predict(X_test_scaled[i]), i))

    #k-NN
    k=5
    knn_clf = []
    for i in range(len(os_list)):
        knn_clf.append(KNeighborsClassifier(n_neighbors=k).fit(os_list[i][0], os_list[i][1]))
        
    for i in range(len(knn_clf)):
        prob = knn_clf[i].predict_proba(X_test_scaled[i])[:,1]
        pred_tmp.append(predict_multiclass.calc_metrics(y_test, knn_clf[i].predict(X_test_scaled[i]), i))
  
    pred_df = pd.DataFrame(pred_tmp)
    pred_df.columns = ['os', 'Sensitivity', 'Specificity', 'Geometric mean', 'F-1', 'MCC']
   
    # export resualt
    pred_df.to_csv(save_path, index=False)