#!/usr/bin/env python3

# Train decision tree to detect phishing
# Dataset: https://archive.ics.uci.edu/ml/datasets/Phishing+Websites#

import csv
import arff
import numpy as np
from sklearn import *
from sklearn import tree
from sklearn.metrics import accuracy_score


# Put the arff dataset into a csv
def saveDataToCSV(inf, outf):
    arffin = csv.reader(open(inf, 'r'), delimiter=',')
    arffout = csv.writer(open(outf, 'w'))
    arffout.writerows(arffin)



# Train decision tree
def traintree(incsv):
    training_data = np.genfromtxt(incsv, delimiter=',', dtype=np.int32) # load the dataset
    inputs = training_data[:, :-1] # all attributes except the last one
    outputs = training_data[:, -1] # last attribute

    # Divide dataset into training and testing
    training_inputs = inputs[:2000]
    training_outputs = outputs[:2000]
    testing_inputs = inputs[2000:]
    testing_outputs = outputs[2000:]

    classifier = tree.DecisionTreeClassifier()
    classifier.fit(training_inputs, training_outputs)
    predictions = classifier.predict(testing_inputs)
    accuracy = 100.0 * accuracy_score(testing_outputs, predictions)

    print("Accuracy: " + str(accuracy))

def main():
    saveDataToCSV("dataset.arff", "dataset.csv")
    traintree("dataset.csv")

if __name__ == '__main__':
    main()