DotaTest.py

import Perceptron as pc
import VotedPerceptron as vp
import numpy as np
import pandas as p
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix
from sklearn.metrics import accuracy_score
from sklearn.metrics import classification_report

def test():
    print ("DOTA TEST")

    y = []
    X = []
    Y = []

    dt = p.read_csv("datasets/dota2Train.csv")
    dt.columns = ['E', 'cluster ID', 'game mode', 'game type', 'Anti-Mage', 'Axe', 'bane', 'Bane', 'Bloodseeker',
                  'Crystal Maiden', 'Drow Ranger', 'Earthshaker', 'Juggernaut', 'Mirana', 'Shadow Fiend', 'Morphling',
                  'Phantom Lancer', 'Puck', 'Pudge', 'Razor'

        , 'Sand King'

        , 'Storm Spirit'

        , 'Sven'

        , 'Tiny'

        , 'Vengeful Spirit'

        , 'Windranger'

        , 'Zeus'

        , 'Kunkka'

        , 'Lina'

        , 'Lich'

        , 'Lion'

        , 'Shadow Shaman'

        , 'Slardar'

        , 'Tidehunter'

        , 'Witch Doctor'

        , 'Riki'

        , 'Enigma'

        , 'Tinker'

        , 'Sniper'

        , 'Necrophos'

        , 'Warlock'

        , 'Beastmaster'

        , 'Queen of Pain'

        , 'Venomancer'

        , 'Faceless Void'

        , 'Skeleton King'

        , 'Death Prophet'

        , 'Phantom Assassin'

        , 'Pugna'

        , 'Templar Assassin'

        , 'Viper'

        , 'Luna'

        , 'Dragon Knight'

        , 'Dazzle'

        , 'Clockwerk'

        , 'Leshrac'

        , 'Natures Prophet'

        , 'Lifestealer'

        , 'Dark Seer'

        , 'Clinkz'

        , 'Omniknight'

        , 'Enchantress'

        , 'Huskar'

        , 'Night Stalker'

        , 'Broodmother'

        , 'Bounty Hunter'

        , 'Weaver'

        , 'Jakiro'

        , 'Batrider'

        , 'Chen'

        , 'Spectre'

        , 'Doom'

        , 'Ancient Apparition'

        , 'Ursa'

        , 'Spirit Breaker'

        , 'Gyrocopter'

        , 'Alchemist'

        , 'Invoker'

        , 'Silencer'

        , 'Outworld Devourer'

        , 'Brewmaster'

        , 'Shadow Demon'

        , 'Lone Druid'

        , 'Chaos Knight'

        , 'Meepo'

        , 'Treant Protector'

        , 'Ogre Magi'

        , 'Undying'

        , 'Rubick'

        , 'Disruptor'

        , 'Nyx Assassin'

        , 'Naga Siren'

        , 'Keeper of the Light'

        , 'Wisp'

        , 'Visage'

        , 'Slark'

        , 'Medusa'

        , 'Troll Warlord'

        , 'Centaur Warrunner'

        , 'Magnus'

        , 'Timbersaw'

        , 'Bristleback'

        , 'Tusk'

        , 'Skywrath Mage'

        , 'Abaddon'

        , 'Elder Titan'

        , 'Legion Commander'

        , 'Ember Spirit'

        , 'Earth Spirit'

        , 'Abyssal Underlord'

        , 'Terrorblade'

        , 'Phoenix'

        , 'Techies'

        , 'Oracle'

        , 'Winter Wyvern'

        , 'Arc Warden', 'giack']

    dtest = p.read_csv('datasets/dota2Test.csv')
    dtest.columns = ['E', 'cluster ID', 'game mode', 'game type', 'Anti-Mage', 'Axe', 'bane', 'Bane', 'Bloodseeker',
                     'Crystal Maiden', 'Drow Ranger', 'Earthshaker', 'Juggernaut', 'Mirana', 'Shadow Fiend',
                     'Morphling', 'Phantom Lancer', 'Puck', 'Pudge', 'Razor'

        , 'Sand King'

        , 'Storm Spirit'

        , 'Sven'

        , 'Tiny'

        , 'Vengeful Spirit'

        , 'Windranger'

        , 'Zeus'

        , 'Kunkka'

        , 'Lina'

        , 'Lich'

        , 'Lion'

        , 'Shadow Shaman'

        , 'Slardar'

        , 'Tidehunter'

        , 'Witch Doctor'

        , 'Riki'

        , 'Enigma'

        , 'Tinker'

        , 'Sniper'

        , 'Necrophos'

        , 'Warlock'

        , 'Beastmaster'

        , 'Queen of Pain'

        , 'Venomancer'

        , 'Faceless Void'

        , 'Skeleton King'

        , 'Death Prophet'

        , 'Phantom Assassin'

        , 'Pugna'

        , 'Templar Assassin'

        , 'Viper'

        , 'Luna'

        , 'Dragon Knight'

        , 'Dazzle'

        , 'Clockwerk'

        , 'Leshrac'

        , 'Natures Prophet'

        , 'Lifestealer'

        , 'Dark Seer'

        , 'Clinkz'

        , 'Omniknight'

        , 'Enchantress'

        , 'Huskar'

        , 'Night Stalker'

        , 'Broodmother'

        , 'Bounty Hunter'

        , 'Weaver'

        , 'Jakiro'

        , 'Batrider'

        , 'Chen'

        , 'Spectre'

        , 'Doom'

        , 'Ancient Apparition'

        , 'Ursa'

        , 'Spirit Breaker'

        , 'Gyrocopter'

        , 'Alchemist'

        , 'Invoker'

        , 'Silencer'

        , 'Outworld Devourer'

        , 'Brewmaster'

        , 'Shadow Demon'

        , 'Lone Druid'

        , 'Chaos Knight'

        , 'Meepo'

        , 'Treant Protector'

        , 'Ogre Magi'

        , 'Undying'

        , 'Rubick'

        , 'Disruptor'

        , 'Nyx Assassin'

        , 'Naga Siren'

        , 'Keeper of the Light'

        , 'Wisp'

        , 'Visage'

        , 'Slark'

        , 'Medusa'

        , 'Troll Warlord'

        , 'Centaur Warrunner'

        , 'Magnus'

        , 'Timbersaw'

        , 'Bristleback'

        , 'Tusk'

        , 'Skywrath Mage'

        , 'Abaddon'

        , 'Elder Titan'

        , 'Legion Commander'

        , 'Ember Spirit'

        , 'Earth Spirit'

        , 'Abyssal Underlord'

        , 'Terrorblade'

        , 'Phoenix'

        , 'Techies'

        , 'Oracle'

        , 'Winter Wyvern'

        , 'Arc Warden', 'giack']

    y = dt[['Tusk']]
    Y_train = [1 if y.Tusk == 1 else -1 for y in y.itertuples()]

    ytest = dtest[['Tusk']]
    Y_test = [1 if y.Tusk == 1 else -1 for y in ytest.itertuples()]
    X = (dt.iloc[:, 0:116])

    # X_train, X_test, Y_train, Y_test = train_test_split(X, Y,test_size=0.2 , random_state=0) we do not use this because we already have a dataset divided in test and train

    X_train = np.insert(np.array(dt), 0, 1, axis=1)
    X_test = np.insert(np.array(dtest), 0, 1, axis=1)

    oPer = pc.Perceptron(0.25, 50)
    oPer.training(X_train, Y_train)

    predY = [oPer.guess(z) for z in X_test]

    cm = confusion_matrix(Y_test, predY)
    np.set_printoptions(precision=3)

    print("accuracy of a layer: %.2f%%" % (100 * accuracy_score(Y_test, predY)))

    print

    print("confusion matrix, no normalization")
    print cm
    print
    print

    print ('normalized confusion matrix')
    print cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
    print
    print

    print(classification_report(Y_test, predY))

    print("VOTED PERCEPTRON")

    o_vper = vp.VotedPerceptron(0.25, 50)

    wei_voted, i_voted, length_array = o_vper.training(X_train, Y_train)

    wei_Y_vper = [o_vper.guess(z) for z in X_test]

    accu = 100 * accuracy_score(Y_test, wei_Y_vper)
    print("accuracy of a layer voted: %.2f%%" % (accu))

    cmv = confusion_matrix(Y_test, wei_Y_vper)
    np.set_printoptions(precision=3)

    print

    print("confusion matrix voted not normalized")
    print cmv
    print
    print

    print("normalized confusion matrix")
    print cmv.astype('float') / cmv.sum(axis=1)[:, np.newaxis]
    print
    print

    print(classification_report(Y_test, wei_Y_vper))