# Enrico Costanzo (201124039)

import os
from math import pow, sqrt, fabs
from numpy import std, mean

clear = lambda: os.system('cls')

class Point:
    def __init__(self, x, y, c):
        self.x = x
        self.y = y
        self.c = c # class 1 or -1

class ResultObject:
    def __init__(self,
                 mean_x_class0, mean_y_class0, mean_x_class1, mean_y_class1,
                 stddev_x_class0, stddev_y_class0, stddev_x_class1, stddev_y_class1):
        self.mean_x_class0 = mean_x_class0
        self.mean_y_class0 = mean_y_class0
        self.mean_x_class1 = mean_x_class1
        self.mean_y_class1 = mean_y_class1

        self.stddev_x_class0 = stddev_x_class0
        self.stddev_y_class0 = stddev_y_class0
        self.stddev_x_class1 = stddev_x_class1
        self.stddev_y_class1 = stddev_y_class1

def die(error_message):
    raise Exception(error_message)

def read_data(filename):
    f = open(filename)
    data_line = False
    data = []
    for l in f:
        l = l.strip()
        if data_line:
            content = [float(x) for x in l.split(',')]
            if len(content) == 3:
                data.append(content)
        else:
            if l.startswith('@DATA'):
                data_line = True
    return data

def xmean(data, axis): # axis: 0=x 1=y
    tmp = []
    for d in data:
        tmp.append(d.x) if axis == 0 else tmp.append(d.y)

    return mean(tmp)

def stddev(data, axis): # axis: 0=x 1=y
    tmp = []
    for d in data:
        tmp.append(d.x) if axis == 0 else tmp.append(d.y)

    return std(tmp)


def train(data):
    class0 = [] # Class 1
    class1 = [] # Class -1

    for d in data:
        p = Point(d[0], d[1], d[2])
        class0.append(p) if(p.c > 0) else class1.append(p)

    # Mean X and Y for Class 1
    mean_x_class0 = xmean(class0, 0) #x
    mean_y_class0 = xmean(class0, 1)

    # Mean X and Y for Class -1
    mean_x_class1 = xmean(class1, 0) #x
    mean_y_class1 = xmean(class1, 1)

    # Standard Deviation X and Y for Class 1
    stddev_x_class0 = stddev(class0, 0) #x
    stddev_y_class0 = stddev(class0, 1)

    # Standard Deviation X and Y for Class -1
    stddev_x_class1 = stddev(class1, 0) #x
    stddev_y_class1 = stddev(class1, 1)

    res = ResultObject(
        mean_x_class0, mean_y_class0, mean_x_class1, mean_y_class1,
        stddev_x_class0, stddev_y_class0, stddev_x_class1, stddev_y_class1
    )

    return res

def test(p, r):
    # Vorhersage anhand der Standardabweichung
    if (fabs(p.x - r.mean_x_class0) < fabs(p.x - r.mean_x_class1)) or (fabs(p.y - r.mean_y_class0) < fabs(p.y - r.mean_y_class1)):
        prediction1 = 1
    else:
        prediction1 = -1

    # Vorhersage anhand des Durchschnittswertes
    l1 = sqrt(fabs(pow(p.x - r.mean_x_class1, 2)) + pow(fabs(p.y - r.mean_y_class1), 2))
    l2 = sqrt(fabs(pow(p.x - r.mean_x_class0, 2)) + pow(fabs(p.y - r.mean_y_class0), 2))

    prediction2 = -1 if l1 <= l2 else 1

    if prediction1 == prediction2:
        # 2/2 richtige Vorhersagen
        # Egal welche, da beide gleich sind
        prediction = prediction1
    elif prediction1 != prediction2:
        # 1/2 richtige Vorhersagen
        #Im Zweifel verlassen wir uns auf den Durchschnittswert
        prediction = prediction2

    return prediction == p.c

def print_results(r):
    clear()
    print
    print "Class -1"
    print
    print "Attribute X"
    print "Mean:   ", r.mean_x_class1
    print "StdDev: ", r.stddev_x_class1
    print
    print "Attribute Y"
    print "Mean:   ", r.mean_y_class1
    print "StdDev: ", r.stddev_y_class1
    print
    print
    print "Class 1"
    print
    print "Attribute X"
    print "Mean:   ", r.mean_x_class0
    print "StdDev: ", r.stddev_x_class0
    print
    print "Attribute Y"
    print "Mean:   ", r.mean_y_class0
    print "StdDev: ", r.stddev_y_class0

def eval_data(data, r):
    error_count = 0
    num_samples = 0

    for d in data:
        p = Point(d[0], d[1], d[2])

        num_samples += 1

        if not test(p, r):
            error_count += 1

    return error_count, num_samples

if __name__ == "__main__":
    train_data = read_data("train.arff")
    train_results = train(train_data)
    print_results(train_results)

    test_data = read_data("eval.arff")
    error_absolute, num_samples = eval_data(test_data, train_results)
    error_quota = ((100 * error_absolute) / num_samples)

    print
    print "Correctly Classified Instances\t\t"  ,num_samples - error_absolute, "\t", 100 - error_quota ,"%"
    print "Incorrectly Classified Instances\t",error_absolute,      "\t", error_quota,"%"
    print "Number of samples\t\t\t", num_samples