# -*- coding: utf-8 -*-
"""
Created on Wed Oct  4 08:14:52 2017

@author: Miguel Almeida
"""

import numpy as np
import matplotlib.pyplot as plt
from random import shuffle
import matplotlib.patches as mpatches
from sklearn.cross_validation import train_test_split
from sklearn.cross_validation import KFold
from sklearn.cross_validation import StratifiedKFold
from sklearn.linear_model import LogisticRegression


def load_data(file_name):
    """Return matrix with orbital radius and period"""
    matrix = np.loadtxt(file_name, delimiter=',')
        
    return matrix

def poly_16features(X):
    """Expand data polynomially
    """
    X_exp = np.zeros((X.shape[0],X.shape[1]+14))
    X_exp[:,:2] = X 
    X_exp[:,2] = X[:,0]*X[:,1]
    X_exp[:,3] = X[:,0]**2
    X_exp[:,4] = X[:,1]**2
    X_exp[:,5] = X[:,0]**3
    X_exp[:,6] = X[:,1]**3
    X_exp[:,7] = X[:,0]**2*X[:,1]
    X_exp[:,8] = X[:,1]**2*X[:,0]
    X_exp[:,9] = X[:,0]**4
    X_exp[:,10] = X[:,1]**4
    X_exp[:,11] = X[:,0]**3*X[:,1]
    X_exp[:,12] = X[:,1]**3*X[:,0]
    X_exp[:,13] = X[:,0]**2*X[:,0]**2
    X_exp[:,14] = X[:,0]**5
    X_exp[:,15] = X[:,1]**5        
    return X_exp

def calc_fold(feats, X,Y, train_ix,valid_ix,C=1e12):
    """return error for train and validation sets"""
    reg = LogisticRegression(C=C, tol=1e-10)
    reg.fit(X[train_ix,:feats],Y[train_ix])
    prob = reg.predict_proba(X[:,:feats])[:,1]
    squares = (prob-Y)**2
    return np.mean(squares[train_ix]),np.mean(squares[valid_ix])


matrix = load_data("data.txt")

#shuffle
np.random.shuffle(matrix)

#normalizar os dados
means = np.mean(matrix[:,1:], axis=0)
devs = np.std(matrix[:,1:], axis=0)

matrix[:,1:] = (matrix[:,1:]-means)/devs


#split data do train
features = matrix[:,1:]
classes = matrix[:,0]

#expandir matrix
features = poly_16features(features)

feature_tr , feature_t, classe_tr, classe_t = train_test_split (features, classes, test_size = 0.33, stratify = classes)

folds = 5
stratKf = StratifiedKFold(classe_tr, folds)

errorTrain = []
errorValidation = []

for features in range(2,16):
    tr_err = va_err = 0
    for tr_ix, val_ix in stratKf:
        r, v = calc_fold(features, feature_tr,  classe_tr, tr_ix, val_ix)
        tr_err += r
        va_err += v
    errorTrain.append(tr_err/folds)
    errorValidation.append(va_err/folds)
    print(features, ':', tr_err/folds, va_err/folds)
        
        
fig = plt.plot( errorTrain, color="red", linewidth=1.0)
red_patch = mpatches.Patch(color='red', label='Training Error')

fig = plt.plot( errorValidation, color="black", linewidth=1.0)
black_patch = mpatches.Patch(color='black', label='Validation Error')

plt.legend(handles=[red_patch, black_patch])

plt.title('Train and Validation Error', fontsize=18)
plt.xlabel('', fontsize=18)
plt.ylabel('Error Value', fontsize=18)