import pandas as pd
import numpy as np

# The competition datafiles are in the directory ../input
# Read competition data files:
train = pd.read_csv("../input/train.csv")
test  = pd.read_csv("../input/test.csv")

# Write to the log:
label_train=train['label']
train=train.drop('label', axis=1)
train = train / 255
test = test / 255
train['label'] = label_train

import matplotlib.pyplot as plt
from sklearn import decomposition
from sklearn import datasets
pca = decomposition.PCA(n_components=200) #Finds first 200 PCs
pca.fit(train.drop('label', axis=1))
plt.plot(pca.explained_variance_ratio_)
plt.ylabel('% of variance explained')

pca = decomposition.PCA(n_components=50) #use first 3 PCs (update to 100 later)
pca.fit(train.drop('label', axis=1))
PCtrain = pd.DataFrame(pca.transform(train.drop('label', axis=1)))
PCtrain['label'] = train['label']

PCtest = pd.DataFrame(pca.transform(test))

from sklearn.neural_network import MLPClassifier
y = PCtrain['label'][0:20000]
X=PCtrain.drop('label', axis=1)[0:20000]
clf = MLPClassifier(solver='lbfgs', alpha=1e-5,hidden_layer_sizes=(50,), random_state=1)
clf.fit(X, y)

from sklearn import  metrics
predicted = clf.predict(PCtrain.drop('label', axis=1)[20001:420000])
expected = PCtrain['label'][20001:42000]

print("Classification report for classifier %s:\n%s\n"% (clf, metrics.classification_report(expected, predicted)))
print("Confusion matrix:\n%s" % metrics.confusion_matrix(expected, predicted))


output = pd.DataFrame(clf.predict(PCtest), columns =['Label'])
output.reset_index(inplace=True)
output.rename(columns={'index': 'ImageId'}, inplace=True)
output['ImageId']=output['ImageId']+1
output.to_csv('output.csv', index=False)