import numpy as np
import sklearn
from sklearn.ensemble import RandomForestClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.datasets import load_iris
from sklearn.cross_validation import train_test_split

# Read some CSV file with floats, ignore header
mydata = [line.split(',') for line in open('mydata.csv').read().strip().split('\n')]

print "type(mydata)", type(mydata)

print "mydata", mydata

# Save the headers
headers = mydata[0]

# So this is mydata has a first list of headers and then
# a bunch of floats.
# All except the headers go into the matrix
data = np.array([np.array([float(value) for value in line[:-1]]) for line in mydata[1:]])

# Target (what we're predicting) goes into its own array
targets =  np.array([int(line[-1]) for line in mydata[1:]])

print "\n\nWe will be predicting these:"
print targets

print "\n\nUsing these:"
print data

# print "\n\nTo be explicit, each element in these two arrays has a 1-to-1 relationship like this:"
# print zip(targets, data)


# Sklearn can shuffle the data for us keeping the X-y correspondence intact
X_train, X_test, y_train, y_test = train_test_split(data, targets, train_size=0.8)

print "\n\ntrain_test_split both shuffles and splits the data, while maintaining the 1-to-1 relationship"

print "\n\nOur training set looks like this:"

print zip(y_train, X_train)


print "\n\nNow that the data is parsed we can start training different classifiers:"

clf = DecisionTreeClassifier()
clf.fit(X_train, y_train)

print "\n\nFor these targets in our test set:"
print y_test

print "\n\nOur classifier predicts:"
print clf.predict(X_test)

print "\n\nIt ranks the following feature importances:"

feature_importances = zip(clf.feature_importances_, headers)
top_features = sorted(feature_importances, key=lambda x: x[0], reverse=True)

for i in top_features:
	print i



# *SAMPLE RANDOM FOREST*




clf = RandomForestClassifier()

clf.fit(X_train, y_train)
clf.predict(X_test)


# *EASY EVALUATION*

from sklearn.metrics import precision_score, recall_score, accuracy_score
from sklearn.metrics import confusion_matrix, classification_report

classification_predictions = clf.predict(X_test)
classification_scores = clf.predict_proba(X_test)

print accuracy_score(y_test, classification_predictions)
print precision_score(y_test, classification_predictions)
print recall_score(y_test, classification_predictions)

print classification_report(y_test, classification_predictions)

# from sklearn.datasets import load_boston

# boston = load_boston()