final_svm.py

#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Created on Wed Jul  5 00:19:18 2017

@author: saliksyed
"""
import json
from sklearn import svm

routes = json.loads(open("routes.json").read())

country_to_country_code = json.loads(open("country_name_to_country_code.json").read())

populations = json.loads(open("population_by_country_code.json").read())

gdps = json.loads(open("gdp_by_country_code.json").read())

country_code_to_feature = {}
for country in routes:
    if country in country_to_country_code:
        code = country_to_country_code[country]
        to_amount = sum(routes[country]["to"].values())
        from_amount = sum(routes[country]["from"].values())
        country_code_to_feature[code] = [to_amount, from_amount]
        

X = []
y = []

# read the data in to the X, y vectors:
for code in country_code_to_feature:
    if populations[code] != None: # skip examples without gdp data
        X.append(country_code_to_feature[code])
        if populations[code] > 100000000:
            y.append(1)
        else:
            y.append(0)
    

# now we'll keep most of the data is training data, and keep the rest to "test"
# if our algorithm actually works on new examples it hasn't seen before

num_train = int(len(X)/2.0)

X_train = X[:num_train]
y_train = y[:num_train]

X_test = X[num_train:]
y_test = y[num_train:]
    
clf = svm.SVC()
clf.fit(X_train, y_train)


predictions = clf.predict(X_test)
print predictions

# Now let's check our predictions:
correct = 0
for i, prediction in enumerate(predictions):
    if prediction == y_test[i]:
        correct += 1

print "Percentage correct:"
print float(correct) / len(predictions) * 100