logistic_regression

gistfile1.txt

import math ;
import matplotlib.pyplot as plt
import re;

data = [] ;


class instance:
	def __init__(self,array):
		self.id = array[0]
		self.label = int(array[-1]) ;
		self.feature = [int(i) for i in array[1:-1]]
	

for line in open("dataset.txt","r"):
	if line[0]=='#':
		continue ;
	i = instance(re.split('\s+',line[:-1])) ;
	data.append(i) ;


weights = [0.0 for i in range(5)] ;

rate = 0.0001 ;

max_iter = 3000 ;

def sigmoid(value):
	return 1/(1+math.exp(-value))


def classify(instance):
	logit = 0.0 ;
	for index in range(len(weights)):
#		print index , instance.feature, instance.id
		logit = logit + weights[index]*instance.feature[index] ;
	return sigmoid(logit) ;



for i in range(max_iter):
	global weights ;
	lik = 0.0
	for n ,instance in enumerate(data):
		predicted = classify(instance)
		for index,weight in enumerate(weights):
			weights[index] = weight+rate*(instance.label-predicted)*instance.feature[index] ;

		lik = instance.label*math.log(classify(instance))+(1-instance.label)*math.log(1-classify(instance))

	print "iterator ",n , weights , lik 

		
for i in data:
	print classify(i), i.label