naive bayes implementation by Ajay

06-Naive-Bayes/bayes_msg_filter.py

@@ -0,0 +1,133 @@
+
+from __future__ import division
+from collections import defaultdict
+import pickle
+import re
+import sys
+import math
+
+count  = 0
+
+
+def rr():
+    return 0.000000000000000000000001
+
+def dd():
+    return defaultdict(rr) 
+
+def train_line(line,data_dict,prior):
+    (label,text) = line.split("\t")
+    # Split the text by the punctuation words
+    text = re.split(r"[^\w]|[\s]",text)
+    prior[label.lower()] += 1
+    for word in text:
+        if word == "":
+            continue
+        data_dict[label.lower()][word.lower()] += 1
+
+
+def normalize_prob(data_dict,prior):
+    ham_word_count = sum(data_dict["ham"].itervalues())
+    spam_word_count = sum(data_dict["spam"].itervalues())
+    for word in data_dict["spam"]:
+        data_dict["spam"][word] /= spam_word_count
+    for word in data_dict["ham"]:
+        data_dict["ham"][word] /= ham_word_count
+
+    prior["ham"] /= (prior["ham"] + prior["spam"])
+    prior["spam"] /= (prior["ham"] + prior["spam"])
+
+def classify_msg(msg,data_dict,prior):
+    spam_prob= 0.0
+    ham_prob  = 0.0
+
+    words = re.split("[^\w]|[/s]",msg)
+    #print words
+    for word in words: 
+        word=word.lower()
+        if word == "":
+            continue
+        ham_prob += math.log10(data_dict["ham"][word])
+        spam_prob += math.log10(data_dict["spam"][word])
+
+    ham_prob += math.log10(prior["ham"])
+    spam_prob += math.log10(prior["spam"])
+
+        #print "Word :{}--Spam Prob->{} Ham Prob -> {}".format(word,spam_prob,ham_prob)
+    if spam_prob > ham_prob:
+        guess = "spam"
+    else:
+        guess = "ham"
+
+    return guess
+
+
+
+def train(filename):
+    data_dict = defaultdict(dd)
+    prior = defaultdict(rr)
+    with open(filename,'r') as inp:
+        for line in inp:
+            train_line(line,data_dict,prior)
+
+    normalize_prob()
+    #print ham_words
+    with open("m_brain", "wb") as fout:
+        pickle.dump((data_dict,prior), fout)
+
+
+def test(filename):
+    with open("m_brain","r") as fin:
+        (data_dict,prior) = pickle.load(fin)
+
+
+    count  = 0
+    success = 0
+    with open(filename, "r") as inp:
+        for line in inp:
+            count +=1
+            label,text = line.split("\t")
+            guess = classify_msg(text,data_dict,prior)
+            if label.lower() == guess:
+                success += 1
+
+
+    print "Success rate = {}/{}".format(success,count)            
+    print "Success rate %= {}".format(success/count * 100)                      
+
+def train_and_test():
+    data_dict = defaultdict(dd)
+    prior = defaultdict(rr)
+
+    with open("corpus/SMSSpamCollection.txt") as inp:
+        lines = inp.readlines()
+        train_len = 5550
+        train_data = lines[:train_len]
+        test_data = lines[train_len:]
+        success = 0
+        for line in train_data:
+            train_line(line,data_dict,prior)
+        normalize_prob(data_dict,prior)
+
+        for line in test_data:
+            label,text = line.split("\t")
+            guess = classify_msg(text,data_dict,prior)
+            if label.lower() == guess:
+                success += 1
+
+    print "Success rate = {}/{}".format(success,len(test_data))            
+    print "Success rate %= {}".format(success/len(test_data) * 100)            
+
+def main():
+    if len(sys.argv) == 3:
+        if sys.argv[1] == "-train" or sys.argv[1] == "--t":
+            train(sys.argv[2])
+        elif sys.argv[1] == "-run" or sys.argv[1] == "--r":
+            test(sys.argv[2])
+    else:
+        train_and_test()
+
+
+
+if __name__ == "__main__":
+    main()