def predict(self, test_file, output_file):
total = 0.0
wfp = open(output_file, "w")
fp = open(test_file)
positive = 0
negative = 0
true_positive = 0
false_positive = 0
true_negative = 0
false_negative = 0
sys.stderr.write("start to predict ...\n")
me = model_evaluate.ModelEvaluate()
while True:
line = fp.readline()
if len(line) <= 0:
break
arr = line.strip('\r\n').split('\t')
label = int(arr[0])
test_sample = Sample()
test_sample.load_sample(arr[1:], label, self.__word_dict)
p = self.__classifier(test_sample)
wfp.write("%d\t%f\n" % (label, p))
total += 1
if p > 0.5:
p = 1
else:
p = -1
me.add(int(label), p)
me.report()
fp.close()
wfp.close()
parser.add_argument("-t", "--test_file", help="test file")
parser.add_argument("-o", "--output_file", help="output file")
args = parser.parse_args()
train_file = args.train_file
test_file = args.test_file
dict_file = args.dict_file
output_file = args.output_file
if not train_file or not os.path.exists(train_file):
parser.print_help()
sys.exit(1)
if not test_file or not os.path.exists(test_file):
parser.print_help()
sys.exit(1)
if not dict_file or not os.path.exists(dict_file):
parser.print_help()
sys.exit(1)
me = model_evaluate.ModelEvaluate()
naive_bayes = NaiveBayes()
naive_bayes.train_model(train_file, dict_file)
fp = open(test_file)
while True:
line = fp.readline()
if len(line) <= 0:
break
label, words = naive_bayes.get_articel_words(line)
l, p = naive_bayes.predict(words)
me.add(int(label), int(l))
me.report()