evaluate.py

import pandas
import numpy
from tobow import tobow
from sklearn.preprocessing import LabelEncoder
from keras.models import Sequential, Dense
from keras.models import model_from_json

def evaluate(model_path, weights_path, test_data_path, test_label_path, encoder_path):
    # load test data and label
    X_test = pandas.read_csv(test_data_path, header=None)
    Y_test = pandas.read_csv(test_label_path, header=None)
    test_data = X_test.values
    test_label = Y_test.values

    # load class encoder
    encoder = LabelEncoder()
    encoder.classes_ = numpy.load(encoder_path)
    json_file = open(model_path, "r")
    loaded_model_json = json_file.read()
    json_file.close()

    # load classifier model
    model = model_from_json(loaded_model_json)
    model.load_weights(weights_path)
    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

    # classification process
    labels = []
    for sentence in test_data:
        vector = tobow(sentence[0])
        predict = model.predict(numpy.array([vector[0]]))
        index = numpy.argmax(predict[0])
        labels.append(encoder.classes_[index])
    
    # evaluate accuracy
    correct = 0
    for i in xrange(len(labels)):
        print test_label[i] + "===>" + labels[i]
        if test_label[i] == labels[i]:
            correct += 1
    print "correct: " + str(correct) + " from " + str(len(test_data))
    print "Accuracy: " + str(float(correct)/len(test_data)*100)
    return correct/len(test_data)

model_path = "classifier/classifier.json"
weights_path = "classifier/weights.h5"
encoder_path = "classifier/encoder.npy"

test_data_path = "data/test/test-data.csv"
test_label_path = "data/test/test-label.csv"
evaluate(model_path,weights_path,test_data_path,test_label_path,encoder_path)