Ejemplo n.º 1
0
def create_trained_model():
    genres = [
        'classical',
        'jazz',
        # 'opera',
        # 'reggae',
        'rock',
        'house',
        'hip hop'
    ]
    required_features = [
        'Centroid_AVG',
        'Centroid_SD',
        'RollOff_AVG',
        'RollOff_SD',
        'Flux_AVG',
        'Flux_SD',
        'BPM',
        'Noise',
        'NoiseTwo'
    ]
    db_service = DatabaseService()
    training_set = db_service.get_training_set(genres, required_features, 100, verbose=True)
    learner = LearnerSVC()
    learner.fit(training_set)

    save_model(learner, features=required_features, genres=genres)
Ejemplo n.º 2
0
def main(argv):

    ####################################################
    #   Constants
    ####################################################

    songs_per_genre = 200

    # songs_per_genre -100
    # c, r, h, hh = %83.59
    # c, j, r, h, hh = %77.83

    genres = [
        'classical',
        'jazz',
        # 'opera',
        # 'reggae',
        'rock',
        'house',
        'hip hop'
    ]
    required_features = [
        'Centroid_AVG',
        'Centroid_SD',
        'RollOff_AVG',
        'RollOff_SD',
        'Flux_AVG',
        'Flux_SD',
        'BPM',
        'Noise',
        # 'NoiseTwo'
    ]

    ####################################################
    #   Training Set Fetch
    ####################################################

    db_service = DatabaseService()

    training_set = db_service.get_training_set(genres, required_features, songs_per_genre, verbose=True)

    cross_val = CrossValidation(training_set, k_folds=6)

    average_hit_rate = 0
    count = 0

    validator = Validator()
    for ts, vs in cross_val:
        learner = LearnerSVC()
        learner.fit(ts)
        results = validator.validate_next(learner, vs)

        average_hit_rate += results.hit_rate()
        count += 1.0
        print results

    print "Total samples: " + str(training_set.total_samples())
    print "Training/Validation: " + str(cross_val.training_size()) + " / " + str(cross_val.validation_size())
    print "Average hit rate: %" + "%.2f" % (100 * average_hit_rate / count)