run.py

import csv
import glob
import numpy as np

from src.init import*
from src.utils import*
from src.train_or_evaluate import preprocess_dataset


def major_vote():
    """
    Performs majority voting on prediction files + creates a final submission file
    """
    pred_f = glob.glob('models/neural_netwroks/predictions/*.csv')
    preds = np.zeros((10000, 2))
    
    for file in pred_f:
        with open(file, 'r') as f:
            lines = f.readlines()[1:]
            currents = np.array([int(l.split(',')[1]) for l in lines])
            currents[currents < 0] = 0
            preds[range(10000), currents] += 1

    preds = np.argmax(preds, axis=1)
    preds[preds < 1 ] = -1

    path = os.path.join('models/neural_networks/predictions', f'pred_cnn.csv')
    create_submission(preds, path)

    return 0



def create_submission(y_pred, path):
    ids=[i for i in range(1, len(y_pred)+1)]
    with open(path, 'w', newline='') as csvfile:
        fileds = ['Id', 'Prediction']
        w = csv.DictWriter(csvfile, delimiter=",", fieldnames=fileds)
        w.writeheader()
        for r1, r2 in zip(ids, y_pred):
            w.writerow({'Id':int(r1),'Prediction':int(r2)})



if __name__ == '__main__':
    cnn = load_best("cnn", 'normal')
    # lstm = load_best("lstm", 'normal')
    # bi_lstm = load_best("bi_lstm", 'normal')
    # gru = load_best("gru", 'normal')

    to_pred = preprocess_tweets_to_predict()
    _, _, _, _, _, tokenizer, _, _, _ = preprocess_dataset('normal', 200, 100)

    pred_cnn = compute_model_predictions(to_pred, cnn, tokenizer, maxlen=100)
    # pred_lstm = compute_model_predictions(to_pred, lstm, tokenizer, maxlen=50)
    # pred_bi_lstm = compute_model_predictions(to_pred, bi_lstm, tokenizer, maxlen=100)
    # pred_gru = compute_model_predictions(to_pred, gru, tokenizer, maxlen=100)

    path = os.path.join('models/neural_networks/predictions', f'pred_cnn.csv')
    create_submission(pred_cnn, path)
    # create_submission(pred_lstm, "'models/neural_netwroks/predictions/pred_lstm.csv")
    # create_submission(pred_bi_lstm, "'models/neural_netwroks/predictions/pred_bi_lstm.csv")
    # create_submission(pred_gru, "'models/neural_netwroks/predictions/pred_gru.csv")

    major_vote()