import numpy as np
from math260 import data_prep, recommend, score
GAMES_FILE = "data/games.csv"
REVIEWS_FILE = "data/reviews.csv"
if __name__ == "__main__":
games, users = data_prep.parse_data(GAMES_FILE, REVIEWS_FILE, verbose=True)
games_map, users_map, rating_matrix, bool_matrix \
= data_prep.create_review_matrix(games, users, sparse=False, verbose=True)
# testing removing 10% from each user and predicting using average score
def rf():
return np.random.randint(0, 10)
predictor = recommend.RandomPredictor(rf)
rmse, errors = score.rmsecv(0.1, rating_matrix, bool_matrix,
predictor.predict)
print('RMSE:\t\t{}'.format(rmse))
def cosine_similarity(u1, u2, rating_matrix, bool_matrix):
return 1 - dist.cosine(rating_matrix[u1], rating_matrix[u2])
if __name__ == "__main__":
games, users = data_prep.parse_data(GAMES_FILE, REVIEWS_FILE, verbose=True)
games_map, users_map, rating_matrix, bool_matrix \
= data_prep.create_review_matrix(games, users, sparse=False, verbose=True)
results = []
for i in range(25):
k = 20 * (i+1)
print(f'using {k} neighbors')
cosine_predictor = recommend.SimilarityPredictor(cosine_similarity, k)
msd_predictor = recommend.SimilarityPredictor(msd_similarity, k)
cosine_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
cosine_predictor.predict, users=range(0, 1000))
sim_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
msd_predictor.predict, users=range(0, 1000))
result = {'k':k, 'cosine': cosine_rmse, 'sim': sim_rmse}
results.append(result)
print(result)
json.dump(results, open(output, mode='w'))
combo_predictor = recommend.SimilarityPredictor(combo_sim, 120)
adjusted_cosine_predictor = recommend.SimilarityPredictor(adjusted_cosine_similarity, 120)
cosine_predictor = recommend.SimilarityPredictor(cosine_similarity, 120)
msd_predictor = recommend.SimilarityPredictor(msd_similarity, 120)
jaccard_predictor = recommend.SimilarityPredictor(jaccard_similarity, 120)
avg_predictor = recommend.AveragePredictor(rating_matrix, bool_matrix)
uavg_predictor = recommend.AveragePredictor(rating_matrix, bool_matrix)
gavg_predictor = recommend.GlobalAveragePredictor(rating_matrix, bool_matrix)
tavg_predictor = recommend.TwoWayAveragePredictor(rating_matrix, bool_matrix)
## adjusted_cosine_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
## adjusted_cosine_predictor.predict, users=range(0, 1000))
## cosine_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
## cosine_predictor.predict, users=range(0, 1000))
combo_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
combo_predictor.predict, users=range(0, 1000))
jaccard_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
jaccard_predictor.predict, users=range(0, 1000))
sim_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
msd_predictor.predict, users=range(0, 1000))
avg_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
avg_predictor.predict, users=range(0, 1000))
uavg_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
uavg_predictor.predict, users=range(0, 1000))
gavg_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
gavg_predictor.predict, users=range(0, 1000))
tavg_rmse, _ = score.rmsecv(0.1, rating_matrix, bool_matrix,
tavg_predictor.predict, users=range(0, 1000))
# print(f"Adjusted cosine predictor RMSE: {adjusted_cosine_rmse}")
# print(f"Cosine predictor RMSE: {cosine_rmse}")