# %% Choose settings and classifier
test_season = '2018-2019' # hold-out season for validation
level = 'match' # match or team level features to use
shuffle = True # whether to shuffle or not the data in k-fold cross validation
norm = True # whether to normalise or not the features
min_round = 5 # minimum number of first rounds to skip in every season
nsplits = 5 # number of folds in k-fold cross validation
random_state = 10 # random state for the classifier
params = {
'n_estimators': np.arange(5, 200, 5),
# 'learning_rate': np.arange(0.3, 1.5, 0.1)}
} # params for the grid search
model = AdaBoostClassifier(random_state=random_state)
# %% load feature data
df = load_features(level)
# choose features
if level == 'match':
feats = [
'Position_x', 'Position_y', 'Offence_x', 'Offence_y', 'Defence_x',
'Defence_y', 'form_x', 'form_y', 'Diff_x', 'Diff_y', 'Home F4',
'Away F4'
]
elif level == 'team':
feats = [
'Home', 'Away', 'Position', 'Offence', 'Defence', 'form', 'F4', 'Diff'
]
n_feats = len(feats)
# seasons for calibration
knn = ItemBase(reviews_data)
result = knn.get_recommended_games(game_id, games_data)
return render_template(
'predict.html',
tables=[result.to_html(index=False, classes='table')],
titles="Suggestion for " + name + "'s players")
else:
username = request.form.get('username')
result = user_base.get_predict(username)
return render_template(
'predict.html',
tables=[result.to_html(index=False, classes='table')],
titles="Suggestion for user:"******"No suggestion for ")
if __name__ == '__main__':
games_data = data_processing.load_games_data()
game_features = data_processing.load_features(games_data,
method='Content-based')
reviews_data = data_processing.load_reviews()
tf_idf = TF_IDF(games_data, game_features)
user_base = UserBase(reviews_data, games_data)
app.run(debug=True)