def learn(self, edge, strategy):
self.display = []
with open('Model.csv', 'w', newline='') as stream:
writer = csv.writer(stream)
self.get_model().train(input_fn=self.train_input_fn(), steps=100)
evaluation_result = self.get_model().evaluate(
input_fn=self.test_input_fn())
predictions = list(
self.get_model().predict(input_fn=self.test_input_fn()))
for pred in predictions:
self.test_results_info[3].append(
round(pred['probabilities'][0], 2))
self.test_results_info[4].append(
round(pred['probabilities'][1], 2))
self.test_results_info[5].append(
round(pred['probabilities'][2], 2))
self.test_results_info[6].append(str(pred['classes']))
prediction_result = betting.test_betting_stategy(
predictions, self.test_features, self.test_labels,
self.test_results_info, edge, strategy)
writer.writerow([
evaluation_result['accuracy'],
evaluation_result['average_loss'],
prediction_result['performance']
])
self.display.append(prediction_result)
self.display.append(evaluation_result)
with open('Predictions.csv', 'w', newline='') as out:
writer = csv.writer(out)
writer.writerow(
["Home", "Away", "Date", "P(H)", "P(D)", "P(A)", "P"])
for j in range(len(self.test_results_info[0])):
writer.writerow([
self.test_results_info[0][j],
self.test_results_info[1][j],
self.test_results_info[2][j].strftime('%d-%m-%Y'),
self.test_results_info[3][j],
self.test_results_info[4][j],
self.test_results_info[5][j],
self.test_results_info[6][j].replace("b", "").replace(
"['", "").replace("']", "")
])
return self.display
def main(argv):
os.system('rm -rf model/*')
data = dataset.Dataset('data/liga.csv')
train_results_len = int(TRAINING_SET_FRACTION *
len(data.processed_results))
train_results = data.processed_results[:train_results_len]
test_results = data.processed_results[train_results_len:]
def map_results(results):
features = {}
for result in results:
for key in result.keys():
if key not in features:
features[key] = []
features[key].append(result[key])
for key in features.keys():
features[key] = np.array(features[key])
return features, features['result']
train_features, train_labels = map_results(train_results)
test_features, test_labels = map_results(test_results)
train_input_fn = tf.estimator.inputs.numpy_input_fn(x=train_features,
y=train_labels,
batch_size=500,
num_epochs=None,
shuffle=True)
test_input_fn = tf.estimator.inputs.numpy_input_fn(x=test_features,
y=test_labels,
num_epochs=1,
shuffle=False)
feature_columns = []
for mode in ['home', 'away']:
feature_columns = feature_columns + [
tf.feature_column.numeric_column(key='{}-wins'.format(mode)),
tf.feature_column.numeric_column(key='{}-draws'.format(mode)),
tf.feature_column.numeric_column(key='{}-losses'.format(mode)),
tf.feature_column.numeric_column(key='{}-goals'.format(mode)),
tf.feature_column.numeric_column(
key='{}-opposition-goals'.format(mode)),
tf.feature_column.numeric_column(key='{}-shots'.format(mode)),
tf.feature_column.numeric_column(
key='{}-shots-on-target'.format(mode)),
tf.feature_column.numeric_column(
key='{}-opposition-shots'.format(mode)),
tf.feature_column.numeric_column(
key='{}-opposition-shots-on-target'.format(mode)),
]
model = tf.estimator.DNNClassifier(
model_dir='model/',
hidden_units=[10],
feature_columns=feature_columns,
n_classes=3,
label_vocabulary=['H', 'D', 'A'],
optimizer=tf.train.ProximalAdagradOptimizer(
learning_rate=0.1, l1_regularization_strength=0.001))
with open('training-log.csv', 'w') as stream:
csvwriter = csv.writer(stream)
for i in range(0, 400):
model.train(input_fn=train_input_fn, steps=100)
evaluation_result = model.evaluate(input_fn=test_input_fn)
predictions = list(model.predict(input_fn=test_input_fn))
prediction_result = betting.test_betting_stategy(
predictions, test_features, test_labels)
csvwriter.writerow([(i + 1) * 100, evaluation_result['accuracy'],
evaluation_result['average_loss'],
prediction_result['performance']])
def main(argv):
data = dataset.Dataset('data/book.csv')
train_results_len = int(TRAINING_SET_FRACTION *
len(data.processed_results))
train_results = data.processed_results[:train_results_len]
test_results = data.processed_results[train_results_len:]
team_details = data.team_details[train_results_len:]
def map_results(results):
features = {}
for result in results:
for key in result.keys():
if key not in features:
features[key] = []
features[key].append(result[key])
for key in features.keys():
features[key] = np.array(features[key])
return features, features['result']
train_features, train_labels = map_results(train_results)
test_features, test_labels = map_results(test_results)
train_input_fn = tf.estimator.inputs.numpy_input_fn(x=train_features,
y=train_labels,
batch_size=500,
num_epochs=None,
shuffle=True)
test_input_fn = tf.estimator.inputs.numpy_input_fn(x=test_features,
y=test_labels,
num_epochs=1,
shuffle=False)
feature_columns = []
for mode in ['home', 'away']:
feature_columns = feature_columns + [
tf.feature_column.numeric_column(key='{}-wins'.format(mode)),
tf.feature_column.numeric_column(key='{}-draws'.format(mode)),
tf.feature_column.numeric_column(key='{}-losses'.format(mode)),
tf.feature_column.numeric_column(key='{}-goals'.format(mode)),
tf.feature_column.numeric_column(
key='{}-opposition-goals'.format(mode)),
tf.feature_column.numeric_column(key='{}-shots'.format(mode)),
tf.feature_column.numeric_column(
key='{}-shots-on-target'.format(mode)),
tf.feature_column.numeric_column(
key='{}-opposition-shots'.format(mode)),
tf.feature_column.numeric_column(
key='{}-opposition-shots-on-target'.format(mode)),
]
model = tf.estimator.DNNClassifier(
model_dir='model/',
hidden_units=[10],
feature_columns=feature_columns,
n_classes=3,
label_vocabulary=['H', 'D', 'A'],
optimizer=tf.train.ProximalAdagradOptimizer(
learning_rate=0.1, l1_regularization_strength=0.001))
_accuracy = -1
FILENAME = 'results.csv'
with open('training-log.csv', 'w') as stream:
csvwriter = csv.writer(stream)
for i in range(0, 2):
model.train(input_fn=train_input_fn, steps=100)
evaluation_result = model.evaluate(input_fn=test_input_fn)
predictions = list(model.predict(input_fn=test_input_fn))
prediction_result = betting.test_betting_stategy(
predictions, test_features, test_labels)
if (evaluation_result['accuracy'] > _accuracy):
_accuracy = evaluation_result['accuracy']
f = open(FILENAME, 'w')
f.write('HomeTeam,AwayTeam,H,D,A,Prediction\n')
for idx, val in enumerate(predictions):
lst = team_details[idx]
lst.extend(list(val['probabilities']))
lst.append(val['classes'][0].decode('utf-8'))
values = list(map(str, lst))
f.write(','.join(values) + '\n')
del lst
del values
f.close()
csvwriter.writerow([(i + 1) * 100, evaluation_result['accuracy'],
evaluation_result['average_loss'],
prediction_result['performance']])
