for i in range(min_num_examples,max_num_examples,10): (train_error,test_error) = train_and_test(i) all_train_errors.append(train_error) all_test_errors.append(test_error) num_examples.append(i) plt.plot(num_examples, all_train_errors, 'r^', label="Training Error") plt.plot(num_examples, all_test_errors, "bo", label="Test Error") plt.legend() plt.show() #sets dp.qb_perfs with performances from given QBs dp.make_perfs_bag(season_long_qbs) TOTAL_PERF_RECORDS = len(dp.qb_perfs) print TOTAL_PERF_RECORDS def train_and_test(num_examples): TOTAL_TRAINING_RECORDS = int(TRAINING_RATIO * num_examples) TOTAL_TEST_RECORDS = int((1-TRAINING_RATIO) * num_examples) #TRAINING selectedTrainingPerfs = dp.train_gameday_LR(TOTAL_TRAINING_RECORDS) #Calculate training error training_error_sum = 0 for idx in selectedTrainingPerfs: actual_perf_score = get_actual_perf(idx) predicted_perf_score = predict_perf_gameday(idx).item(0)
for i in range(min_num_examples, max_num_examples, 10):
(train_error, test_error) = train_and_test(i)
all_train_errors.append(train_error)
all_test_errors.append(test_error)
num_examples.append(i)
plt.plot(num_examples, all_train_errors, 'r^', label="Training Error")
plt.plot(num_examples, all_test_errors, "bo", label="Test Error")
plt.legend()
plt.show()
#sets dp.qb_perfs with performances from given QBs
dp.make_perfs_bag(season_long_qbs)
TOTAL_PERF_RECORDS = len(dp.qb_perfs)
print TOTAL_PERF_RECORDS
def train_and_test(num_examples):
TOTAL_TRAINING_RECORDS = int(TRAINING_RATIO * num_examples)
TOTAL_TEST_RECORDS = int((1 - TRAINING_RATIO) * num_examples)
#TRAINING
selectedTrainingPerfs = dp.train_gameday_LR(TOTAL_TRAINING_RECORDS)
#Calculate training error
training_error_sum = 0
for idx in selectedTrainingPerfs:
actual_perf_score = get_actual_perf(idx)
for i in range(min_num_examples,max_num_examples,5): (train_error,test_error) = train_and_test(i) all_train_errors.append(train_error) all_test_errors.append(test_error) num_examples.append(i) plt.plot(num_examples, all_train_errors, 'r^', label="Training Error") plt.plot(num_examples, all_test_errors, "bo", label="Test Error") plt.legend() plt.show() # Obtain season data dp.make_perfs_bag() TOTAL_PERF_RECORDS = len(dp.qb_perfs) def train_and_test(num_examples): TOTAL_TRAINING_RECORDS = int(TRAINING_RATIO * num_examples) TOTAL_TEST_RECORDS = int((1-TRAINING_RATIO) * num_examples) #TRAINING selectedTrainingPerfs = dp.trainGamedayLR(TOTAL_TRAINING_RECORDS) #Calculate training error training_error_sum = 0 for idx in selectedTrainingPerfs: actual_perf_score = get_actual_perf(idx) predicted_perf_score = predict_perf_gameday(idx).item(0) training_error_sum += abs(actual_perf_score - predicted_perf_score)
