def one_day_prediction_svm_18(ETFtable,
ETF_list,
window,
interval,
kernel='rbf',
gamma=1,
C=1.0,
coef0=0):
ETF_score = []
for ETF_code in ETF_list:
ETF_target = load_ETF(ETFtable, ETF_code)
ETF_target_rate = get_change_rate(ETF_target)
get_rise_fall(ETF_target_rate, interval)
X_train, y_train, X_test, y_test = load_data(ETF_target_rate, window)
# Model
clf = svm.SVC(kernel=kernel, gamma=gamma, C=C, coef0=coef0)
clf.fit(X_train, y_train)
y_predict_test = clf.predict(X_test)
rbf_test = sum(y_predict_test == y_test) * 0.5 / (y_test.shape[0] *
1.0)
y_predict_train = clf.predict(X_train)
rbf_train = sum(y_predict_train == y_train) * 0.5 / (y_train.shape[0] *
1.0)
# print('Average train one day score is ',sum(y_predict_train == y_train)*0.5/(y_train.shape[0]*1.0))
# print('Average test one day score is ',sum(y_predict_test == y_test)*0.5/(y_test.shape[0]*1.0))
print(f'{ETF_code} average one day score is ,', rbf_test)
ETF_score.append(rbf_test)
print('Total ETF one day rise_fall score is', sum(ETF_score))
print('Average ETF one day rise_fall score is', sum(ETF_score) / 18.0)
def one_day_prediction_regression_18(ETFtable, ETF_list, window, lamb,
interval):
ETF_score = []
for ETF_code in ETF_list:
ETF_target = load_ETF(ETFtable, ETF_code)
#price data
ETF_target_N = normalize_data(ETF_target)
X_train_p, y_train_p, X_test_p, y_test_p = load_data(
ETF_target_N, window)
#rate data
ETF_target_rate = get_change_rate(ETF_target)
get_rise_fall(ETF_target_rate, interval)
X_train_r, y_train_r, X_test_r, y_test_r = load_data(
ETF_target_rate, window)
### Model
lr = LinearRegressionReg()
lr.fit(X_train_p, y_train_p, lamb)
### score
p = lr.predict(X_test_p)
y_predict_de = denormalize(ETF_target, p, 'close').reshape(
(y_test_p.shape[0]))
y_predict_rise_fall = []
for day, price in enumerate(y_predict_de):
if (day != 0):
y_predict_rise_fall.append(
sign(y_predict_de[day] - y_predict_de[day - 1]))
y_predict_rise_fall = np.array(y_predict_rise_fall).astype(np.float64)
if (y_test_r.shape[0] > y_predict_rise_fall.shape[0]):
y_test_r_short = y_test_r[y_test_r.shape[0] -
y_predict_rise_fall.shape[0]:]
score = (sum(y_test_r_short == y_predict_rise_fall) *
0.5) / (y_test_r_short.shape[0] * 1.0)
else:
score = (sum(y_test_r == y_predict_rise_fall) *
0.5) / (y_test_r.shape[0] * 1.0)
print(f'{ETF_code} average one day score is', score)
ETF_score.append(score)
print('Total ETF one day rise_fall score is', sum(ETF_score))
print('Average ETF one day rise_fall score is', sum(ETF_score) / 18.0)
return ETF_score
def five_day_prediction_18(ETFtable, ETF_list, window, lamb):
ETF_score = []
ETF_week_score = []
for ETF_code in ETF_list:
ETF_target = load_ETF(ETFtable, ETF_code)
ETF_target_N = normalize_data(ETF_target)
fiveday_score_array, fiveday_score = week_score(
ETF_target, ETF_target_N, window, lamb)
print(f'{ETF_code} average Fiveday_score is ', fiveday_score)
ETF_score.append(fiveday_score)
ETF_week_score.append(fiveday_score_array)
print('Total ETF five day price score is', sum(ETF_score))
print('Average ETF five day price score is', sum(ETF_score) / 18.0)
return ETF_week_score
def predict_ETF_close(ETFtable, ETF_list, window, lamb):
ETF_week_close = []
ETF_week_score = []
ETF_lastday_close = []
for ETF_code in ETF_list:
ETF_target = load_ETF(ETFtable, ETF_code)
ETF_target_N = normalize_data(ETF_target)
close_price, lastday_close = get_fiveday_close(ETF_target_N,
ETF_target, window,
lamb)
ETF_week_close.append(close_price)
ETF_lastday_close.append(lastday_close)
ETF_week_close = np.array(ETF_week_close)
ETF_lastday_close = np.array(ETF_lastday_close)
return ETF_week_close, ETF_lastday_close
def one_day_prediction_18(ETFtable, ETF_list, window, lamb):
ETF_score = []
ETF_model = []
for ETF_code in ETF_list:
ETF_target = load_ETF(ETFtable, ETF_code)
ETF_target_N = normalize_data(ETF_target)
X_train, y_train, X_test, y_test = load_data(ETF_target_N, window)
### Model
lr = LinearRegressionReg()
lr.fit(X_train, y_train, lamb)
### score
score = lr.score(X_test, y_test, ETF_target)
print(f'{ETF_code} average one day score is', score)
ETF_score.append(score)
ETF_model.append(lr)
ETF_features = ETF_target.columns
print('Total ETF one day price score is', sum(ETF_score))
print('Average ETF one day price score is', sum(ETF_score) / 18.0)
return [ETF_score, ETF_model, ETF_features]