def sim_func(results, dates, param):
model_name = param['model'][0]
x_norm = TRAINED_MODEL[model_name]['x_norm']
all_dates = TRAINED_MODEL[model_name]['dates'] + dates
all_results = subset_results_for_dates(results, all_dates)
names, X, Y, vol_normal, full_tickers, full_dates, x_norm = get_signal(all_results,
with_y=True,
vol_normal=TRAINED_MODEL[model_name]['vol_normalization'],
x_norm=x_norm,
normalize_y=False)
full_dates = [dt.strftime('%Y%m%d') for dt in full_dates]
model = TRAINED_MODEL[model_name]['model']
if not CLASSIFY:
y_pred_mean = param['y_pred_mean']
y_pred_std = param['y_pred_std']
#import pdb; pdb.set_trace()
prediction = model.predict(X)
print prediction[-10:]
trade_idx = np.where(prediction > (Trade_Thresh * y_pred_std + y_pred_mean) )
else :
prediction = model.predict(X)
print prediction[-10:]
trade_idx = np.where(prediction > 0)
pnl = np.array(Y)[trade_idx] - Cost
full_dates = np.array(full_dates)[trade_idx]
new_pnl = reorganize_pnl(pnl, full_dates, dates)
print_summary(new_pnl, 'OutSample')
return new_pnl
def sim_func(result, dates, param):
mv_back = param['mv_back'][0]
mv_front = param['mv_front'][0]
tf_back = param['tf_back'][0]
tf_front = param['tf_front'][0]
model_name = param['model_name'][0]
pre_dates = TRAINED_MODEL[model_name]['dates']
mv_all_dates = pre_dates[(-mv_back-mv_front):] + dates
mv_all_dates_dt = [datetime.datetime.strptime(dt, '%Y%m%d').date() for dt in mv_all_dates]
mv_new_result = subset_result_for_dates(result, mv_all_dates_dt)
mv_new_result_mat = to_matrix(mv_new_result)
mv_trade_dates, mv_position = generating_pos_position(mv_new_result_mat, TRAINED_MODEL[model_name]['loadings'][param['mv_factor_ind'][0]],
mv_all_dates, mv_back, mv_front, mean_revert=True)
tf_all_dates = pre_dates[(-tf_back-tf_front):] + dates
tf_all_dates_dt = [datetime.datetime.strptime(dt, '%Y%m%d').date() for dt in tf_all_dates]
tf_new_result = subset_result_for_dates(result, tf_all_dates_dt)
tf_new_result_mat = to_matrix(tf_new_result)
tf_trade_dates, tf_position = generating_pos_position(tf_new_result_mat, TRAINED_MODEL[model_name]['loadings'][param['tf_factor_ind'][0]],
tf_all_dates, tf_back, tf_front, mean_revert=True)
mv_position = np.array(mv_position)
tf_position = np.array(tf_position)
#if param['mv_value'] < 0:
#mv_position = np.zeros(mv_position.shape)
#if param['tf_value'] < 0:
#tf_position = np.zeros(tf_position.shape)
position = mv_position + tf_position
future_return = mv_new_result_mat[(mv_back+mv_front-1):,:]
pnl_mat = position * future_return
pnl = np.sum(pnl_mat, axis=1)
pos = np.sum(position, axis=1)
pnl_array = np.array(zip(dates, pnl, pos, pos, pos), dtype=[('date', 'S8'), ('total_pnl', float),
('volume', float), ('max_position', float),
('min_position', float)])
print print_summary(pnl_array, 'OutSample')
return pnl_array