/
mft2_sim_experiment.py
270 lines (235 loc) · 12 KB
/
mft2_sim_experiment.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
from mfipy.roll_sim import RollSim, print_summary
from mfipy.data_loader import readCSV
import numpy as np
import os.path as op
from sklearn.grid_search import GridSearchCV
from sklearn.cross_validation import StratifiedKFold, KFold
from sklearn import svm
from sklearn.metrics import mean_squared_error, r2_score, classification_report
from collections import defaultdict
from matplotlib import mlab
import datetime
from sklearn.decomposition import PCA
from sklearn.grid_search import IterGrid
TRAINED_MODEL = {}
mv_back_range = range(11, 12, 1)
mv_front_range = range(11, 12, 1)
mv_factor_range = [4]
mv_param_space = {'back': mv_back_range, 'front': mv_front_range}
#mv_param_list = list(IterGrid(mv_param_space))
mv_param_list = [{'back':item[0], 'front':item[1]} for item in zip(mv_back_range, mv_front_range)]
tf_back_range = range(1, 2, 1)
tf_front_range = range(1, 2, 1)
tf_factor_range = [3]
tf_param_space = {'back': tf_back_range, 'front': tf_front_range}
#tf_param_list = list(IterGrid(tf_param_space))
tf_param_list = [{'back':item[0], 'front':item[1]} for item in zip(tf_back_range, tf_front_range)]
def open_to_open_ret(rec, name=''):
ret = np.diff(rec['Open']) / rec['Open'][:-1]
dates = rec['Date'][1:]
new_rec = np.array(zip(dates, ret), dtype=[('Date', datetime.date),
('Ret_'+name, float)])
return new_rec
def recs_inner_join(key, recs, postfixes):
new_rec = recs[0]
for ii in range(1, len(recs)):
r1postfix='1'
r2postfix='2'
new_rec = mlab.rec_join(key, new_rec, recs[ii], jointype='inner',
defaults=None, r1postfix=r1postfix, r2postfix=r2postfix)
return new_rec
def to_matrix(rec):
new_rec = mlab.rec_drop_fields(rec, ['Date'])
new_mat = np.c_[[new_rec[nm] for nm in new_rec.dtype.names]].T
return new_mat
class Mft2RollSim(RollSim):
def __init__(self, name, train_n, sim_n):
super(Mft2RollSim, self).__init__(name, train_n, sim_n)
def load_all_data(self, filenames, ret_func, min_date='20040101'):
result = {}
dates = []
for fname in filenames:
assert 'daily_' in fname
ticker = op.split(fname)[-1].split('_')[1]
rec = readCSV(fname, date_name='Date', date_format='%Y%m%d')
rec = np.sort(rec, 0, order=['Date'])
result[ticker] = ret_func(rec, ticker)
tickers, recs = zip(*result.items())
self._tickers = tickers
self._result = recs_inner_join('Date', recs,
postfixes=tickers)
self._result = np.sort(self._result, 0, order=['Date'])
self._dates = [dt.strftime('%Y%m%d') for dt in self._result['Date'] if dt.strftime('%Y%m%d') >= min_date ]
print "Tickers: "
print " ", ", ".join(result.keys())
def subset_result_for_dates(result, dates):
new_result = None
for adate in dates:
if new_result is None:
new_result = result[np.where(result['Date'] == adate)]
else :
new_result = np.append(new_result, result[np.where(result['Date'] == adate)])
return new_result
def auto_correlation_score(mat, loading, back, front):
factor = np.dot(mat, loading)
nn = len(factor)
back_weights = np.ones(back) * 1.0 / back
front_weights = np.ones(front) * 1.0 / front
factor_back = np.convolve(factor, back_weights)[(back-1):(nn-front)]
factor_front = np.convolve(factor, front_weights)[(back+front-1):nn]
value = np.mean(factor_back * factor_front) / np.std(factor_back) / np.std(factor_front)
return value
def pos_trading_score(mat, loading, back, front, mean_revert=True):
factor = np.dot(mat, loading)
nn = len(factor)
back_weights = np.ones(back) * 1.0 / back
front_weights = np.ones(front) * 1.0 / front
factor_back = np.convolve(factor, back_weights)[(back-1):(nn-front)]
pos_loading = np.copy(loading)
pos_loading[np.where(loading < 0)] = 0.0
pos_factor_front = np.convolve(np.dot(mat, pos_loading), front_weights)[(back+front-1):nn]
neg_loading = np.copy(loading) * (-1.0)
neg_loading[np.where(loading < 0)] = 0.0
neg_factor_front = np.convolve(np.dot(mat, neg_loading), front_weights)[(back+front-1):nn]
if mean_revert:
trade_factor = np.copy(pos_factor_front)
trade_factor[np.where(factor_back) > 0] = neg_factor_front[np.where(factor_back) > 0]
value = np.mean(np.fabs(factor_back) * trade_factor) / np.std(factor_back) / np.std(trade_factor)
else :
trade_factor = np.copy(pos_factor_front)
trade_factor[np.where(factor_back) < 0] = neg_factor_front[np.where(factor_back) < 0]
value = np.mean(np.fabs(factor_back) * trade_factor) / np.std(factor_back) / np.std(trade_factor)
return value
def generating_pos_position(mat, loading, dates, back, front, mean_revert=True):
factor = np.dot(mat, loading)
nn = len(factor)
back_weights = np.ones(back) * 1.0 / back
factor_back = np.convolve(factor, back_weights)[(back-1):nn]
date_used = dates[(back-1):nn]
pos_loading = np.copy(loading)
pos_loading[np.where(loading < 0)] = 0.0
neg_loading = np.copy(loading)
neg_loading[np.where(loading > 0)] = 0.0
pos_loading *= 10.0
neg_loading *= -10.0
position = []
for ii in range(front-1, len(date_used)):
cur_pos = np.zeros(len(pos_loading))
for jj in range(front):
signal = factor_back[ii-jj]
if mean_revert:
if signal > 0:
cur_pos = cur_pos + signal * neg_loading
else :
cur_pos = cur_pos + signal * (-1.0) * pos_loading
else :
if signal > 0:
cur_pos = cur_pos + signal * pos_loading
else :
cur_pos = cur_pos + signal * (-1.0) * neg_loading
position.append(cur_pos)
date_trade = date_used[(front-1):]
return date_trade[:-1], position[:-1]
def train_func(result, dates):
dates_dt = [datetime.datetime.strptime(dt, '%Y%m%d').date() for dt in dates]
new_result = subset_result_for_dates(result, dates_dt)
new_result_mat = to_matrix(new_result)
pca = PCA()
pca.fit(new_result_mat)
mv_est = []
mv = True
for ind in mv_factor_range:
for param in mv_param_list:
value = pos_trading_score(new_result_mat, pca.components_[ind], param['back'], param['front'], mv)
mv_est.append((ind, param['back'], param['front'], mv, value))
mv_est = np.array(mv_est, dtype=[('factor_ind', int), ('back', int), ('front', int), ('mean_revert', bool), ('value', float)])
tf_est = []
mv = False
for ind in tf_factor_range:
for param in tf_param_list:
value = pos_trading_score(new_result_mat, pca.components_[ind], param['back'], param['front'], mv)
tf_est.append((ind, param['back'], param['front'], mv, value))
tf_est = np.array(tf_est, dtype=[('factor_ind', int), ('back', int), ('front', int), ('mean_revert', bool), ('value', float)])
model_name = max(dates)
mv_ind_max = np.argmax(mv_est['value'])
tf_ind_max = np.argmax(tf_est['value'])
TRAINED_MODEL[model_name] = {'mv_factor_ind': mv_est['factor_ind'][mv_ind_max],
'mv_back': mv_est['back'][mv_ind_max],
'mv_front': mv_est['front'][mv_ind_max],
'mv_est': mv_est,
'tf_factor_ind': tf_est['factor_ind'][tf_ind_max],
'tf_back': tf_est['back'][tf_ind_max],
'tf_front': tf_est['front'][tf_ind_max],
'tf_est': tf_est,
'loadings': pca.components_,
'dates': dates}
param = np.array([(model_name, mv_est['factor_ind'][mv_ind_max], mv_est['back'][mv_ind_max], mv_est['front'][mv_ind_max], mv_est['value'][mv_ind_max],
tf_est['factor_ind'][tf_ind_max], tf_est['back'][tf_ind_max], tf_est['front'][tf_ind_max], tf_est['value'][tf_ind_max])],
dtype = [('model_name', 'S8'), ('mv_factor_ind', int), ('mv_back', int), ('mv_front', int), ('mv_value', float),
('tf_factor_ind', int), ('tf_back', int), ('tf_front', int), ('tf_value', float)])
print "optimized param"
print param
return param
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
rsim = Mft2RollSim('test_pca', 100, 100)
rsim.TrainFunc = train_func
rsim.SimFunc = sim_func
#fileroot = '/home/mingyuan/Projects/DZH_data/Daily'
#filelist = [op.join(fileroot, 'daily_000029_20120623215017.csv')]
sector_filelist = [
'/home/mingyuan/Projects/DZH_data/Daily/daily_600016_20120928144733.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_002230_20120928145038.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_601857_20120928145224.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_000002_20120928145525.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_000799_20120928145811.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_000538_20120928150033.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_000060_20120928150309.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_000099_20120928150501.csv',
]
bank_filelist = [
'/home/mingyuan/Projects/DZH_data/Daily/daily_600016_20121001134109.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_600000_20121001134258.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_600036_20121001134436.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_601169_20121001134635.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_601998_20121001134751.csv',
'/home/mingyuan/Projects/DZH_data/Daily/daily_601009_20121001134859.csv',
]
filelist = sector_filelist
rsim.load_all_data(filelist, open_to_open_ret, min_date='20040101')
rsim.split_dates()
rsim.run()
rsim.print_pnl_summary()