-
Notifications
You must be signed in to change notification settings - Fork 86
/
vadj_old.py
154 lines (116 loc) · 6.42 KB
/
vadj_old.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
#!/usr/bin/env python
from alphacalc import *
from dateutil import parser as dateparser
def calc_vadj(full_df, horizon):
print "Caculating daily vadj..."
result_df = full_df.reset_index()
result_df = filter_expandable(result_df)
result_df = result_df[ ['close', 'pbeta', 'tradable_volume', 'tradable_med_volume_21', 'log_ret', 'date', 'ind1', 'sid' ] ]
print "Calculating vadj0..."
rv = np.log(result_df['tradable_volume'] / result_df['tradable_med_volume_21'])
result_df['vadj0'] = rv * result_df['log_ret'] / result_df['pbeta']
result_df['vadj0_B'] = winsorize(result_df['vadj0'])
demean = lambda x: (x - x.mean())
indgroups = result_df[['vadj0_B', 'date', 'ind1']].groupby(['date', 'ind1'], sort=False).transform(demean)
result_df['vadj0_B_ma'] = indgroups['vadj0_B']
result_df.set_index(keys=['date', 'sid'], inplace=True)
print "Calulating lags..."
for lag in range(1,horizon+1):
shift_df = result_df.unstack().shift(lag).stack()
result_df['vadj' + str(lag) + '_B_ma'] = shift_df['vadj0_B_ma']
result_df = pd.merge(full_df, result_df, how='left', left_index=True, right_index=True, sort=False, suffixes=['', '_dead'])
result_df = remove_dup_cols(result_df)
return result_df
def calc_vadj_intra(full_df):
print "Calculating vadj intra..."
result_df = full_df.reset_index()
result_df = filter_expandable(result_df)
result_df = result_df[ ['iclose_ts', 'dopen', 'iclose', 'pbeta', 'dvolume', 'overnight_log_ret', 'tradable_med_volume_21_y', 'dvol_frac_med_21', 'date', 'ind1', 'sid' ] ]
result_df = result_df.dropna(how='any')
print "Calulating vadjC..."
rv = np.log((result_df['dvolume'] / result_df['dvol_frac_med_21']) / result_df['tradable_med_volume_21_y'])
result_df['vadjC'] = rv * (result_df['overnight_log_ret'] + (np.log(result_df['iclose']/result_df['dopen']))) / result_df['pbeta']
result_df['vadjC_B'] = winsorize(result_df['vadjC'])
print "Calulating vadjC_ma..."
demean = lambda x: (x - x.mean())
indgroups = result_df[['vadjC_B', 'date', 'ind1']].groupby(['date', 'ind1'], sort=False).transform(demean)
result_df['vadjC_B_ma'] = indgroups['vadjC_B']
#important for keeping NaTs out of the following merge
del result_df['date']
print "Merging..."
result_df.set_index(keys=['iclose_ts', 'sid'], inplace=True)
result_df = pd.merge(full_df, result_df, how='left', left_index=True, right_index=True, sort=True, suffixes=['_dead', ''])
result_df = remove_dup_cols(result_df)
return result_df
def vadj_fits(daily_df, intra_df, full_df, horizon, name):
regress_intra_df = intra_df
regress_daily_df = daily_df
# middate = intra_df.index[0][0] + (intra_df.index[len(intra_df)-1][0] - intra_df.index[0][0]) / 2
# print "Setting fit period before {}".format(middate)
# regress_intra_df = intra_df[ intra_df['date'] < middate ]
intra_horizon = 3
fits_df = pd.DataFrame(columns=['horizon', 'coef', 'indep', 'tstat', 'nobs', 'stderr'])
fitresults_df, intraForwardRets_df = regress_alpha_intra(regress_intra_df, 'vadjC_B_ma', intra_horizon)
fits_df = fits_df.append(fitresults_df, ignore_index=True)
plot_fit(fits_df, "vadj_intra_"+name+"_" + df_dates(regress_intra_df))
fits_df = pd.DataFrame(columns=['horizon', 'coef', 'indep', 'tstat', 'nobs', 'stderr'])
if 'vadj' not in full_df.columns:
print "Creating forecast columns..."
full_df[name] = np.nan
full_df[ 'vadjC_B_ma_coef' ] = np.nan
for lag in range(0, horizon+1):
full_df[ 'vadj' + str(lag) + '_B_ma_coef' ] = np.nan
for lag in range(1,horizon+1):
fitresults_df, dailyForwardRets_df = regress_alpha_daily(regress_daily_df, 'vadj0_B_ma', lag)
full_df = merge_intra_data(dailyForwardRets_df, full_df)
fits_df = fits_df.append(fitresults_df, ignore_index=True)
plot_fit(fits_df, "vadj_daily_"+name+"_" + df_dates(regress_daily_df))
fits_df.set_index(keys=['indep', 'horizon'], inplace=True)
coef0 = fits_df.ix['vadj0_B_ma'].ix[horizon].ix['coef']
full_df.ix[ intra_df.index, 'vadjC_B_ma_coef' ] = coef0
for lag in range(1,horizon+1):
coef = coef0 - fits_df.ix['vadj0_B_ma'].ix[lag].ix['coef']
print "Coef{}: {}".format(lag, coef)
full_df.ix[ intra_df.index, 'vadj'+str(lag)+'_B_ma_coef' ] = coef
full_df.ix[ intra_df.index, 'vadj'] = full_df['vadjC_B_ma'] * full_df['vadjC_B_ma_coef']
for lag in range(1,horizon):
full_df.ix[ intra_df.index, 'vadj'] += full_df['vadj'+str(lag)+'_B_ma'] * full_df['vadj'+str(lag)+'_B_ma_coef']
return full_df
def calc_vadj_forecast(daily_df, intra_df, horizon):
daily_df = calc_vadj(daily_df, horizon)
intra_df = calc_vadj_intra(intra_df)
full_df = merge_intra_data(daily_df, intra_df)
sector_name = 'Energy'
print "Running vadj for sector {}".format(sector_name)
sector_df = daily_df[ daily_df['sector_name'] == sector_name ]
sector_intra_df = intra_df[ intra_df['sector_name'] == sector_name ]
full_df = vadj_fits(sector_df, sector_intra_df, full_df, horizon, "vadj")
print "Running vadj for sector {}".format(sector_name)
sector_df = daily_df[ daily_df['sector_name'] != sector_name ]
sector_intra_df = intra_df[ intra_df['sector_name'] != sector_name ]
full_df = vadj_fits(sector_df, sector_intra_df, full_df, horizon, "vadj")
return full_df
if __name__=="__main__":
parser = argparse.ArgumentParser(description='G')
parser.add_argument("--start",action="store",dest="start",default=None)
parser.add_argument("--end",action="store",dest="end",default=None)
args = parser.parse_args()
start = args.start
end = args.end
lookback = 30
horizon = 5
start = dateparser.parse(start)
end = dateparser.parse(end)
uni_df = get_uni(start, end, lookback)
barra_df = load_barra(uni_df, start, end)
price_df = load_prices(uni_df, start, end)
daily_df = merge_barra_data(price_df, barra_df)
# volume_df = load_volume_profile(uni_df, start, end)
daybar_df = load_daybars(uni_df, start, end)
intra_df = merge_intra_data(daily_df, daybar_df)
intra_df = calc_vol_profiles(intra_df)
# intra_df = pd.merge(daybar_df, volume_df, left_index=True, right_index=True, sort=False, suffixes=['', '_dead'])
# intra_df = remove_dup_cols(intra_df)
full_df = calc_vadj_forecast(daily_df, intra_df, horizon)
dump_alpha(full_df, 'vadj')
dump_all(full_df)