-
Notifications
You must be signed in to change notification settings - Fork 3
/
prediction.py
208 lines (149 loc) · 8.31 KB
/
prediction.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
import numpy as np
import pandas as pd
import os
import stat_models
import distance_models
import utils
from tqdm import trange
from functools import reduce
from joblib import Parallel, delayed
def difference(df_stocks, tv_split=60):
df_diff = df_stocks.diff().iloc[1:] # First row is NA
df_diff["monthID"] = df_stocks["monthID"].iloc[1:]
train_ids = np.unique(df_diff.monthID)[:-tv_split]
test_ids = np.unique(df_diff.monthID)[-tv_split:-1]
return df_diff, train_ids, test_ids
def indexing(df_stocks, tv_split=60):
df_index = df_stocks.copy()
df_index.index = pd.to_datetime(df_index.index)
last_month_closes = df_index.loc[df_index.groupby(df_index.index.to_period('M')).apply(lambda x: x.index.max())]
ids = last_month_closes["monthID"]
last_month_closes = last_month_closes["Close"].shift(1)
for i in range(len(ids)):
mask = df_index["monthID"] == ids[i]
df_index.loc[mask, "Close"] = df_index.loc[mask, "Close"] / last_month_closes[i]
df_index = df_index.dropna()
train_ids = np.unique(df_index.monthID)[:-tv_split]
test_ids = np.unique(df_index.monthID)[-tv_split:-1]
return df_index, train_ids, test_ids
# https://stats.stackexchange.com/questions/178626/how-to-normalize-data-between-1-and-1
def normalizing(df_stocks, tv_split):
df_norm = df_stocks.copy()
ids = np.unique(df_stocks["monthID"])
for i in range(len(ids)):
mask = df_norm["monthID"] == ids[i]
x = df_norm.loc[mask, "Close"]
df_norm.loc[mask, "Close"] = 2 * ((x - np.min(x)) / (np.max(x)-np.min(x))) - 1
df_norm = df_norm.dropna()
train_ids = np.unique(df_norm.monthID)[:-tv_split]
test_ids = np.unique(df_norm.monthID)[-tv_split:-1]
return df_norm, train_ids, test_ids
def calculate_distances(train_x, train_y, test_x, train_labels, predict_id, settings_dict):
distances_f = np.zeros([len(train_labels), 6])
for count, unique_id in enumerate(train_labels):
train = train_x[train_x["monthID"] == unique_id]["Close"]
test = test_x[test_x["monthID"] == predict_id]["Close"]
distances_f[count, 0] = unique_id
if settings_dict["lcss_delta"] == "variable":
settings_dict["lcss_delta"] = np.max([1, np.round(0.25 * np.mean([len(train), len(test)]), 0)]) # At least 1, 25% of the length of the average length of train&test.
if settings_dict["lcss_epsilon"] == "variable":
settings_dict["lcss_epsilon"] = np.min([np.std(train), np.std(test)]) # Smallest standard deviation of train/test
if "dtw" in settings_dict["list_of_distance_models"]:
distances_f[count, 1] = distance_models.dtw(train, test, window=15)
else:
distances_f[count, 1] = np.nan
if "twed" in settings_dict["list_of_distance_models"]:
distances_f[count, 2] = distance_models.twed(train,
test,
nu=float(settings_dict["twed_nu"]),
_lambda=float(settings_dict["twed_lambda"]),)
else:
distances_f[count, 2] = np.nan
if "lcss" in settings_dict["list_of_distance_models"]:
distances_f[count, 3] = distance_models.lcss(train,
test,
delta=int(settings_dict["lcss_delta"]),
epsilon=float(settings_dict["lcss_epsilon"]))
else:
distances_f[count, 3] = np.nan
if "corrd" in settings_dict["list_of_distance_models"]:
distances_f[count, 4] = distance_models.corrd(train, test)
else:
distances_f[count, 4] = np.nan
distances_f[count, 5] = train_y[list(train_y.index) == unique_id]["Returns"]
output = pd.DataFrame(distances_f)
output.columns = ["monthID", "dtw", "twed", "lcss","corrd", "returns"]
return output
def multiprocessing_predict_trades(predict_id, train_labels, test_labels, stocks, returns, type_train, settings_dict):
train_x = stocks[stocks["monthID"].isin(train_labels)].copy()
train_y = returns[returns.index.isin(train_labels)].copy()
test_x = stocks[stocks["monthID"].isin(test_labels)].copy()
distances = calculate_distances(train_x, train_y, test_x, train_labels, predict_id, settings_dict)
output = pd.DataFrame(columns=list(["monthID", "instrument", "data_normalization", "distance_model", "stat_model", "result"]))
for a_distance_model in settings_dict["list_of_distance_models"]:
for a_stat_model in settings_dict["stat_models"]:
if a_stat_model == "knn":
output.loc[len(output)] = [predict_id, settings_dict["instrument"], type_train, a_distance_model, a_stat_model,
stat_models.knn(distances, a_distance_model)]
else:
output.loc[len(output)] = [predict_id, settings_dict["instrument"], type_train, a_distance_model, a_stat_model,
stat_models.kstar(distances, a_distance_model)]
return output
def predict_trades(type_train, stocks, returns, settings_dict, months_out_of_sample=60):
if type_train == "None":
train_labels = np.unique(stocks["monthID"])[:-months_out_of_sample]
test_labels = np.unique(stocks["monthID"])[-months_out_of_sample:-1]
elif type_train == "Difference":
stocks, train_labels, test_labels = difference(stocks, tv_split=months_out_of_sample)
elif type_train == "Index":
stocks, train_labels, test_labels = indexing(stocks, tv_split=months_out_of_sample)
elif type_train == "Normalization":
stocks, train_labels, test_labels = normalizing(stocks, tv_split=months_out_of_sample)
else:
raise TypeError("No/Wrong type_train chosen.")
labels = np.append(train_labels, test_labels)
output = pd.concat(Parallel(n_jobs=-1)(
delayed(multiprocessing_predict_trades)
(
test_labels[i],
labels[:(-months_out_of_sample + i + 1)],
labels[-(months_out_of_sample + 1 + i):],
stocks,
returns,
type_train,
settings_dict
) for i in trange(len(test_labels), desc="".join([settings_dict["instrument"], " ", type_train]))))
return output
# Perfomance metric = `accuracy` / `total_returns`
def take_best_parameters(config_f, performance_metric='profitability'):
try:
best_twed = pd.read_csv("data/param_test/{}_{}_test_acc_table.csv".format(config_f["instrument"], "twed")).sort_values(performance_metric, ascending=False)[['nu', 'lambda']].head(1).values[0] # nu / lambda
best_lcss = pd.read_csv("data/param_test/{}_{}_test_acc_table.csv".format(config_f["instrument"], "lcss")).sort_values(performance_metric, ascending=False)[['epsilon', 'delta']] .head(1).values[0] # Epsilon / Delta
config_f["twed_nu"] = best_twed[0]
config_f["twed_lambda"] = best_twed[1]
config_f["lcss_epsilon"] = best_lcss[0]
config_f["lcss_delta"] = best_lcss[1]
return config_f
except FileNotFoundError:
return config_f
def main(settings_dict):
stocks, returns = utils.read_data(settings_dict["instrument"])
trade_predictions = [predict_trades(
normalization_type,
stocks,
returns,
settings_dict,
months_out_of_sample= int(settings_dict["months_out_of_sample"])) for normalization_type in settings_dict["normalization_types"]]
if not os.path.exists('data/predictions'):
os.makedirs('data/predictions')
out_filename = "data/predictions/{}_predictions.csv".format(settings_dict["instrument"])
df_merged = reduce(lambda left, right: pd.merge(left, right, how='outer'), trade_predictions)
pd.DataFrame.to_csv(df_merged, out_filename, sep=',', na_rep='.', index=False)
instruments = ["^GSPC", "^DJI", "^GDAXI", "^N225", "^FCHI"] # ["Brent Oil", "Natural Gas", "Gasoline RBOB", "Carbon Emissions", "Gold", "Copper", "London Wheat"]
if __name__ == '__main__':
config = utils.read_config("run_parameters")
config["normalization_types"] = ["Index"]
for stock_index in instruments:
config["instrument"] = stock_index
config = take_best_parameters(config)
main(config)