-
Notifications
You must be signed in to change notification settings - Fork 2
/
xgboostmodel.py
291 lines (268 loc) · 16.6 KB
/
xgboostmodel.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
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
# -*- coding: utf-8 -*-
"""
Created on Fri Apr 22 10:01:10 2022
Last Modified: Tue 21 Jun 2022 12:36:36 AM PST
@author: chenxiong888
"""
import xgboost as xgb
# from sklearn.model_selection import KFold, cross_val_score as CVS, train_test_split as TTS
# from sklearn.metrics import mean_squared_error as MSE
import pandas as pd
import datetime
# import time
import re
# import tushare as ts
import talib
import logging
from sometools import tools
import getData
LOGGER = logging.getLogger(__name__)
class BuildFeature(object):
def __init__(self, ):
self.tool=tools()
def splitbars(self,bars, y_days=5):
###
# para1: stock code
# para2: accumulate reward in days,which means feature y
# return k_lines of before y_days and y for y_days's reward
###
df1 = bars[1:-y_days-1]
df2 = bars['close'][-y_days-2:-1].tolist()
# y = round(((df2[-1] - df2[-y_days-1]) / df2[-1]) * 100, 2)
y = round(((df2[-1] - df2[-y_days-1]) / df2[-1]) * 100, 3)
return df1,y
def getfeatures(self,df,y):
open_values=df['open'].values
close_values=df['close'].values
low_values=df['low'].values
high_values=df['high'].values
pct_chg=df['pct_chg'].values
feature={}
feature.update({"股票收益":y})
feature.update({"股票代码":df.iloc[1]['ts_code']})
feature.update({"股票名称":df.iloc[1]['name']})
#==============================================
ma5=talib.SMA(close_values,timeperiod=5)
deg=self.tool.data_to_deg(ma5[-11:-1])
feature.update({"5日角度":deg})
ma20=talib.SMA(close_values,timeperiod=20)
deg=self.tool.data_to_deg(ma20[-11:-1])
feature.update({"20日角度":deg})
ma60=talib.SMA(close_values,timeperiod=60)
deg=self.tool.data_to_deg(ma60[-11:-1])
feature.update({"60日角度":deg})
if ma5[-2]<ma20[-2] and ma5[-1]>=ma20[-1]:
corss5x20=1
elif ma5[-2]>ma20[-2] and ma5[-1]<=ma20[-1]:
corss5x20=-1
else :
corss5x20=0
feature.update({"5日20日均线":corss5x20})
if ma20[-2]<ma60[-2] and ma20[-1]>=ma60[-1]:
corss20x60=1
elif ma20[-2]>ma60[-2] and ma20[-1]<=ma60[-1]:
corss20x60=-1
else :
corss20x60=0
feature.update({"20日60日均线":corss20x60})
pos=round((close_values[-1]-min(close_values))/(max(close_values)-min(close_values))*100,2)
feature.update({"价格分位数":pos})
#==============================================
feature.update({'两只乌鸦': talib.CDL2CROWS(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({'三只乌鸦': talib.CDL3BLACKCROWS(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '三内部上涨和下跌': talib.CDL3INSIDE(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '三线打击': talib.CDL3LINESTRIKE(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '三外部上涨和下跌': talib.CDL3OUTSIDE(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '南方三星': talib.CDL3STARSINSOUTH(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '三个白兵': talib.CDL3WHITESOLDIERS(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '弃婴': talib.CDLABANDONEDBABY(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '大敌当前': talib.CDLADVANCEBLOCK(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '捉腰带线': talib.CDLBELTHOLD(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '脱离': talib.CDLBREAKAWAY(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '收盘缺影线': talib.CDLCLOSINGMARUBOZU(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '藏婴吞没': talib.CDLCONCEALBABYSWALL(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '反击线': talib.CDLCOUNTERATTACK(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '乌云压顶': talib.CDLDARKCLOUDCOVER(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '十字': talib.CDLDOJI(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '十字星': talib.CDLDOJISTAR(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '蜻蜓十字/T形十字': talib.CDLDRAGONFLYDOJI(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '吞噬模式': talib.CDLENGULFING(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '十字暮星': talib.CDLEVENINGDOJISTAR(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '暮星': talib.CDLEVENINGSTAR(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '向上/下跳空并列阳线': talib.CDLGAPSIDESIDEWHITE(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '墓碑十字/倒T十字': talib.CDLGRAVESTONEDOJI(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '锤头': talib.CDLHAMMER(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '上吊线': talib.CDLHANGINGMAN(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '母子线': talib.CDLHARAMI(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '十字孕线': talib.CDLHARAMICROSS(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '风高浪大线': talib.CDLHIGHWAVE(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '陷阱': talib.CDLHIKKAKE(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '修正陷阱': talib.CDLHIKKAKEMOD(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '家鸽': talib.CDLHOMINGPIGEON(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '三胞胎乌鸦': talib.CDLIDENTICAL3CROWS(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '颈内线': talib.CDLINNECK(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '倒锤头': talib.CDLINVERTEDHAMMER(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '反冲形态': talib.CDLKICKING(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '由较长缺影线决定的反冲形态': talib.CDLKICKINGBYLENGTH(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '梯底': talib.CDLLADDERBOTTOM(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '长脚十字': talib.CDLLONGLEGGEDDOJI(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '长蜡烛': talib.CDLLONGLINE(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '光头光脚/缺影线': talib.CDLMARUBOZU(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '相同低价': talib.CDLMATCHINGLOW(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '铺垫': talib.CDLMATHOLD(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '十字晨星': talib.CDLMORNINGDOJISTAR(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '晨星': talib.CDLMORNINGSTAR(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '颈上线': talib.CDLONNECK(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '刺透形态': talib.CDLPIERCING(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '黄包车夫': talib.CDLRICKSHAWMAN(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '上升/下降三法': talib.CDLRISEFALL3METHODS(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '分离线': talib.CDLSEPARATINGLINES(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '射击之星': talib.CDLSHOOTINGSTAR(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '短蜡烛': talib.CDLSHORTLINE(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '纺锤': talib.CDLSPINNINGTOP(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '停顿形态': talib.CDLSTALLEDPATTERN(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '条形三明治': talib.CDLSTICKSANDWICH(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '探水竿': talib.CDLTAKURI(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '跳空并列阴阳线': talib.CDLTASUKIGAP(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '插入': talib.CDLTHRUSTING(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '三星': talib.CDLTRISTAR(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '奇特三河床': talib.CDLUNIQUE3RIVER(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '向上跳空的两只乌鸦': talib.CDLUPSIDEGAP2CROWS(open_values, high_values, low_values, close_values)[-1]/100})
feature.update({ '上升/下降跳空三法': talib.CDLXSIDEGAP3METHODS(open_values, high_values, low_values, close_values)[-1]/100})
# if realtime==False:
# self.features=self.features.append(feature,ignore_index=True)
# else:
# self.realtimefeatures=self.realtimefeatures.append(feature,ignore_index=True)
# print(self.features)
return feature
def run(self,days=[3,5], testing=False) -> list:#注意：days列表里必须至少有一项，且不为0
gd=getData.GetData()
feature0=pd.DataFrame()
allstock=gd.GetAllStock()
for k in days:#定义几个空的dataframe
exec(f"feature{k}=pd.DataFrame()")
# print("feature%s=pd.DataFrame()"%k)
j=0
for code, name in zip(allstock.ts_code, allstock.name):
try:
# time.sleep(0.15)
if (not self.check_stock(code)) or (self.check_stock(name, "^[*S]|退$")):
# 排除北交所，ST、退是整理
LOGGER.info(f"bypass {code} {name}")
continue
bars=gd.GetAStockData(code)# 得到K线
bars['name'] = name
# LOGGER.info(f"{code} {name} {bars.iloc[1]}")
if len(bars)>100:
featureret=self.getfeatures(bars,y=0) #get today's feature
# feature0=feature0.append(featureret,ignore_index=True)
# feature0.append(featureret,ignore_index=True)
# print(f"{type(featureret)=}")
# print(f"{featureret=}")
# print(f"{pd.DataFrame.from_dict(featureret, orient='index').T}")
# feature0=pd.concat(feature0, pd.DataFrame(featureret), ignore_index=True)
# feature01=pd.concat(feature01, pd.DataFrame(featureret), ignore_index=True)
feature0=pd.concat([feature0, pd.DataFrame.from_dict(featureret, orient='index').T])
feature0.reset_index(drop=True, inplace=True)
# feature0=feature0.convert_dtypes()
self._adjust_columns_type(feature0)
if (j % 5 == 0):
print(f"|{j} {code}", end=" ", flush=True)
else:
print(f"|{j} {code}", end=" ")
for k in days:
df,y=self.splitbars(bars,y_days=k)
df['name'] = name
featureret=self.getfeatures(df,y) #get learning feature,which includes rewards
exec(f"feature{k}=pd.concat([feature{k}, pd.DataFrame.from_dict(featureret, orient='index').T]);feature{k}.reset_index(drop=True, inplace=True)")
exec(f"self._adjust_columns_type(feature{k})")
# exec("feature%s=feature%s.append(featureret,ignore_index=True)"%(k,k))
j+=1
if testing and j>100 :
# 当testing为真时，测试数据为100
break; #这两行代码测试时使用
else:
print(f"{code}", end=" - ")
except OSError:
pass
result=[]
result.append(feature0)
for k in days:#定义几个空的dataframe
exec ("result.append(feature%s)"%k)
return result
@classmethod
def _adjust_columns_type(cls, a_df:pd.DataFrame) -> None:
"""调整pandas.DataFrame columns 数据类型
"""
#不需要改变类型的列名
exclude_columns=["股票代码",
"股票名称",
]
for col in a_df.columns:
if col in exclude_columns:
continue
a_df[col] = a_df[col].astype(float)
@classmethod
def check_stock(cls, code_or_name:str="", pattern="^[036].*"):
"""设定代码规则
指定某些股票代码是否剔除(例如默认pattern, 保留深证、创业板、沪市，剔除北交所代码)
return:
"""
# 返回股票代码以 0 3 6起始
res = re.search(pattern, code_or_name)
# res = re.search("^[036].*[^退]$", code)
# if res is None:
# LOGGER.info(f"{code_or_name} not match pattern {pattern}")
return res
if __name__ == "__main__":
bf=BuildFeature()
days=[1,3,5,10,20]
ret=bf.run(days=days) #得到特征矩阵，ret是一个list，每个元素是一个dataframe
predict_data=ret[0]
today = datetime.date.today()
today = today.strftime('%Y%m%d')
try:
writer = pd.ExcelWriter(today+'result.xlsx')
for i in range(len(days)):
data=ret[i+1]
y=data['股票收益']
# x=data.iloc[:,3:]
x=data.drop(['股票收益','股票代码','股票名称'],axis=1)
print(f"{x=}\n{y=}")
# Xtrain,Xtest,Ytrain,Ytest = TTS(x,y,test_size=0.3,random_state=420)
dfull = xgb.DMatrix(x,y)
param1 = {'silent':True
,'obj':'reg:linear'
,"subsample":1
,"max_depth":5
,"eta":0.1
,"gamma":2
,"lambda":0.2
,"alpha":0
,"colsample_bytree":1
,"colsample_bylevel":1
,"colsample_bynode":1
,"nfold":5}
num_round = 200
print("......正在学习模型......")
bst=xgb.train(param1,dfull,num_round)
print("......正在预测......")
xtest=predict_data.drop(['股票收益','股票代码','股票名称'],axis=1)
feature=xgb.DMatrix(xtest)
pred=bst.predict(feature)
predict_data['股票收益']=pred
today = datetime.date.today()
today = today.strftime('%Y%m%d')
print("预测完成！")
columns=['股票代码','股票名称','股票收益']
sorted=predict_data.sort_values(by="股票收益",ascending=False)[columns]
sorted['股票收益']=sorted['股票收益'].round(2)
print(f"{days[i]}日涨幅最大股票预测结果：")
print(sorted.head(20) )
top20=sorted.head(20)
top20.to_excel(writer,sheet_name="%s日涨幅"%(days[i]))
except Exception as e:
print(e.args)
finally:
writer.save()
writer.close()