当前位置：首页 > news >正文

认识一下Qlib的158因子特征

news 2025/7/26 10:07:19

from qlib.data import D
from qlib.data.dataset import DatasetH
from qlib.data.dataset.handler import DataHandlerLP
from qlib.data.dataset.loader import QlibDataLoader
from qlib.data.dataset.processor import DropnaLabel, Fillna, ZScoreNorm# ================== 第一部分：数据准备 ==================
instruments = ['sh000300']  # 沪深300指数
base_fields = ['$open', '$high', '$low', '$close', '$volume']  # 基础OHLCV数据# ================== 第二部分：特征工程 ==================
# 注意：QLib要求特征表达式必须用双引号包裹，且不能包含as关键字
technical_fields = [# 价格波动类特征"($close - $open) / $open",  # KWID"($high - $low) / $open",    # KLEN"($high - $close) / $close", # KWID2# 动量类特征"($close - Ref($close, 1)) / Ref($close, 1)",  # KUP"($high - Ref($close, 1)) / Ref($close, 1)",   # KUP2"($low - Ref($close, 1)) / Ref($close, 1)",    # KLOW"($low - Ref($low, 1)) / Ref($low, 1)",        # KLOW2# 形态类特征"($close - $open) / ($high - $low + 1e-5)",  # KSFT"Ref($close, 1) / ($open + 1e-5)",           # KSFT2# 开盘特征"$open / (Ref($close, 1) + 1e-5)",           # OPENG# 成交量特征"Mean($volume, 10) / ($volume + 1e-5)",      # VSUWN10"Mean($volume, 20) / ($volume + 1e-5)",      # VSUWN20"Mean($volume, 30) / ($volume + 1e-5)",      # VSUWN30"Mean($volume, 60) / ($volume + 1e-5)",      # VSUWN60"($volume - Ref($volume, 5)) / (Ref($volume, 5) + 1e-5)",  # VSUWD5# 其他特征"($volume - Ref($volume, 10)) / (Ref($volume, 10) + 1e-5)",  # VSUWD10"($volume - Ref($volume, 20)) / (Ref($volume, 20) + 1e-5)",  # VSUWD20"($volume - Ref($volume, 30)) / (Ref($volume, 30) + 1e-5)",  # VSUWD30"($volume - Ref($volume, 60)) / (Ref($volume, 60) + 1e-5)",  # VSUWD60# 标签"Ref($close, 5) / ($close + 1e-5) - 1"       # LABEL0
]# 对应的特征名称
feature_names = ["KWID", "KLEN", "KWID2","KUP", "KUP2", "KLOW", "KLOW2","KSFT", "KSFT2", "OPENG","VSUWN10", "VSUWN20", "VSUWN30", "VSUWN60", "VSUWD5","VSUWD10", "VSUWD20", "VSUWD30", "VSUWD60","LABEL0"
]# ================== 第三部分：数据加载配置 ==================
data_loader_config = {"feature": technical_fields[:-1],  # 所有特征字段（排除标签）"label": [technical_fields[-1]]    # 最后一个字段作为标签
}# ================== 第四部分：创建数据处理器 ==================
handler = DataHandlerLP(instruments=instruments,start_time="2010-01-01",end_time="2020-12-31",data_loader=QlibDataLoader(config=data_loader_config),infer_processors=[],learn_processors=[{"class": DropnaLabel, "kwargs": {}},{"class": Fillna, "kwargs": {}},{"class": ZScoreNorm, "kwargs": {"fit_start_time": "2010-01-01","fit_end_time": "2015-12-31"}}],
)# ================== 第五部分：数据集划分 ==================
segments = {"train": ("2010-01-01", "2015-12-31"),"valid": ("2016-01-01", "2017-12-31"),"test": ("2018-01-01", "2020-12-31"),
}# ================== 第六部分：创建数据集 ==================
dataset = DatasetH(handler=handler, segments=segments)# 准备训练数据
train_df = dataset.prepare("train")# 添加列名（因为原始表达式没有as别名）
train_df.columns = feature_namesprint("训练集数据样例：")
print(train_df.head())

训练集数据样例：KWID      KLEN     KWID2       KUP      KUP2  \
datetime   instrument                                                     
2010-01-04 sh000300   -0.015933  0.017403  0.017685 -0.011313  0.006172   
2010-01-05 sh000300    0.005317  0.022529  0.003785  0.008149  0.011965   
2010-01-06 sh000300   -0.004769  0.013393  0.013299 -0.006260  0.006956   
2010-01-07 sh000300   -0.020236  0.029858  0.025090 -0.019841  0.004752   
2010-01-08 sh000300    0.006717  0.016020  0.000560  0.002498  0.003059   KLOW     KLOW2      KSFT     KSFT2     OPENG  \
datetime   instrument                                                     
2010-01-04 sh000300   -0.011313 -0.002018 -0.915402  0.995324  1.004693   
2010-01-05 sh000300   -0.010627 -0.010627  0.235979  0.997188  1.002815   
2010-01-06 sh000300   -0.006417  0.012440 -0.355992  1.001498  0.998499   
2010-01-07 sh000300   -0.025118 -0.024963 -0.677694  0.999594  1.000401   
2010-01-08 sh000300   -0.012894 -0.007550  0.419210  1.004206  0.995806   VSUWN10   VSUWN20   VSUWN30   VSUWN60    VSUWD5  \
datetime   instrument                                                     
2010-01-04 sh000300    0.902123  0.976117  1.196088  1.193749  0.408967   
2010-01-05 sh000300    0.738603  0.748769  0.918195  0.924786  0.549599   
2010-01-06 sh000300    0.838915  0.815925  0.974689  1.017141  0.413020   
2010-01-07 sh000300    0.859500  0.804227  0.937711  0.993251 -0.091365   
2010-01-08 sh000300    1.130150  1.065647  1.202639  1.312900 -0.122060   VSUWD10   VSUWD20   VSUWD30   VSUWD60    LABEL0  
datetime   instrument                                                    
2010-01-04 sh000300   -0.000150 -0.507190 -0.302369  0.162298 -0.031245  
2010-01-05 sh000300    0.775501 -0.059368 -0.087044  0.455339 -0.024023  
2010-01-06 sh000300    0.454750 -0.053867 -0.468220  0.547375 -0.011576  
2010-01-07 sh000300    0.671713  0.172625 -0.298880 -0.007756  0.025174  
2010-01-08 sh000300   -0.055727  0.056051 -0.524639  0.003769  0.027453

初学qlib ，查看阿尔法158和阿尔法360多因子信息，

"""from qlib.contrib.data.handler import Alpha158
import qlib
#查看Qlib官方文档#检查qlib/contrib/data/handler.py源文件中的Alpha158类定义def main():qlib.init(provider_uri='D:/qlib/qlib_data/cn_data')handler = Alpha158(instruments='csi300', start_time='2010-01-01', end_time='2020-12-31')# 使用 get_cols() 获取因子名称feature_names = handler.get_cols()print(feature_names)print(len(feature_names))  # 应该输出158if __name__ == '__main__':main()"""
from qlib.contrib.data.handler import Alpha360
import qlibdef main():# 初始化qlibqlib.init(provider_uri='D:/qlib/qlib_data/cn_data')# 定义时间范围start_time = '2010-01-01'end_time = '2020-12-31'# 创建Alpha360数据处理器handler = Alpha360(instruments='csi300',start_time=start_time,end_time=end_time,# Alpha360不需要显式设置fit_start_time和fit_end_timeinfer_processors=[],learn_processors=[],)# 获取所有因子名称feature_names = handler.get_cols()# 打印因子信息print("Alpha360因子列表:")print(feature_names)print(f"因子总数: {len(feature_names)}")  # 应该输出360# 获取实际数据示例df = handler.fetch()print("\n数据示例:")print(df.head())if __name__ == '__main__':main()

查看全文

http://www.lryc.cn/news/599719.html