Add high-frequency feature engineering code

qlib/contrib/data/highfreq_handler.py

@@ -0,0 +1,164 @@
+from qlib.data.dataset.handler import DataHandler, DataHandlerLP
+
+EPSILON = 1e-4
+
+
+class HighFreqHandler(DataHandlerLP):
+    def __init__(
+        self,
+        instruments="csi300",
+        start_time=None,
+        end_time=None,
+        infer_processors=[],
+        learn_processors=[],
+        fit_start_time=None,
+        fit_end_time=None,
+        drop_raw=True,
+    ):
+        def check_transform_proc(proc_l):
+            new_l = []
+            for p in proc_l:
+                p["kwargs"].update(
+                    {
+                        "fit_start_time": fit_start_time,
+                        "fit_end_time": fit_end_time,
+                    }
+                )
+                new_l.append(p)
+            return new_l
+
+        infer_processors = check_transform_proc(infer_processors)
+        learn_processors = check_transform_proc(learn_processors)
+
+        data_loader = {
+            "class": "QlibDataLoader",
+            "kwargs": {
+                "config": self.get_feature_config(),
+                "swap_level": False,
+                "freq": "1min",
+            },
+        }
+        super().__init__(
+            instruments=instruments,
+            start_time=start_time,
+            end_time=end_time,
+            data_loader=data_loader,
+            infer_processors=infer_processors,
+            learn_processors=learn_processors,
+            drop_raw=drop_raw,
+        )
+
+    def get_feature_config(self):
+        fields = []
+        names = []
+
+        template_if = "If(IsNull({1}), {0}, {1})"
+        template_paused = "Select(Gt($hx_paused_num, 1.001), {0})"
+
+        def get_normalized_price_feature(price_field, shift=0):
+            # norm with the close price of 237th minute of yesterday.
+            if shift == 0:
+                template_norm = "{0}/DayLast(Ref({1}, 243))"
+            else:
+                template_norm = "Ref({0}, " + str(shift) + ")/DayLast(Ref({1}, 243))"
+
+            template_fillnan = "FFillNan({0})"
+            # calculate -> ffill -> remove paused
+            feature_ops = template_paused.format(
+                template_fillnan.format(
+                    template_norm.format(template_if.format("$close", price_field), template_fillnan.format("$close"))
+                )
+            )
+            return feature_ops
+
+        fields += [get_normalized_price_feature("$open", 0)]
+        fields += [get_normalized_price_feature("$high", 0)]
+        fields += [get_normalized_price_feature("$low", 0)]
+        fields += [get_normalized_price_feature("$close", 0)]
+        fields += [get_normalized_price_feature("$vwap", 0)]
+        names += ["$open", "$high", "$low", "$close", "$vwap"]
+
+        fields += [get_normalized_price_feature("$open", 240)]
+        fields += [get_normalized_price_feature("$high", 240)]
+        fields += [get_normalized_price_feature("$low", 240)]
+        fields += [get_normalized_price_feature("$close", 240)]
+        fields += [get_normalized_price_feature("$vwap", 240)]
+        names += ["$open_1", "$high_1", "$low_1", "$close_1", "$vwap_1"]
+
+        # calculate and fill nan with 0
+        template_gzero = "If(Ge({0}, 0), {0}, 0)"
+        fields += [
+            template_gzero.format(
+                template_paused.format(
+                    "If(IsNull({0}), 0, {0})".format("{0}/Ref(DayLast(Mean({0}, 7200)), 240)".format("$volume"))
+                )
+            )
+        ]
+        names += ["$volume"]
+
+        fields += [
+            template_gzero.format(
+                template_paused.format(
+                    "If(IsNull({0}), 0, {0})".format(
+                        "Ref({0}, 240)/Ref(DayLast(Mean({0}, 7200)), 240)".format("$volume")
+                    )
+                )
+            )
+        ]
+        names += ["$volume_1"]
+
+        return fields, names
+
+
+class HighFreqBacktestHandler(DataHandler):
+    def __init__(
+        self,
+        instruments="csi300",
+        start_time=None,
+        end_time=None,
+    ):
+        data_loader = {
+            "class": "QlibDataLoader",
+            "kwargs": {
+                "config": self.get_feature_config(),
+                "swap_level": False,
+                "freq": "1min",
+            },
+        }
+        super().__init__(
+            instruments=instruments,
+            start_time=start_time,
+            end_time=end_time,
+            data_loader=data_loader,
+        )
+
+    def get_feature_config(self):
+        fields = []
+        names = []
+
+        template_if = "If(IsNull({1}), {0}, {1})"
+        template_paused = "Select(Gt($hx_paused_num, 1.001), {0})"
+        # template_paused = "{0}"
+        template_fillnan = "FFillNan({0})"
+        fields += [
+            template_fillnan.format(template_paused.format("$close")),
+        ]
+        names += ["$close0"]
+
+        fields += [
+            template_paused.format(
+                template_if.format(
+                    template_fillnan.format("$close"),
+                    "$vwap",
+                )
+            )
+        ]
+        names += ["$vwap0"]
+
+        fields += [template_paused.format("If(IsNull({0}), 0, {0})".format("$volume"))]
+        names += ["$volume0"]
+
+        fields += [template_paused.format("If(IsNull({0}), 0, {0})".format("$factor"))]
+        names += ["$factor0"]
+
+        return fields, names

qlib/contrib/data/highfreq_processor.py

@@ -0,0 +1,81 @@
+import os
+
+import numpy as np
+import pandas as pd
+from qlib.data.dataset.processor import Processor
+from qlib.data.dataset.utils import fetch_df_by_index
+from typing import Dict
+
+
+class HighFreqTrans(Processor):
+    def __init__(self, dtype: str = "bool"):
+        self.dtype = dtype
+
+    def fit(self, df_features):
+        pass
+
+    def __call__(self, df_features):
+        if self.dtype == "bool":
+            return df_features.astype(np.int8)
+        else:
+            return df_features.astype(np.float32)
+
+
+class HighFreqNorm(Processor):
+    def __init__(
+        self,
+        fit_start_time: pd.Timestamp,
+        fit_end_time: pd.Timestamp,
+        feature_save_dir: str,
+        norm_groups: Dict[str, int],
+    ):
+
+        self.fit_start_time = fit_start_time
+        self.fit_end_time = fit_end_time
+        self.feature_save_dir = feature_save_dir
+        self.norm_groups = norm_groups
+
+    def fit(self, df_features) -> None:
+        if os.path.exists(self.feature_save_dir) and len(os.listdir(self.feature_save_dir)) != 0:
+            return
+        os.makedirs(self.feature_save_dir)
+        fetch_df = fetch_df_by_index(df_features, slice(self.fit_start_time, self.fit_end_time), level="datetime")
+        del df_features
+        index = 0
+        names = {}
+        for name, dim in self.norm_groups.items():
+            names[name] = slice(index, index + dim)
+            index += dim
+        for name, name_val in names.items():
+            df_values = fetch_df.iloc(axis=1)[name_val].values
+            if name.endswith("volume"):
+                df_values = np.log1p(df_values)
+            self.feature_mean = np.nanmean(df_values)
+            np.save(self.feature_save_dir + name + "_mean.npy", self.feature_mean)
+            df_values = df_values - self.feature_mean
+            self.feature_std = np.nanstd(np.absolute(df_values))
+            np.save(self.feature_save_dir + name + "_std.npy", self.feature_std)
+            df_values = df_values / self.feature_std
+            np.save(self.feature_save_dir + name + "_vmax.npy", np.nanmax(df_values))
+            np.save(self.feature_save_dir + name + "_vmin.npy", np.nanmin(df_values))
+        return
+
+    def __call__(self, df_features):
+        if "date" in df_features:
+            df_features.droplevel("date", inplace=True)
+        df_values = df_features.values
+        index = 0
+        names = {}
+        for name, dim in self.norm_groups.items():
+            names[name] = slice(index, index + dim)
+            index += dim
+        for name, name_val in names.items():
+            feature_mean = np.load(self.feature_save_dir + name + "_mean.npy")
+            feature_std = np.load(self.feature_save_dir + name + "_std.npy")
+
+            if name.endswith("volume"):
+                df_values[:, name_val] = np.log1p(df_values[:, name_val])
+            df_values[:, name_val] -= feature_mean
+            df_values[:, name_val] /= feature_std
+        df_features = pd.DataFrame(data=df_values, index=df_features.index, columns=df_features.columns)
+        return df_features.fillna(0)