mirror of
https://github.com/microsoft/qlib.git
synced 2026-07-12 15:26:54 +08:00
Refine DDG-DA (#1472)
* Run ddg-da successfully * Support include valid; More parameters * Support L2 reg & visualization * Blackformat * Enable fill_method * Support specify handler & optim dataset * Fix Pylint
This commit is contained in:
you-n-g
@@ -10,8 +10,10 @@ import pandas as pd
|
||||
import fire
|
||||
import sys
|
||||
import pickle
|
||||
from typing import Optional
|
||||
from qlib import auto_init
|
||||
from qlib.model.trainer import TrainerR
|
||||
from qlib.typehint import Literal
|
||||
from qlib.utils import init_instance_by_config
|
||||
from qlib.workflow import R
|
||||
from qlib.tests.data import GetData
|
||||
@@ -30,7 +32,33 @@ class DDGDA:
|
||||
- `rm -r mlruns`
|
||||
"""
|
||||
|
||||
def __init__(self, sim_task_model="linear", forecast_model="linear"):
|
||||
def __init__(
|
||||
self,
|
||||
sim_task_model: Literal["linear", "gbdt"] = "linear",
|
||||
forecast_model: Literal["linear", "gbdt"] = "linear",
|
||||
h_path: Optional[str] = None,
|
||||
test_end: Optional[str] = None,
|
||||
train_start: Optional[str] = None,
|
||||
meta_1st_train_end: Optional[str] = None,
|
||||
task_ext_conf: Optional[dict] = None,
|
||||
alpha: float = 0.0,
|
||||
proxy_hd: str = "handler_proxy.pkl",
|
||||
):
|
||||
"""
|
||||
|
||||
Parameters
|
||||
----------
|
||||
|
||||
train_start: Optional[str]
|
||||
the start datetime for data. It is used in training start time (for both tasks & meta learing)
|
||||
test_end: Optional[str]
|
||||
the end datetime for data. It is used in test end time
|
||||
meta_1st_train_end: Optional[str]
|
||||
the datetime of training end of the first meta_task
|
||||
alpha: float
|
||||
Setting the L2 regularization for ridge
|
||||
The `alpha` is only passed to MetaModelDS (it is not passed to sim_task_model currently..)
|
||||
"""
|
||||
self.step = 20
|
||||
# NOTE:
|
||||
# the horizon must match the meaning in the base task template
|
||||
@@ -38,10 +66,19 @@ class DDGDA:
|
||||
self.meta_exp_name = "DDG-DA"
|
||||
self.sim_task_model = sim_task_model # The model to capture the distribution of data.
|
||||
self.forecast_model = forecast_model # downstream forecasting models' type
|
||||
self.rb_kwargs = {
|
||||
"h_path": h_path,
|
||||
"test_end": test_end,
|
||||
"train_start": train_start,
|
||||
"task_ext_conf": task_ext_conf,
|
||||
}
|
||||
self.alpha = alpha
|
||||
self.meta_1st_train_end = meta_1st_train_end
|
||||
self.proxy_hd = proxy_hd
|
||||
|
||||
def get_feature_importance(self):
|
||||
# this must be lightGBM, because it needs to get the feature importance
|
||||
rb = RollingBenchmark(model_type="gbdt")
|
||||
rb = RollingBenchmark(model_type="gbdt", **self.rb_kwargs)
|
||||
task = rb.basic_task()
|
||||
|
||||
with R.start(experiment_name="feature_importance"):
|
||||
@@ -69,7 +106,7 @@ class DDGDA:
|
||||
fi = self.get_feature_importance()
|
||||
col_selected = fi.nlargest(topk)
|
||||
|
||||
rb = RollingBenchmark(model_type=self.sim_task_model)
|
||||
rb = RollingBenchmark(model_type=self.sim_task_model, **self.rb_kwargs)
|
||||
task = rb.basic_task()
|
||||
dataset = init_instance_by_config(task["dataset"])
|
||||
prep_ds = dataset.prepare(slice(None), col_set=["feature", "label"], data_key=DataHandlerLP.DK_L)
|
||||
@@ -96,7 +133,7 @@ class DDGDA:
|
||||
"kwargs": {"config": DIRNAME / "fea_label_df.pkl"},
|
||||
}
|
||||
)
|
||||
handler.to_pickle(DIRNAME / "handler_proxy.pkl", dump_all=True)
|
||||
handler.to_pickle(DIRNAME / self.proxy_hd, dump_all=True)
|
||||
|
||||
@property
|
||||
def _internal_data_path(self):
|
||||
@@ -108,7 +145,7 @@ class DDGDA:
|
||||
This function will dump the input data for meta model
|
||||
"""
|
||||
# According to the experiments, the choice of the model type is very important for achieving good results
|
||||
rb = RollingBenchmark(model_type=self.sim_task_model)
|
||||
rb = RollingBenchmark(model_type=self.sim_task_model, **self.rb_kwargs)
|
||||
sim_task = rb.basic_task()
|
||||
|
||||
if self.sim_task_model == "gbdt":
|
||||
@@ -122,24 +159,27 @@ class DDGDA:
|
||||
with self._internal_data_path.open("wb") as f:
|
||||
pickle.dump(internal_data, f)
|
||||
|
||||
def train_meta_model(self):
|
||||
def train_meta_model(self, fill_method="max"):
|
||||
"""
|
||||
training a meta model based on a simplified linear proxy model;
|
||||
"""
|
||||
|
||||
# 1) leverage the simplified proxy forecasting model to train meta model.
|
||||
# - Only the dataset part is important, in current version of meta model will integrate the
|
||||
rb = RollingBenchmark(model_type=self.sim_task_model)
|
||||
rb = RollingBenchmark(model_type=self.sim_task_model, **self.rb_kwargs)
|
||||
sim_task = rb.basic_task()
|
||||
train_start = self.rb_kwargs.get("train_start", "2008-01-01")
|
||||
train_end = "2010-12-31" if self.meta_1st_train_end is None else self.meta_1st_train_end
|
||||
test_start = (pd.Timestamp(train_end) + pd.Timedelta(days=1)).strftime("%Y-%m-%d")
|
||||
proxy_forecast_model_task = {
|
||||
# "model": "qlib.contrib.model.linear.LinearModel",
|
||||
"dataset": {
|
||||
"class": "qlib.data.dataset.DatasetH",
|
||||
"kwargs": {
|
||||
"handler": f"file://{(DIRNAME / 'handler_proxy.pkl').absolute()}",
|
||||
"handler": f"file://{(DIRNAME / self.proxy_hd).absolute()}",
|
||||
"segments": {
|
||||
"train": ("2008-01-01", "2010-12-31"),
|
||||
"test": ("2011-01-01", sim_task["dataset"]["kwargs"]["segments"]["test"][1]),
|
||||
"train": (train_start, train_end),
|
||||
"test": (test_start, sim_task["dataset"]["kwargs"]["segments"]["test"][1]),
|
||||
},
|
||||
},
|
||||
},
|
||||
@@ -156,7 +196,7 @@ class DDGDA:
|
||||
segments=0.62, # keep test period consistent with the dataset yaml
|
||||
trunc_days=1 + self.horizon,
|
||||
hist_step_n=30,
|
||||
fill_method="max",
|
||||
fill_method=fill_method,
|
||||
rolling_ext_days=0,
|
||||
)
|
||||
# NOTE:
|
||||
@@ -165,12 +205,15 @@ class DDGDA:
|
||||
# So the misalignment will not affect the effectiveness of the method.
|
||||
with self._internal_data_path.open("rb") as f:
|
||||
internal_data = pickle.load(f)
|
||||
|
||||
md = MetaDatasetDS(exp_name=internal_data, **kwargs)
|
||||
|
||||
# 3) train and logging meta model
|
||||
with R.start(experiment_name=self.meta_exp_name):
|
||||
R.log_params(**kwargs)
|
||||
mm = MetaModelDS(step=self.step, hist_step_n=kwargs["hist_step_n"], lr=0.001, max_epoch=100, seed=43)
|
||||
mm = MetaModelDS(
|
||||
step=self.step, hist_step_n=kwargs["hist_step_n"], lr=0.001, max_epoch=100, seed=43, alpha=self.alpha
|
||||
)
|
||||
mm.fit(md)
|
||||
R.save_objects(model=mm)
|
||||
|
||||
@@ -203,7 +246,7 @@ class DDGDA:
|
||||
hist_step_n = int(param["hist_step_n"])
|
||||
fill_method = param.get("fill_method", "max")
|
||||
|
||||
rb = RollingBenchmark(model_type=self.forecast_model)
|
||||
rb = RollingBenchmark(model_type=self.forecast_model, **self.rb_kwargs)
|
||||
task_l = rb.create_rolling_tasks()
|
||||
|
||||
# 2.2) create meta dataset for final dataset
|
||||
@@ -233,13 +276,13 @@ class DDGDA:
|
||||
"""
|
||||
with self._task_path.open("rb") as f:
|
||||
tasks = pickle.load(f)
|
||||
rb = RollingBenchmark(rolling_exp="rolling_ds", model_type=self.forecast_model)
|
||||
rb = RollingBenchmark(rolling_exp="rolling_ds", model_type=self.forecast_model, **self.rb_kwargs)
|
||||
rb.train_rolling_tasks(tasks)
|
||||
rb.ens_rolling()
|
||||
rb.update_rolling_rec()
|
||||
|
||||
def run_all(self):
|
||||
# 1) file: handler_proxy.pkl
|
||||
# 1) file: handler_proxy.pkl (self.proxy_hd)
|
||||
self.dump_data_for_proxy_model()
|
||||
# 2)
|
||||
# file: internal_data_s20.pkl
|
||||
|
||||
Reference in New Issue
Block a user