liuyu/qlib
1
0
Fork 0
mirror of https://github.com/microsoft/qlib.git synced 2026-06-06 05:51:17 +08:00
Code Activity

optimize_CI (#1314)

Browse Source
This commit is contained in:
Linlang
2022-10-20 08:38:05 +08:00
committed by GitHub
parent 3c62d131a5
commit 1861c8edaf
11 changed files with 70 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/rl/scripts/collect_pickle_dataframe.py
View File

@@ -13,7 +13,7 @@ for tag in ("backtest", "feature"):
df = pd.concat(list(df.values())).reset_index()
df["date"] = df["datetime"].dt.date.astype("datetime64")
instruments = sorted(set(df["instrument"]))
os.makedirs(os.path.join("data", "pickle_dataframe", tag), exist_ok=True)
for instrument in tqdm(instruments):
cur = df[df["instrument"] == instrument].sort_values(by=["datetime"])

20
examples/rl/scripts/gen_backtest_orders.py
View File

@@ -22,19 +22,21 @@ instruments = sorted(set(df["instrument"]))
df_list = []
for instrument in instruments:
print(instrument)
cur_df = df[df["instrument"] == instrument]
dates = sorted(set([str(d).split(" ")[0] for d in cur_df["date"]]))
n = args.num_order
df_list.append(
pd.DataFrame({
"date": sorted(np.random.choice(dates, size=n, replace=False)),
"instrument": [instrument] * n,
"amount": np.random.randint(low=3, high=11, size=n) * 100.0,
"order_type": np.random.randint(low=0, high=2, size=n),
}).set_index(["date", "instrument"]),
pd.DataFrame(
{
"date": sorted(np.random.choice(dates, size=n, replace=False)),
"instrument": [instrument] * n,
"amount": np.random.randint(low=3, high=11, size=n) * 100.0,
"order_type": np.random.randint(low=0, high=2, size=n),
}
).set_index(["date", "instrument"]),
)
total_df = pd.concat(df_list)

4
examples/rl/scripts/gen_pickle_data.py
View File

@@ -30,8 +30,8 @@ if __name__ == "__main__":
if "backtest_conf" in conf:
backtest = provider._gen_dataframe(deepcopy(provider.backtest_conf))
provider.feature_conf['path'] = os.path.splitext(provider.feature_conf['path'])[0] + '/'
provider.backtest_conf['path'] = os.path.splitext(provider.backtest_conf['path'])[0] + '/'
provider.feature_conf["path"] = os.path.splitext(provider.feature_conf["path"])[0] + "/"
provider.backtest_conf["path"] = os.path.splitext(provider.backtest_conf["path"])[0] + "/"
# Split by date
if args.split == "date" or args.split == "both":
provider._gen_day_dataset(deepcopy(provider.feature_conf), "feature")

16
examples/rl/scripts/gen_training_orders.py
View File

@@ -23,15 +23,17 @@ for group, n in zip(("train", "valid", "test"), (args.train_size, args.valid_siz
path = os.path.join("data", "pickle", f"backtest{group}.pkl")
df = pickle.load(open(path, "rb")).reset_index()
df["date"] = df["datetime"].dt.date.astype("datetime64")
dates = sorted(set([str(d).split(" ")[0] for d in df["date"]]))
data_df = pd.DataFrame({
"date": sorted(np.random.choice(dates, size=n, replace=False)),
"instrument": [args.stock] * n,
"amount": np.random.randint(low=3, high=11, size=n) * 100.0,
"order_type": [0] * n,
}).set_index(["date", "instrument"])
data_df = pd.DataFrame(
{
"date": sorted(np.random.choice(dates, size=n, replace=False)),
"instrument": [args.stock] * n,
"amount": np.random.randint(low=3, high=11, size=n) * 100.0,
"order_type": [0] * n,
}
).set_index(["date", "instrument"])
os.makedirs(os.path.join("data", "training_order_split", group), exist_ok=True)
pickle.dump(data_df, open(os.path.join("data", "training_order_split", group, f"{args.stock}.pkl"), "wb"))