Silars
Silars 是基于 Polars 的因子分析与回测工具库，提供数据预处理、组合权重、策略和回测入口。
安装
Silars 需要 Python 3.12 或更高版本；CI 持续验证 Python 3.12 和 3.13。
pip install silars
使用 uv：
uv add silars
最小示例
下面按每个 (date, time) 截面选择得分最高的两个资产，并生成等权组合：
from copy import deepcopy
import polars as pl
from silars.alphalens import top_k
scores = pl.DataFrame(
{
"date": ["2026-01-02"] * 3,
"time": ["09:31:00"] * 3,
"asset": ["A", "B", "C"],
"score": [0.2, 0.8, 0.5],
}
)
selector = deepcopy(top_k).set_params(num=2)
weights = selector.transform(scores)
print(weights.select("date", "time", "asset", "target_weight"))
主要入口
Preprocessor及预处理函数：因子清洗、标准化和中性化。top_k、qcut、MFEs/MFEConfig：组合权重生成。Strategy、FactorStrategy：策略编排。BacktestEngine、bt：回测。Zoo、zoo：因子数据工作区。
这些入口均从 silars.alphalens 导入。
研究期 tree shortlist
silars.feature_selection.select_features 用固定的单特征浅树，按独立验证期的每日
prediction RankIC 对已物化数值特征做粗筛：
from silars.feature_selection import select_features
ranking = select_features(
train,
valid,
feature_names,
"forward_return",
max_features=100,
)
它衡量的是 tree(feature) 对下游模型的排序潜力，不是原始特征 RankIC、独立 OOS、
显著性结论或可直接传给 MFEs 的 score。PIT、purge/embargo 和后续使用边界由
调用方保证。该入口也不同于
silars.alphalens.evaluate.select_features 的多重检验与相关性筛选。
同一模块也可以把一组显式特征通过一棵相同参数的浅树融合为 predict 区间的新特征：
from silars.feature_selection import fuse_features
fused = fuse_features(
train,
valid,
["feature_a", "feature_b", "feature_c"],
target="forward_return",
output_name="tree_fusion",
)
valid = valid.with_columns(fused["tree_fusion"])
该入口只返回 predict prediction，不返回模型或 train in-sample prediction。若输入 特征是利用同一 predict 区间的 target 选出的，结果仍有 post-selection bias，不是 独立 OOS；完整历史融合列应由调用方 walk-forward/cross-fit 生成。
若 valid 含 target，可用同一棵 train-only 树得到一行 holdout 摘要及可复制的 Polars 表达式：
from silars.feature_selection import evaluate_fusion
report = evaluate_fusion(
train,
valid,
["feature_a", "feature_b", "feature_c"],
target="forward_return",
output_name="tree_fusion",
)
返回列为 feature、source_features、model_rank_ic_mean、
model_rank_icir、expression。这里只能证明 train/valid 时间分离；PIT、label
实现时点、purge/embargo、valid 前冻结特征集合且未重复使用 valid 选 winner 均由
调用方保证。holdout 审查只调用 evaluate_fusion，未来预测再选择
fuse_features 或冻结 expression，避免对同一区间重复 fit。
基准对冲回测
对已加载的 Zoo，可复用原多头回测并得到 1:1 对冲收益。传入指数代码时读取对应
指数日收益：
from silars.alphalens import FactorStrategy, zoo
results = zoo.hedge(
["KMID"],
FactorStrategy(),
index_code="000300",
times=["09:31:00", "10:00:00"],
)
print(results["KMID"]["ret"])
print(results["KMID"]["metric"])
hedge() 会立即完成回测、显示多头/基准/对冲组合净值图，并返回每个因子的
pos、ret 和对冲组合 metric，不是延迟任务生成器。
index_code="" 时不读取指数，直接从回测输入的 prev_close/open/close 构造等权市场
日收益，结果列为 mkt 和 long-mkt。指数路径运行时需要当前环境可导入
dc.data.base.ds_index_retC2C；period 仅控制原组合持仓周期，不缩放日度基准收益。
许可证
Silars 使用 MIT License。silars/empyrical 包含 Apache-2.0 许可的第三方代码，详见
THIRD_PARTY_NOTICES.md 和 LICENSES/Apache-2.0.txt。
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file silars-2026.7.30.1.tar.gz.
File metadata
- Download URL: silars-2026.7.30.1.tar.gz
- Upload date:
- Size: 181.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via:
uv/0.11.14 {"installer":{"name":"uv","version":"0.11.14","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
a0c243a692b2078e98e948418f6422f0c0bb25044d5d18e032cee724bfc08b83
|
|
| MD5 |
454a2a660879e628dd9460a4fbc278b8
|
|
| BLAKE2b-256 |
468326f167a07a252b50228f6ad1883e4dc1322c56b30a9470b64e0e420cdf8e
|
File details
Details for the file silars-2026.7.30.1-py3-none-any.whl.
File metadata
- Download URL: silars-2026.7.30.1-py3-none-any.whl
- Upload date:
- Size: 120.1 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via:
uv/0.11.14 {"installer":{"name":"uv","version":"0.11.14","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
5a0cd652c6f79a46438c487b45dcfefc064f830e3ba16c314def38b4f60c9708
|
|
| MD5 |
7ff88c2341e8af14c5b4bc353317e5a9
|
|
| BLAKE2b-256 |
01c82c53f70b3c3e25dc83c2fe120798c9091b44bdf80aea3623f1fd0c08aca9
|
