Qlib Transformer Workflow for CSI 300 Alpha360 Ranking
Summary
This configuration lays out a Qlib workflow for training a Transformer model on Alpha360 features and ranking CSI 300 stocks. It specifies Chinese market data, a close-to-close forward return label, robust feature normalization, missing-value filling, and cross-sectional label ranking. The data is divided into training, validation, and test periods, with model inference and portfolio analysis configured separately.
For portfolio evaluation, the setup uses a top-k dropout strategy that holds 50 names and replaces up to five, with the CSI 300 index as benchmark. The backtest settings include close-price dealing, transaction costs, a minimum fee, and a price-limit threshold. Signal analysis and portfolio analysis records are enabled. The document is configuration only: it provides no reported metrics, comparison against a baseline, or evidence that the model produces an edge. Results would depend on data quality, feature construction, execution assumptions, and choices such as the forward label and rebalance behavior.
Key ideas
- The workflow trains a Transformer model using Qlib's Alpha360 handler for CSI 300 stocks.
- Features are robustly normalized, missing values are filled, and labels are cross-sectionally ranked.
- A close-price forward return serves as the prediction target, with separate train, validation, and test periods.
- Portfolio analysis uses a top-k dropout approach, a benchmark, and configured trading costs and price limits.
- The configuration reports no model or backtest results, so it does not establish profitability.
Tags
Full text
# workflow_config_transformer_Alpha360.yaml
```yaml
qlib_init:
provider_uri: "~/.qlib/qlib_data/cn_data"
region: cn
market: &market csi300
benchmark: &benchmark SH000300
data_handler_config: &data_handler_config
start_time: 2008-01-01
end_time: 2020-08-01
fit_start_time: 2008-01-01
fit_end_time: 2014-12-31
instruments: *market
infer_processors:
- class: RobustZScoreNorm
kwargs:
fields_group: feature
clip_outlier: true
- class: Fillna
kwargs:
fields_group: feature
learn_processors:
- class: DropnaLabel
- class: CSRankNorm
kwargs:
fields_group: label
label: ["Ref($close, -2) / Ref($close, -1) - 1"]
port_analysis_config: &port_analysis_config
strategy:
class: TopkDropoutStrategy
module_path: qlib.contrib.strategy
kwargs:
signal: <PRED>
topk: 50
n_drop: 5
backtest:
start_time: 2017-01-01
end_time: 2020-08-01
account: 100000000
benchmark: *benchmark
exchange_kwargs:
limit_threshold: 0.095
deal_price: close
open_cost: 0.0005
close_cost: 0.0015
min_cost: 5
task:
model:
class: TransformerModel
module_path: qlib.contrib.model.pytorch_transformer
kwargs:
d_feat: 6
seed: 0
dataset:
class: DatasetH
module_path: qlib.data.dataset
kwargs:
handler:
class: Alpha360
module_path: qlib.contrib.data.handler
kwargs: *data_handler_config
segments:
train: [2008-01-01, 2014-12-31]
valid: [2015-01-01, 2016-12-31]
test: [2017-01-01, 2020-08-01]
record:
- class: SignalRecord
module_path: qlib.workflow.record_temp
kwargs:
model: <MODEL>
dataset: <DATASET>
- class: SigAnaRecord
module_path: qlib.workflow.record_temp
kwargs:
ana_long_short: False
ann_scaler: 252
- class: PortAnaRecord
module_path: qlib.workflow.record_temp
kwargs:
config: *port_analysis_config
```Shown in full with attribution under the source's licence. Licence: MIT
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.