Qlib TabNet Alpha360 Training and CSI 300 Backtest Configuration
Summary
This configuration specifies a Qlib workflow for a TabNet model using the Alpha360 feature handler and CSI 300 instruments, with the index as benchmark. It sets a two-day-ahead close-price return label, applies robust normalization and missing-value filling to features, and cross-sectionally rank-normalizes labels. The data are divided into pretraining, validation, training, and test periods; the test segment also defines the portfolio backtest window.
The portfolio uses a top-k dropout strategy that holds 50 names and replaces up to five, with trading costs, a minimum fee, a closing-price deal assumption, and a price-limit threshold specified. Signal, signal-analysis, and portfolio-analysis records are configured. This is an experiment recipe rather than a report: it includes no prediction metrics, portfolio returns, or comparison against alternatives. The setup alone therefore does not establish that TabNet or these choices produce useful out-of-sample performance.
Key ideas
- The workflow trains a TabNet model on Qlib’s Alpha360 features for CSI 300 stocks.
- The label is the forward two-day close-price return, and features and labels receive separate normalization steps.
- The configured portfolio holds 50 stocks and allows five positions to be dropped or replaced.
- The backtest specifies a benchmark, trading costs, minimum fees, a deal-price assumption, and a price-limit threshold.
- The YAML describes an experimental setup but reports no model or portfolio results.
Tags
Full text
# workflow_config_TabNet_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: TabnetModel
module_path: qlib.contrib.model.pytorch_tabnet
kwargs:
d_feat: 360
pretrain: True
seed: 993
dataset:
class: DatasetH
module_path: qlib.data.dataset
kwargs:
handler:
class: Alpha360
module_path: qlib.contrib.data.handler
kwargs: *data_handler_config
segments:
pretrain: [2008-01-01, 2014-12-31]
pretrain_validation: [2015-01-01, 2016-12-31]
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.