Qlib TCN Workflow for CSI 300 Alpha158 Stock Selection
Summary
This configuration defines a Qlib workflow for predicting short-horizon CSI 300 stock returns with a temporal convolutional network (TCN). It uses Alpha158 features, filters a specified set of feature columns, applies robust cross-sectional normalization, and fills missing feature values. The label is based on the change from the next close to the close two trading days ahead. Training, validation, and test periods are separated, with the test segment covering 2017 through mid-2020.
The workflow records signal analysis and a portfolio backtest. Its portfolio strategy holds the top 50 ranked stocks and replaces a limited number of holdings at a time; the configuration also specifies a benchmark, trading costs, a price-limit assumption, and closing-price execution. These details make the experiment more reproducible, but the document contains no model scores, returns, or comparison with a baseline. Results would depend on the data, framework implementation, assumptions, and possible issues such as survivorship or point-in-time data handling.
Key ideas
- The workflow trains a TCN on Alpha158 features for CSI 300 stocks.
- The label measures a short forward close-to-close return, with distinct training, validation, and test segments.
- The backtest uses a top-ranked stock portfolio with gradual holding replacement and specified trading assumptions.
- The configuration describes an experiment but provides no evidence of predictive or portfolio performance.
Tags
Full text
# workflow_config_tcn_Alpha158.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: FilterCol
kwargs:
fields_group: feature
col_list: ["RESI5", "WVMA5", "RSQR5", "KLEN", "RSQR10", "CORR5", "CORD5", "CORR10",
"ROC60", "RESI10", "VSTD5", "RSQR60", "CORR60", "WVMA60", "STD5",
"RSQR20", "CORD60", "CORD10", "CORR20", "KLOW"
]
- 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: TCN
module_path: qlib.contrib.model.pytorch_tcn_ts
kwargs:
d_feat: 20
num_layers: 5
n_chans: 32
kernel_size: 7
dropout: 0.5
n_epochs: 200
lr: 1e-4
early_stop: 20
batch_size: 2000
metric: loss
loss: mse
optimizer: adam
n_jobs: 20
GPU: 0
dataset:
class: TSDatasetH
module_path: qlib.data.dataset
kwargs:
handler:
class: Alpha158
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]
step_len: 20
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.