Qlib TCN Workflow for CSI 300 Stock Ranking and Backtesting
Summary
This Qlib configuration defines a temporal convolutional network workflow for ranking CSI 300 stocks. It uses Alpha360 features, robust feature normalization, missing-value filling, and cross-sectional rank normalization of labels. The prediction target is based on the relative movement of closing prices over adjacent future days. Data is divided into training, validation, and test periods spanning 2008 through 2020, with model fitting limited to the training period. The workflow specifies a five-layer TCN trained with mean squared error and Adam, then evaluates signals and simulates a top-50 portfolio that can replace up to five holdings at a time. The backtest uses the CSI 300 benchmark, closing prices, transaction costs, and a price-limit threshold. These settings describe an experiment, not evidence of profitability: the document supplies no performance results, and its assumptions, data quality, costs, and potential look-ahead or survivorship issues would need independent review.
Key ideas
- The workflow trains a temporal convolutional network on Alpha360 features for CSI 300 stocks.
- Feature normalization and label ranking are applied through separate data processing steps.
- The target is defined from closing-price relationships across adjacent future days.
- Portfolio simulation ranks signals, holds up to 50 stocks, and allows five replacements at a time.
- The configuration specifies fees and price limits but provides no evidence of backtest performance.
Tags
Full text
# workflow_config_tcn_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: TCN
module_path: qlib.contrib.model.pytorch_tcn
kwargs:
d_feat: 6
num_layers: 5
n_chans: 128
kernel_size: 3
dropout: 0.5
n_epochs: 200
lr: 1e-3
early_stop: 20
batch_size: 2000
metric: loss
loss: mse
optimizer: adam
GPU: 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.