Qlib Alpha360 Stock Ranking with an ADD-GRU Model
Summary
This Qlib configuration trains an ADD model using a GRU base model and Alpha360 features to rank CSI 300 stocks. Features are robustly normalized and missing values are filled; labels are cross-sectionally rank-normalized after missing labels are dropped. The target is a short-horizon close-to-close return, while the training, validation, and test segments cover distinct historical periods.
The portfolio backtest applies a top-k dropout strategy, holding 50 names and replacing up to five, with the Shanghai 300 index as benchmark. It specifies close-price execution, transaction costs, a minimum fee, and a price-limit threshold. These settings make the file useful as a reproducible experiment template, but the document provides no resulting metrics, so it cannot establish predictive or investment performance. The configuration also relies on Chinese market data and stated sample dates; outcomes may depend on data quality, implementation, and assumptions such as close-price fills.
Key ideas
- The experiment uses Alpha360 features with an ADD model built on a GRU base.
- Robust feature normalization and cross-sectional label ranking are specified.
- The target is a close-price return over the next short interval.
- A top-k dropout portfolio holds 50 stocks and replaces up to five positions.
- The configuration includes transaction costs and a benchmark but reports no performance results.
Tags
Full text
# workflow_config_add_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: ADD
module_path: qlib.contrib.model.pytorch_add
kwargs:
d_feat: 6
hidden_size: 64
num_layers: 2
dropout: 0.1
dec_dropout: 0.0
n_epochs: 200
lr: 1e-3
early_stop: 20
batch_size: 5000
metric: ic
base_model: GRU
gamma: 0.1
gamma_clip: 0.2
optimizer: adam
mu: 0.2
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.