Configuring a GRU Model and Top-K Portfolio in Qlib
Summary
This configuration specifies a Qlib workflow for predicting CSI 300 stock returns with a gated recurrent unit (GRU) neural network. It uses Alpha158 features, filters to a selected feature set, applies robust normalization and missing-value filling, and rank-normalizes labels. The target is a forward close-to-close return over the next interval. A time-series dataset uses 20-step sequences, with training, validation, and test periods separated across 2008–2020.
Portfolio analysis applies a top-k dropout strategy that holds 50 names and replaces up to five positions, with backtesting from 2017 through mid-2020 against the CSI 300 benchmark. The configuration sets transaction costs, a minimum fee, a price limit threshold, and closing-price execution. It also specifies model training parameters and records for signal and portfolio analysis. These are experimental settings rather than reported findings: the document contains no performance results, and the configuration alone does not establish predictive value or account for all implementation risks.
Key ideas
- The workflow trains a GRU model on Alpha158 stock features using 20-step sequences.
- The prediction target is a forward close-to-close return.
- Data segments divide model fitting, validation, and testing across distinct historical periods.
- The portfolio strategy holds 50 stocks and allows five positions to be dropped and replaced.
- Backtesting includes a benchmark, transaction costs, minimum fees, and a price limit threshold, but no results are provided.
Tags
Full text
# workflow_config_gru.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: GeneralPTNN
module_path: qlib.contrib.model.pytorch_general_nn
kwargs:
n_epochs: 200
lr: 2e-4
early_stop: 10
batch_size: 800
metric: loss
loss: mse
n_jobs: 20
GPU: 0
pt_model_uri: "qlib.contrib.model.pytorch_gru_ts.GRUModel"
pt_model_kwargs: {
"d_feat": 20,
"hidden_size": 64,
"num_layers": 2,
"dropout": 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.