-
v0.7.dev
- make server and the scripts work with different frequencies rather than only 1 minute
- restructure config for ML-features making their configuration closer to normal features and normal generators
- possibility to reference arbitrary external functions as generators
- support for parquet storage format for intermediate files
- all data including features, predicted scores and signals are stored in the context and available for further processing
- Improved visualization of historic data in on-line mode by using this common data context
- refactor and improve train signals based on the new structure
- introduce a section with signal generators
- refactoring: implement aggregations and trade logic to conventional column generators
-
v0.6.0 (2023-10-05)
- add visualization of previous transactions
- move algorithm configurations from source code to configuration file algorithms section
- add a feature generator based on TA-lib technical analysis library
- add configurations to feature generators instead of global parameters
- add configuration parameters to label generators
- refactor aggregation logic
-
v0.5.0 (2023-03-26)
- refactor and improve signal training (grid search) by taking into account the new rule model and external config file
- introduce rules with multiple inputs (currently only two dimensional rules with two thresholds)
- introduce a new rule model with different rule type
- introduce multiple aggregation sets
- generalize the aggregation model by separating it from the rule model, refactoring the logic of score combination and other improvements
- implement a new labeling function
-
v0.4.0 (2022-09-04)
- Unified specifications of data_sources, feature_sets and label_sets
- Introduce buy labels and sell labels
- Logging trade transactions and computing performance of the service
- Support for daily data frequency and Yahoo data source (only in batch mode)
-
v0.3.0 (2022-07-09)
-
v0.2.0 (2022-03-19)
-
v0.1.0 (2021-11-01)
-
v0.0.0 (2020-02-23)
- initial commit