Regularization-based asset hedging tool
A regularization-based (RB) hedging tool identifies a recommended hedging portfolio that track a target asset and provides one or metrics indicating a predicted performance of the hedging portfolio relative to the target asset. The RB hedging tool uses a RB hedging model that is trained on price data from an observation period. Initial hyperparameters for the model are selected using asset price data from a validation period and the performance of the model is evaluated by applying it to asset price data from a backtest period. The end-user is presented with one or more metrics indicating the performance of the model and may modify one or more settings (e.g., hyperparameters) of the model. The model is retrained and reapplied to the backtest period, and the metrics are updated. Thus, end-users may tailor the model to their own particular needs and preferences.
1. A method for training and using a regularization-based model to build a hedging portfolio, the method comprising:
identifying, using a processor, a target asset;
applying, using the processor, the regularization-based model to identify a recommended portfolio of assets that track the target asset, wherein applying the regularization-based model includes:
training the regularization-based model using asset price data from an observation period;
choosing initial hyperparameters of the regularization-based model using asset price data from a validation period, the validation period being after the observation period;
identifying an initial portfolio of assets by applying the regularization-based model to asset price data from a backtest period, the backtest period being after the validation period;
providing for display in a user interface information regarding the initial portfolio of assets;
receiving, via the user interface, an end-user modification to one or more hyperparameters of the regularization-based model to obtain updated hyperparameters;
retraining the regularization-based model using the updated hyperparameters as modified by the end-user;
identifying an updated portfolio of assets by applying the retrained regularization-based model to the asset price data from the backtest period, wherein the updated portfolio of assets is the recommended portfolio of assets; and
providing for display in the user interface updated information regarding the recommended portfolio of assets; and
obtaining, using the processor, the recommended portfolio of assets.
2. The method of claim 1 , wherein the hyperparameters include a concentration hyperparameter that impacts a total number of assets in the recommended portfolio of assets.
3. The method of claim 1 , wherein the hyperparameters include a diversity hyperparameter that impacts a range of relative sizes of asset positions in the recommended portfolio of assets.
4. The method of claim 1 , wherein the regularization-based model uses a linear objective function.
5. The method of claim 4 , wherein the linear objective function is solved according to a combination of linear and quadratic constraints.
6. The method of claim 1 , wherein the information regarding the portfolio of assets includes one or more metrics representing a performance of the initial portfolio and the user interface includes controls configured to enable the end-user to provide the modification to the one or more hyperparameters of the regularization-based model.
7. The method of claim 6 , wherein the updated information includes one or more updated metrics representing a performance of the updated portfolio.
8. The method of claim 6 , wherein the one or more metrics include at least one of: a tracking error, a holding error, a daily correlation, a hedge transaction cost, an annual volatility, or a chart indicating relative prices of the initial portfolio and the target asset.
9. The method of claim 6 , wherein the user interface includes one or more asset-selection controls configured to enable the end-user to select the target asset, and wherein identifying the target asset includes receiving an identifier of the target asset from the client device in response to end-user input using the one or more asset-selection controls.
10. The method of claim 1 , wherein retraining the regularization-based model is done in real time.
11. A non-transitory computer-readable medium storing instructions for training and using a regularization-based model to build a hedging portfolio that, when executed by a computing device, cause the computing device to perform operations comprising:
identifying a target asset;
applying the regularization-based model to identify a recommended portfolio of assets that track the target asset, wherein applying the regularization-based model includes:
training the regularization-based model using asset price data from an observation period;
choosing initial hyperparameters of the regularization-based model using asset price data from a validation period, the validation period being after the observation period;
identifying an initial portfolio of assets by applying the regularization-based model to asset price data from a backtest period, the backtest period being after the validation period;
providing for display in a user interface information regarding the initial portfolio of assets;
receiving, via the user interface, an end-user modification to one or more hyperparameters of the regularization-based model to obtain updated hyperparameters;
retraining the regularization-based model using the setting updated hyperparameters as modified by the end-user;
identifying an updated portfolio of assets by applying the retrained regularization-based model to the asset price data from the backtest period, wherein the updated portfolio of assets is the recommended portfolio of assets; and
providing for display in the user interface updated information regarding the recommended portfolio of assets; and
obtaining the recommended portfolio of assets.
12. The non-transitory computer-readable medium of claim 11 , wherein the hyperparameters include a concentration hyperparameter that impacts a total number of assets in the recommended portfolio of assets.
13. The non-transitory computer-readable medium of claim 11 , wherein the hyperparameters include a diversity hyperparameter that impacts a range of relative sizes of asset positions in the recommended portfolio of assets.
14. The non-transitory computer-readable medium of claim 11 , wherein the regularization-based model uses a linear objective function.
15. The non-transitory computer-readable medium of claim 14 , wherein the linear objective function is solved according to a combination of linear and quadratic constraints.
16. The non-transitory computer-readable medium of claim 11 , wherein the information regarding the portfolio of assets includes one or more metrics representing a performance of the initial portfolio and the user interface includes controls configured to enable the end-user to provide the modification to a setting of the regularization-based model.
17. The non-transitory computer-readable medium of claim 16 , wherein the updated information includes one or more updated metrics representing a performance of the updated portfolio.
18. The non-transitory computer-readable medium of claim 16 , wherein the one or more metrics include at least one of: a tracking error, a holding error, a daily correlation, a hedge transaction cost, an annual volatility, or a chart indicating relative prices of the initial portfolio and the target asset.
19. The non-transitory computer-readable medium of claim 11 , wherein the user interface includes one or more asset-selection controls configured to enable the end-user to select the target asset, and wherein identifying the target asset includes receiving an identifier of the target asset from the client device in response to end-user input using the one or more asset-selection controls.
20. The non-transitory computer-readable medium of claim 11 , wherein retraining the regularization-based model is done in real time.