Quantization algorithm for asset allocation optimization in financial field
The quantitative algorithm addresses the limitations of existing asset allocation models by employing Dirichlet distribution-based random point generation and curve fitting to optimize asset allocation, achieving efficient and rational asset-liability alignment and faster convergence in complex financial environments.
Patent Information
- Application Number
- CN202411775696.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-05
- Publication Date
- 2025-03-25
AI Technical Summary
Current asset allocation optimization models in finance, such as Markowitz, Black-Litterman, and risk parity models, face limitations in handling multiple constraints and objectives, leading to unsolvable scenarios and inability to perform multi-objective optimization, especially in complex financial environments with increased regulatory demands and diverse management goals, resulting in incomplete and irrational asset allocation schemes.
A quantitative algorithm for asset allocation optimization that includes steps like user input of optimization targets and constraints, static and dynamic optimization, random point generation using Dirichlet distribution, multi-dimensional indicator calculation, and curve fitting to achieve efficient asset and liability alignment, enabling visualization and interactive operations.
The algorithm efficiently performs multi-objective, multi-constraint optimization across both account levels, achieving faster convergence to optimal results and enhancing asset-liability matching, providing a comprehensive and rational asset allocation framework.
Smart Images
Figure CN119693151A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of asset allocation optimization in the financial field, and in particular to a quantitative algorithm for optimizing asset allocation in the financial field. Background Art
[0002] In the financial field, asset allocation optimization is one of the important links in formulating asset allocation plans, especially in the insurance and pension fields. Asset allocation optimization needs to be carried out separately at the sub-account and general account levels, and consider the asset-liability linkage process, as well as numerous optimization objectives and constraints, including but not limited to investment returns, solvency, volatility, duration and cash flow, etc., which involves a large amount of data transmission, quantitative measurement and visual display.
[0003] Traditional asset allocation optimization models are commonly used in the current market, including the mean-variance model (Markowitz model), the Black-Litterman model (BL model) and the risk parity model. However, these models are all based entirely on mathematical programming solutions. When there are many constraints, there are technical bottlenecks and process limitations such as no solution, inability to perform multi-objective optimization, and inability of pure asset-side models to perform asset-liability linkage. Especially in the context of the increasing strengthening of systemic financial risk supervision, the increasing prominence of pension planning needs involving people's livelihood, the increasingly diverse management objectives of financial institutions such as insurance institutions, and the challenges of asset allocation in the implementation of new international accounting standards, the traditional asset allocation optimization algorithm model can no longer meet the daily needs of asset allocation and asset-liability management, and is accompanied by a large number of manual corrections, thus limiting the integrity, regularity and rationality of the asset allocation plan generated by financial institutions (especially traditional insurance institutions and pension insurance institutions). Therefore, the industry urgently needs to find a more reasonable and efficient asset allocation optimization algorithm that can meet diversified business goals and consider asset-liability linkage. Summary of the invention
[0004] The purpose of the present invention is to provide a quantitative algorithm for optimizing asset allocation in the financial field, which is used to solve the problems raised in the background technology and facilitate promotion.
[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:
[0006] A quantitative algorithm for optimizing asset allocation in the financial field, which mainly includes the following steps:
[0007] S1. Input the optimization objectives and constraints of the asset allocation optimization process at the sub-account level in the system visualization interface;
[0008] S101. The optimization goal selected by the user and the constraints that can be input include but are not limited to the type of asset portfolio, type of liability, asset allocation ratio, investment return rate, volatility, duration, value at risk and solvency;
[0009] S102. In the static optimizer, the user inputs data at a certain optimization point in time, and the model automatically obtains it. In the dynamic optimizer, the user inputs asset and liability data at the evaluation point in time, as well as future rebalancing rules and assumptions about new assets. The model automatically performs dynamic rebalancing, calculates and obtains data at the optimization point in time.
[0010] S2. At the sub-account level, determine the starting point of the random scatter of asset allocation through planning and solving methods;
[0011] S201. Solve the asset allocation planning in the dimension of expected investment return rate-volatility, and find the highest expected investment return rate within the unit volatility step;
[0012] S202, preliminarily generating an effective frontier curve including multiple asset allocation points as a set of fixed points of preliminary asset allocation random scattered points;
[0013] S3. At the sub-account level, randomly generate asset allocation points by means of random scattering;
[0014] S301. Calibrate the parameters of Dirichlet Distribution according to the input parameters of the model.
[0015] S302. Perform statistical probability distribution fitting based on Dirichlet Distribution, randomly simulate multiple asset allocation points, and the sum of the proportions of each asset allocation point is 1, thereby forming a first round of asset allocation point set.
[0016] S4. At the sub-account level, multi-dimensional quantitative indicator calculations are performed for each asset allocation point;
[0017] S401. Conduct quantitative modeling for various types of assets and liabilities and calculate various quantitative indicators, including but not limited to cash flow, present value, rate of return, volatility, value at risk, solvency minimum capital and stress testing;
[0018] S402. Considering the asset and liability modeling information at the same time, taking into account the asset-liability linkage benefits, using the curve interpolation algorithm, calculate the quantitative indicators under the asset-liability linkage, including but not limited to net assets, duration gap, solvency adequacy ratio, net volatility, net value at risk and stress test net value;
[0019] S403, based on the constraints input by the user and the quantitative indicators of each asset allocation point calculated in S401 and S402, the first round of asset allocation point set is screened to generate a second round of asset allocation point set;
[0020] S5. At the sub-account level, the efficient frontier curve is generated through random simulation interpolation and random simulation smoothing calculation of asset allocation points;
[0021] S501. In the expected investment return rate-volatility dimension, a plurality of effective frontier allocation points are randomly selected, and multiple random simulation interpolations are performed to form new asset allocation points, thereby forming a third round of asset allocation point set;
[0022] S502, in the expected investment return rate-volatility dimension, based on the third round of asset allocation point set, select the second round of effective frontier allocation points, and perform smoothing calculation of the effective frontier allocation points to form the final effective frontier allocation points, form an effective frontier curve by connecting the effective frontier allocation points, and generate the fourth round of asset allocation point set;
[0023] S503, the system stores the fourth round of asset allocation point set into the database for subsequent visualization and interactive operation presentation;
[0024] S6. At the sub-account level, all asset allocation points are visualized and interactively presented;
[0025] S601. Based on the horizontal axis (X axis) and vertical axis (Y axis) selected by the user, the fourth round of asset allocation point set is automatically displayed in two-dimensional quadrants, and the displayed content includes the calculated quantitative indicators, and the displayed forms include pictures and tables;
[0026] S602: The user selects upper and lower limits for each quantitative indicator, and the model automatically performs filtering and displays it visually on the system interface;
[0027] S7. At the sub-account level, the asset allocation optimization results are downloaded and exported. Users can distinguish between effective frontier allocation points and ineffective frontier allocation points for the fourth round of asset allocation points, and export the results in the form of pictures and Excel.
[0028] S8. Optimize asset allocation at the total account level, visualize and export results based on sub-account optimization results;
[0029] S801. Generate the possibility of all total account asset allocation points based on the asset allocation points optimized for each sub-account by using an exhaustive method;
[0030] S802, based on the total account configuration point set, calculate each quantitative indicator (the algorithm is the same as S4);
[0031] S803. Visually present all asset allocation points and interactively present them at the level of the total account (the algorithm is the same as S6);
[0032] S804. Export the asset allocation optimization results at the general account level (the algorithm is the same as S7);
[0033] Furthermore, through the random simulation probability distribution fitting algorithm, the possibility of asset allocation points is generated, and the convergence effect is achieved faster, which is closer to the global optimal result.
[0034] Furthermore, with extremely large amounts of data, the calculation, comparison, screening, visualization and operational interaction of all asset and liability quantitative indicators of all asset allocation points can be efficiently completed.
[0035] Furthermore, by using the curve interpolation algorithm, the asset-liability linkage between various asset types and liability cash flows based on multiple investment yield assumptions can be efficiently completed, greatly improving the degree of asset-liability matching in the asset allocation optimization process.
[0036] Furthermore, through the effective frontier random simulation interpolation and random simulation smoothing algorithm, an effective frontier curve with smooth changes in asset allocation plans between adjacent effective frontier asset allocation points is generated.
[0037] Furthermore, a set of algorithms combining sub-account optimization with total account optimization is provided, which enables the possibility of generating total account asset allocation optimization results after sub-account asset allocation optimization based on the exhaustive method.
[0038] As an improvement, the beneficial effects of the present invention are:
[0039] The present invention relates to a quantitative algorithm for optimizing asset allocation in the financial field, which efficiently completes a multi-objective, multi-constraint, and multi-dimensional asset allocation optimization algorithm model at the sub-account level and the total account level, taking into account the linkage between assets and liabilities, and reaches a convergence result based on a random simulation asset allocation optimization process in a shorter time. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] Figure 1 A schematic diagram of the calculation steps and user operation flow of a quantitative algorithm for optimizing asset allocation in the financial field of the present invention;
[0041] FIG2 is a schematic diagram of a page design for a user to input underlying asset information and liability information of a quantitative algorithm for optimizing asset allocation in the financial field according to the present invention;
[0042] Figure 3 A schematic diagram of a page design for a user to input various types of constraints for a quantitative algorithm for optimizing asset allocation in the financial field of the present invention;
[0043] Figure 4A schematic diagram of the page design of an asset allocation optimization result of a quantitative algorithm for asset allocation optimization in the financial field of the present invention;
[0044] Figure 5 A schematic diagram of a page design for a user to input filtering conditions for a quantitative algorithm for optimizing asset allocation in the financial field of the present invention;
[0045] Figure 6 A schematic diagram of a page design for downloading and exporting the results of a quantitative algorithm for optimizing asset allocation in the financial field according to the present invention;
[0046] Figure 7 A schematic diagram of a page design for optimizing total account asset allocation by selecting a quantitative algorithm for optimizing asset allocation in the financial field according to the present invention; DETAILED DESCRIPTION
[0047] In order to make the contents of the present invention more clearly understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.
[0048] like Figure 1 As shown in FIG. 1 , a quantitative algorithm for optimizing asset allocation in the financial field is implemented, and the specific implementation includes the following steps:
[0049] S1. In the system visualization interface, input the optimization objectives and constraints of the asset allocation optimization process at the sub-account level.
[0050] As shown in Figure 2, users enter the underlying asset information (including existing assets and new assets) and insurance account liability information (including existing liability cash flow and new liability cash flow) in the system interface. In the static optimizer, users enter the data at a certain optimization point in time. In the dynamic optimizer, users enter the asset and liability data at the evaluation point in time, as well as future rebalancing rules and new asset assumptions.
[0051] like Figure 3 As shown, the user can select the optimization objectives and input the constraints in the system interface.
[0052] The quantitative indicators that can be selected include but are not limited to asset portfolio type, liability type, asset allocation ratio, investment return rate, volatility, duration, value at risk and solvency, etc.;
[0053] The types of constraints that users can enter include asset constraints, that is, users can customize the upper and lower limits of asset allocation ratios in asset classification; customized label constraints, that is, users can customize the packaging of multiple types of assets and assign labels, and set upper and lower limits of asset allocation ratios and other classification indicators on the labels; equivalent asset label constraints, that is, users can customize the packaging of multiple types of assets and assign labels, and the optimization algorithm will assign the same weight to each asset type under the same label; indicator constraints, that is, users can choose a number of quantitative indicators, including but not limited to investment return rate, solvency, volatility, duration and cash flow, etc., and set the upper and lower limits of each quantitative indicator.
[0054] The model presets a variety of quantitative indicators, and users can set custom quantitative indicators. The model will automatically perform weighted average calculations and participate in the asset allocation optimization process.
[0055] The system automatically reads and saves the user input information (including constraints, etc.) for use in the subsequent step S403.
[0056] In the static optimizer, the user inputs data at a certain optimization point in time, and the model automatically obtains the data, such as the size of the configured funds, the major categories of configurable assets, and the risk and return indicators of each major category of assets.
[0057] In the dynamic optimizer, users input asset and liability data at the evaluation time, as well as future rebalancing rules and assumptions about new assets. The model will automatically perform dynamic rebalancing, calculate and obtain data at the optimization time, including the size of the allocated funds, the major categories of configurable assets, and the risk and return indicators of each major category of assets.
[0058] S2. At the sub-account level, determine the starting point of the random scatter of asset allocation through planning and solving methods;
[0059] S201. Perform asset allocation planning and solve the expected investment return rate-volatility dimension, and find the highest expected investment return rate within a unit 0.1% volatility step.
[0060] S202. Preliminarily generate an effective frontier curve containing 58 asset allocation points as a set of fixed points for preliminary asset allocation random scatter points.
[0061] At the same time, the stock asset information is read, and the current asset allocation point is added to the fixed point set.
[0062] S3. At the sub-account level, randomly generate asset allocation points by means of random scattering;
[0063] S301. Calibrate the parameters of Dirichlet Distribution according to the input parameters of the model.
[0064] S302. Perform statistical probability distribution fitting based on Dirichlet Distribution, randomly simulate 30,000 asset allocation points near each fixed point (i.e., a total of 30,000*59=1,770,000 random asset allocation points are randomly simulated), and the sum of the proportions of each asset allocation point is 1. This forms the first round of asset allocation point set.
[0065] S4. At the sub-account level, multi-dimensional quantitative indicator calculations are performed for each asset allocation point;
[0066] S401. Carry out quantitative modeling for various types of assets and liabilities, and calculate the quantitative indicators of the asset side and the liability side respectively, including but not limited to cash flow, present value, rate of return, volatility, value at risk, minimum solvency capital and stress test, etc., a total of 34 quantitative indicators.
[0067] S402. Considering the asset and liability modeling information at the same time, considering the asset-liability linkage benefits, use the curve interpolation algorithm. Specifically, based on the expected investment yield results obtained by weighting different asset allocation points in parallel, and the liability cash flow under the different expected investment yield actuarial assumptions input by the user, linearly interpolate the liability cash flow that matches the expected investment yield results of the asset allocation points for subsequent quantitative indicator measurement.
[0068] Quantitative indicators under the linkage between assets and liabilities are calculated, including but not limited to net assets, duration gap, solvency ratio, net volatility, net value at risk and stress test net value, totaling 26 quantitative indicators.
[0069] In the static optimizer, the index is calculated at a certain analysis time point selected by the user. In the dynamic optimizer, the rebalancing index is calculated at a certain analysis time point and an optimization time point selected by the user.
[0070] S403. According to the constraints input by the user and the quantitative indicators of each asset allocation point calculated in S4, the first round of asset allocation point set is screened to generate a second round of asset allocation point set.
[0071] S5. At the sub-account level, the efficient frontier curve is generated through random simulation interpolation and random simulation smoothing calculation of asset allocation points;
[0072] S501. In the expected investment return-volatility dimension, within a volatility step of 0.1%, select the point with the highest expected investment return.
[0073] Based on the selected effective frontier configuration points, two random simulated linear interpolations are performed (the effective frontier configuration points are reselected after the first interpolation), that is, two effective frontier configuration points are randomly selected and interpolation of 25%, 50%, and 75% intermediate points is performed.
[0074] This will form new asset allocation points and form the third round of asset allocation point set.
[0075] S502. In the expected investment yield-volatility dimension, based on the third round of asset allocation point set, the second round of effective frontier allocation point selection is performed, and the effective frontier allocation point smoothing calculation is performed. Specifically, in the expected investment yield indicator dimension, an effective frontier asset allocation point set with a maximum step size of 0.02% is generated. The 4 closest allocation points are selected around each effective frontier allocation point, and the weighted average of the proportions of the 4 asset allocation points is performed, and each quantitative indicator is re-weighted and averaged.
[0076] After smoothing calculation, the final effective frontier configuration point is formed. By connecting the effective frontier configuration points, an effective frontier curve is formed. And the fourth round of asset configuration point set is generated.
[0077] S503. The system stores the fourth round of asset allocation point set in a database for subsequent visualization and interactive operation presentation.
[0078] S6. At the sub-account level, all asset allocation points are visualized and interactively presented.
[0079] S601, such as Figure 4 As shown, based on the horizontal axis (X-axis) and vertical axis (Y-axis) selected by the user, the fourth round of asset allocation point set is automatically displayed in two-dimensional quadrants. The display content includes various types of calculated quantitative indicators, and the display forms include pictures and tables.
[0080] After the user modifies the horizontal axis (X axis) and vertical axis (Y axis), the system interface will automatically adjust and re-display.
[0081] S602. The user can filter the upper and lower limits of each quantitative indicator, and the model will automatically filter.
[0082] like Figure 5 As shown, according to the filtering conditions input by the user, the system will automatically filter out the asset allocation points that do not meet the filtering conditions in the fourth round of asset allocation point set.
[0083] Based on the filtered asset allocation points, the system interface is visualized and displayed in the same way as S601.
[0084] S7, such as Figure 6 As shown in the figure, at the sub-account level, the asset allocation optimization results can be downloaded and exported. Users can distinguish between effective frontier allocation points and ineffective frontier allocation points for the fourth round of asset allocation points, and export the results in the form of pictures and Excel.
[0085] S8. Optimize asset allocation, visualize and export results at the total account level based on sub-account optimization results.
[0086] S801. Use the exhaustive method to generate the possibility of all total account asset allocation points based on the optimized asset allocation points of the sub-accounts.
[0087] like Figure 7 As shown, specifically, according to the sub-account information selected by the user, an asset allocation point is selected from the fourth round asset allocation point set in each sub-account by using a permutation and combination method, and the allocation ratio of the total account asset allocation point is formed by a weighted average method.
[0088] Repeat the above method to exhaustively enumerate all the possibilities of total account asset allocation points and generate a set of total account asset allocation points.
[0089] S802. Based on the total account configuration point set, various quantitative indicators are calculated (the algorithm is the same as S4).
[0090] S803. Visualize and interactively present all asset allocation points at the total account level (the algorithm is the same as S6).
[0091] S804. At the general account level, export the asset allocation optimization results (the algorithm is the same as S7).
[0092] The present invention is installed, deployed, debugged and used through a systemized product. The above algorithms, models and operation processes are implemented in a systematic and automated manner.
[0093] After the present invention is systematically implemented, it can flexibly respond to various user usage scenarios and business scenarios without changing the algorithm logic, including but not limited to adding new quantitative indicators, adding new insurance accounts and types, and switching between Chinese and English languages.
[0094] The present invention provides a quantitative algorithm for optimizing asset allocation in the financial field, which can efficiently and reasonably complete the asset allocation optimization process at the sub-account level and the total account level, and quickly calculate the quantitative indicators required for various types of asset allocation and asset-liability management, and then perform user-friendly result visualization.
[0095] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.
Claims
1. A quantitative algorithm for optimizing asset allocation in the financial field, which mainly includes the following steps: S1. Input the optimization objectives and constraints of the asset allocation optimization process at the sub-account level in the system visualization interface; S101. The optimization goal selected by the user and the constraints that can be input include but are not limited to the type of asset portfolio, type of liability, asset allocation ratio, investment return rate, volatility, duration, value at risk and solvency; S102. In the static optimizer, the user inputs data at a certain optimization point in time, and the model automatically obtains it. In the dynamic optimizer, the user inputs asset and liability data at the evaluation point in time, as well as future rebalancing rules and assumptions about new assets. The model automatically performs dynamic rebalancing, calculates and obtains data at the optimization point in time. S2. At the sub-account level, determine the starting point of the random scatter of asset allocation through planning and solving methods; S201. Solve the asset allocation planning in the dimension of expected investment return rate-volatility, and find the highest expected investment return rate within the unit volatility step; S202, preliminarily generating an effective frontier curve including multiple asset allocation points as a set of fixed points of preliminary asset allocation random scattered points; S3. At the sub-account level, randomly generate asset allocation points by means of random scattering; S301. Calibrate the parameters of Dirichlet Distribution according to the input parameters of the model. S302. Perform statistical probability distribution fitting based on Dirichlet Distribution, randomly simulate multiple asset allocation points, and the sum of the proportions of each asset allocation point is 1, thereby forming a first round of asset allocation point set. S4. At the sub-account level, multi-dimensional quantitative indicator calculations are performed for each asset allocation point; S401. Conduct quantitative modeling for various types of assets and liabilities and calculate various quantitative indicators, including but not limited to cash flow, present value, rate of return, volatility, value at risk, solvency minimum capital and stress testing; S402. Considering the asset and liability modeling information at the same time, taking into account the asset-liability linkage benefits, using the curve interpolation algorithm, calculate the quantitative indicators under the asset-liability linkage, including but not limited to net assets, duration gap, solvency adequacy ratio, net volatility, net value at risk and stress test net value; S403, based on the constraints input by the user and the quantitative indicators of each asset allocation point calculated in S401 and S402, the first round of asset allocation point set is screened to generate a second round of asset allocation point set; S5. At the sub-account level, the efficient frontier curve is generated through random simulation interpolation and random simulation smoothing calculation of asset allocation points; S501. In the expected investment return rate-volatility dimension, a plurality of effective frontier allocation points are randomly selected, and multiple random simulation interpolations are performed to form new asset allocation points, thereby forming a third round of asset allocation point set; S502, in the expected investment return rate-volatility dimension, based on the third round of asset allocation point set, select the second round of effective frontier allocation points, and perform smoothing calculation of the effective frontier allocation points to form the final effective frontier allocation points, form an effective frontier curve by connecting the effective frontier allocation points, and generate the fourth round of asset allocation point set; S503, the system stores the fourth round of asset allocation point set into the database for subsequent visualization and interactive operation presentation; S6. At the sub-account level, all asset allocation points are visualized and interactively presented; S601. Based on the horizontal axis (X axis) and vertical axis (Y axis) selected by the user, the fourth round of asset allocation point set is automatically displayed in two-dimensional quadrants, and the displayed content includes the calculated quantitative indicators, and the displayed forms include pictures and tables; S602: The user selects upper and lower limits for each quantitative indicator, and the model automatically performs filtering and displays it visually on the system interface; S7. At the sub-account level, the asset allocation optimization results are downloaded and exported. Users can distinguish between effective frontier allocation points and ineffective frontier allocation points for the fourth round of asset allocation points, and export the results in the form of pictures and Excel. S8. Optimize asset allocation at the total account level, visualize and export results based on sub-account optimization results; S801. Generate the possibility of all total account asset allocation points based on the asset allocation points optimized for each sub-account by using an exhaustive method; S802, based on the total account configuration point set, calculate each quantitative indicator (the algorithm is the same as S4); S803. Visually present all asset allocation points and interactively present them at the level of the total account (the algorithm is the same as S6); S804. At the general account level, export the asset allocation optimization results (the algorithm is the same as S7).
2. A quantitative algorithm for optimizing asset allocation in the financial field according to claim 1, characterized in that: Through random simulation probability distribution fitting algorithm, the possibility of asset allocation points is generated, and convergence effect is achieved faster, which is closer to the global optimal result.
3. A quantitative algorithm for optimizing asset allocation in the financial field according to claim 1, characterized in that: With extremely large amounts of data, the quantitative indicators of all assets and liabilities at all asset allocation points can be efficiently measured, compared, screened, visualized and interactively displayed.
4. A quantitative algorithm for optimizing asset allocation in the financial field according to claim 1, characterized in that: Using the curve interpolation algorithm, we can efficiently complete the asset-liability linkage between various asset types and liability cash flows based on multiple investment yield assumptions, greatly improving the degree of asset-liability matching in the asset allocation optimization process.
5. A quantitative algorithm for optimizing asset allocation in the financial field according to claim 1, characterized in that: Through the effective frontier random simulation interpolation and random simulation smoothing algorithm, an effective frontier curve with smooth changes in asset allocation plans between adjacent effective frontier asset allocation points is generated.
6. A quantitative algorithm for optimizing asset allocation in the financial field according to any one of claims 1 to 5, characterized in that: Provide a set of algorithms that combine sub-account optimization with total account optimization, and generate the possibility of total account asset allocation optimization results after sub-account asset allocation optimization based on the exhaustive method.