Investment portfolio determination method and device

Abstract

The invention discloses an investment portfolio determination method and device which are applied to the field of big data, and the method comprises the steps: obtaining asset data to be recombined, and carrying out the coding, thereby obtaining a plurality of quantum bits; the quantum bits indicate whether the asset data is incorporated into the portfolio; transforming the plurality of quantum bits by using a quantum gate to obtain a superposition state; the superposition state comprises all investment portfolios; for each investment portfolio, inputting the investment portfolio into the corresponding screening model for prediction to obtain fitness corresponding to the investment portfolio; the screening model is constructed based on an income guiding value and a risk guiding value corresponding to the investment portfolio; and screening out the investment portfolio with the highest fitness from all the investment portfolios contained in the superposition state as the optimal investment portfolio. In the application, the screening model provides a clear direction for screening the superposition state, and in a large-scale portfolio, the optimal portfolio can be found more quickly through quantum calculation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data, and particularly relates to a portfolio determination method and device. BACKGROUND

[0002] In the field of financial investment, portfolio optimization is an important task that investors need to focus on when making decisions, aiming to build a portfolio that can meet the yield target and control the risk among various assets (such as stocks, bonds, etc.).

[0003] Currently, traditional portfolio optimization methods mostly rely on classical computer algorithms, such as mean-variance optimization model, Monte Carlo simulation, etc. These methods predict the expected return, risk and correlation of assets through historical data and statistical analysis, and further build an optimal portfolio that meets the investor's target. However, as the size of the portfolio continues to expand (for example, a portfolio containing hundreds of assets or even more), the computational complexity of traditional algorithms grows exponentially, and when the market environment changes suddenly or new data is introduced, a lot of time is needed to recalculate the entire portfolio, which cannot quickly determine the optimal portfolio.

[0004] Therefore, how to quickly determine the optimal portfolio has become a problem that needs to be solved in the field. SUMMARY

[0005] The present application provides a portfolio determination method and device, aiming to quickly determine the optimal portfolio.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme:

[0007] A portfolio determination method, comprising:

[0008] Obtaining asset data to be recombined for encoding to obtain a plurality of qubits; the qubits indicate whether the asset data is included in the portfolio; wherein all portfolios of n assets are converted into qubits to obtain 2 n qubits;

[0009] Transforming the plurality of qubits using quantum gates to obtain a superposition state; the superposition state contains all portfolios; wherein n quantum gates act on the initial n ground states to generate a superposition state containing 2 n states; the ground state is the state of the qubit;

[0010] For each portfolio, input the portfolio into a corresponding screening model for prediction to obtain the fitness of the portfolio corresponding to the portfolio; the screening model is constructed based on the yield-oriented value and the risk-oriented value corresponding to the portfolio.

[0011] From all portfolios contained in the superposition state, the portfolio with the highest fitness is selected as the optimal portfolio.

[0012] Optionally, before the asset data to be recombined is encoded to obtain a plurality of qubits, the method further comprises:

[0013] Obtaining asset information to be recombined;

[0014] Preprocessing the asset information to be recombined to obtain preprocessed asset information;

[0015] Calculating the risk of the preprocessed asset information by using a shrinkage estimation covariance matrix to obtain a risk assessment result of the asset;

[0016] Inputting the preprocessed asset information into a prediction model to obtain an expected return rate of the asset;

[0017] Taking the risk assessment result of the asset and the expected return rate of the asset as the asset data to be recombined.

[0018] Optionally, the screening model is constructed based on the return-oriented value and the risk-oriented value corresponding to the portfolio, comprising:

[0019] Obtaining a preset target return rate and a risk-free return rate;

[0020] Determining a standard deviation and a lower standard deviation according to the risk assessment result and the expected return rate of the asset;

[0021] Calculating a first difference value between the expected return rate of the portfolio and the risk-free return rate, and calculating a ratio between the first difference value and the standard deviation to obtain a return-oriented value;

[0022] Calculating a second difference value between the expected return rate of the portfolio and the preset target return rate, and calculating a ratio between the second difference value and the standard deviation to obtain a risk-oriented value;

[0023] Calculating a sum value between the return-oriented value and the risk-oriented value, and calculating a ratio between the risk-oriented value and the sum value to obtain a weight;

[0024] The screening model is constructed based on the return-oriented value, the risk-oriented value, and the weight.

[0025] Optionally, the determination of the standard deviation and the lower standard deviation according to the risk assessment result and the expected return rate of the asset comprises:

[0026] Extracting the standard deviation from asset data of the asset;

[0027] For each of the portfolios, an expected return rate of the portfolio is calculated according to an expected return rate of assets in the portfolio;

[0028] From the expected return rates of all the portfolios, assets with an expected return rate less than the preset target return rate are screened out and identified as target portfolios;

[0029] A downside standard deviation is calculated according to the expected return rate of the target portfolios.

[0030] Optionally, the step of screening out, from all the portfolios contained in the superposition state, a portfolio with the highest fitness as an optimal portfolio comprises:

[0031] An initial iteration number is obtained;

[0032] The initial iteration number is adjusted according to the fitness to obtain an adjusted iteration number;

[0033] A portfolio with the highest fitness is screened out from the superposition state according to the adjusted iteration number as an optimal portfolio.

[0034] A portfolio determination apparatus comprises:

[0035] An encoding unit is configured to obtain asset data to be recombined and encode the asset data to obtain a plurality of qubits; the qubits indicate whether the asset data is included in a portfolio; wherein all portfolios of n assets are converted into qubits to obtain 2 n qubits;

[0036] A transformation unit is configured to transform the plurality of qubits by using quantum gates to obtain a superposition state; the superposition state contains all portfolios; wherein the superposition state containing 2 n states is generated by applying n quantum gates to initial n ground states; the ground state is a state of the qubit;

[0037] A prediction unit is configured to, for each of the portfolios, input the portfolio into a corresponding screening model to obtain a fitness corresponding to the portfolio; the screening model is constructed based on a return-oriented value and a risk-oriented value corresponding to the portfolio;

[0038] A screening unit is configured to screen out, from all the portfolios contained in the superposition state, a portfolio with the highest fitness as an optimal portfolio.

[0039] Optionally, the apparatus further comprises:

[0040] An obtaining unit is configured to obtain asset information to be recombined;

[0041] a preprocessing unit, configured to preprocess the asset information to be recombined to obtain preprocessed asset information;

[0042] a calculation unit, configured to perform risk calculation on the preprocessed asset information by using a shrinkage estimation covariance matrix to obtain a risk evaluation result of the asset;

[0043] an input unit, configured to input the preprocessed asset information into a prediction model to obtain an expected return rate of the asset;

[0044] a unit, configured to take the risk evaluation result of the asset and the expected return rate of the asset as asset data to be recombined.

[0045] Optionally, the prediction unit comprises:

[0046] an acquisition subunit, configured to acquire a preset target return rate and a risk-free return rate;

[0047] a determination subunit, configured to determine a standard deviation and a downside standard deviation according to the risk evaluation result of the asset and the expected return rate;

[0048] a first calculation subunit, configured to calculate a first difference between the expected return rate of the portfolio and the risk-free return rate, and calculate a ratio between the first difference and the standard deviation to obtain a return-oriented value;

[0049] a second calculation subunit, configured to calculate a second difference between the expected return rate of the portfolio and the preset target return rate, and calculate a ratio between the second difference and the standard deviation to obtain a risk-oriented value;

[0050] a third calculation subunit, configured to calculate a sum of the return-oriented value and the risk-oriented value, and calculate a ratio between the risk-oriented value and the sum to obtain a weight;

[0051] a fourth calculation subunit, configured to construct the screening model based on the return-oriented value, the risk-oriented value and the weight.

[0052] Optionally, the determination subunit is specifically configured to:

[0053] extract the standard deviation from asset data of the asset;

[0054] for each portfolio, calculate an expected return rate of the portfolio according to an expected return rate of an asset in the portfolio;

[0055] screen assets with an expected return rate less than the preset target return rate from expected return rates of all portfolios, and identify the assets as target portfolios;

[0056] According to the expected yield rate of the target portfolio, a downlink standard deviation is calculated.

[0057] Optionally, the screening unit is specifically used for:

[0058] An initial iteration number is obtained.

[0059] The initial iteration number is adjusted according to the fitness to obtain an adjusted iteration number.

[0060] The portfolio with the highest fitness is screened from the superposition state according to the adjusted iteration number as the optimal portfolio.

[0061] The technical solution provided in the application obtains asset data to be recombined for encoding to obtain a plurality of quantum bits; the quantum bits indicate whether the asset data is included in a portfolio; a quantum gate is used to transform the plurality of quantum bits to obtain a superposition state; the superposition state contains all portfolios; for each portfolio, the portfolio is input into a corresponding screening model for prediction to obtain a fitness corresponding to the portfolio; the screening model is constructed based on a yield-oriented value and a risk-oriented value corresponding to the portfolio; from all portfolios contained in the superposition state, a portfolio with the highest fitness is screened as an optimal portfolio. In the application, the screening model provides a clear direction for screening the superposition state, and in a large-scale portfolio, quantum computing can find the optimal portfolio more quickly. BRIEF DESCRIPTION OF DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0063] Figure 1 A flowchart of a portfolio determination method provided in an embodiment of the present application;

[0064] Figure 2 A schematic diagram of a candidate scheme generation flow provided in an embodiment of the present application;

[0065] Figure 3 An architectural schematic diagram of a portfolio determination device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0067] In this application, the terms "comprises," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0068] like Figure 1 FIG. 1 is a flow chart of a method for determining an investment portfolio provided in an embodiment of the present application, including:

[0069] S101: Obtain asset data to be reassembled and encode it to obtain multiple quantum bits.

[0070] Among them, quantum bits indicate whether an asset is included in the portfolio.

[0071] Specifically, the asset data includes at least the expected rate of return and risk assessment results of the assets.

[0072] It can be understood that encoding the asset data to be reassembled can obtain multiple quantum bits, that is, converting all portfolios of n assets (n is a positive integer greater than 0) into quantum bits, and each asset can be included in the portfolio (represented as 1) or not included in the portfolio (represented as 0). Therefore, there are a total of 2 n possible combinations. We encode these combinations as quantum states.

[0073] For example, there are two assets, asset 1 and asset 2. Encoding each asset results in 4 possible combinations, namely (0,0), (0,1), (1,0) and (1,1).

[0074] In addition, when n=30, more than 1 billion combinations can be searched simultaneously. Through quantum computing, all possible combinations can be explored in a very short time.

[0075] Optionally, before step S101 , process A1 to process A4 are also included.

[0076] A1: Obtain the asset information to be reassembled.

[0077] Among them, assets include but are not limited to: stocks and bonds.

[0078] The asset information includes but is not limited to: the historical price of the asset, the historical transaction volume of the asset, and the historical rate of return of the asset.

[0079] A2: The asset information to be reassembled is pre-processed to obtain pre-processed asset information.

[0080] It is understood that the asset information to be reassembled is pre-processed (including data cleaning and data deduplication). Specifically, the asset information to be reassembled is checked to see if it is within the standard deviation range (for example, the mean ± 3 times the standard deviation). If the asset information to be reassembled is not within the standard deviation range, the asset information to be reassembled is deleted, and the remaining asset information is the cleaned asset data. The cleaned asset information is deduplicated using a hash algorithm to obtain the pre-processed asset data.

[0081] It should be noted that by preprocessing asset information, reliable basic data can be provided for subsequent risk calculations and return forecasts, avoiding noise interference in the results.

[0082] A3: Use the shrinkage estimated covariance matrix to calculate the risk of the preprocessed asset information and obtain the asset risk assessment result.

[0083] Among them, the risk assessment results include at least standard deviation and covariance matrix.

[0084] Specifically, the specific form of the shrinkage estimation covariance matrix is ​​shown in formula (1).

[0085] (1).

[0086] In formula (1), The sample covariance matrix reflects the mutual changes between asset returns; The mean of the sample variances; I is the identity matrix; is the shrinkage coefficient, which controls the relationship between the sample covariance matrix and the shrinkage target ( ), Estimated covariance matrix for shrinkage.

[0087] It should be noted that for a single asset, risk is usually the standard deviation of the asset's historical returns; for a portfolio, risk is the square root of the weighted covariance matrix.

[0088] A4: Input the preprocessed asset information into the forecasting model to obtain the expected rate of return of the asset.

[0089] Optionally, the prediction model includes but is not limited to an ARIMA model.

[0090] Specifically, the specific form of the prediction model is shown in formula (2).

[0091] y t =c+φ1*y t-1 +...+φ p *y t-p +θ1*ε t-1 +...+θ q *ε t-q (2).

[0092] In formula (2), y t is the asset price at time t, c is a constant term, φ and θ are model parameters determined by fitting historical data, ε is a white noise sequence representing random fluctuations that cannot be explained by the model, y t-1 is the asset price at historical time, φ p is the autoregressive coefficient, θ q is the moving average coefficient, and ε t-q is the white noise sequence value.

[0093] It should be noted that the preprocessed asset information is input into the prediction model, and the prediction model mines the rules from the historical asset prices to predict the future price trend of the asset, thereby obtaining the expected return rate.

[0094] For example, by analyzing the time series of the past prices of a certain stock, the parameters of the ARIMA model can be determined to predict future stock prices. These predicted asset prices are used to calculate expected returns and other related indicators such as expected return rates. Then, these return indicators are combined with risk indicators calculated by risk to comprehensively present the risk-return characteristics of the investment portfolio. These information constitutes the input data required for quantum computing, enabling quantum computing to consider risk and return when searching for quantum state space, and thus finding the optimal investment portfolio solution to maximize returns under controllable risk or minimize risk under expected returns.

[0095] S102: Transforming a plurality of quantum bits using quantum gates to obtain a superposition state.

[0096] Wherein, the superposition state contains all investment portfolios.

[0097] Specifically, the specific form of the superposition state is shown in formula (3).

[0098] (3).

[0099] In formula (3), |0> represents that n quantum bits are in |0> state, |1> represents that n quantum bits are in |1> state, For each combination of amplitudes, an n-qubit superposition state is generated by applying n Hadamard gates to the initial n-qubit ground state containing 2nstates n For example, when n = 3, it simultaneously encompasses 8 different portfolio states (from |000 to |111.

[0100] It is important to note that all subsequent operations are performed on all possible portfolios represented by this superposition state, searching and optimizing for portfolios that meet investment objectives (such as maximizing returns, minimizing risk, etc.).

[0101] S103: For each portfolio, input the portfolio into the corresponding screening model for prediction, to obtain the fitness of the portfolio corresponding to the portfolio.

[0102] Among them, the screening model is constructed based on the yield-oriented value and the risk-oriented value corresponding to the portfolio.

[0103] Optionally, in another embodiment of the present application, the specific implementation of constructing the screening model based on the yield-oriented value and the risk-oriented value corresponding to the portfolio includes processes B1 to B6:

[0104] B1: Obtain a preset target yield and a risk-free yield.

[0105] Among them, the preset target yield is an expected yield level set by the investor in advance. The investor can determine this value according to his own investment goals, risk tolerance and other factors.

[0106] Optionally, the risk-free yield is an investment return with relatively stable yield and can be regarded as having no risk.

[0107] B2: According to the risk assessment result and the expected yield of the asset, determine the standard deviation and the lower standard deviation.

[0108] Among them, the standard deviation measures the total volatility of the yield, including both positive volatility (yield exceeds average yield) and negative volatility (yield is lower than average yield).

[0109] Optionally, the lower standard deviation only measures the fluctuations that are unfavorable to the investor, that is, only the part below a certain target yield is calculated. It ignores the positive volatility, so it reflects the investor's sensitivity to loss more than the standard deviation.

[0110] Optionally, in another embodiment of the present application, the specific implementation of step B2 includes C1 to C4: ​​

[0111] C1: Extracting the standard deviation from the asset data of the assets.

[0112] It should be noted that since the shrinkage estimation covariance matrix contains the square of the standard deviation, the standard deviation can be directly extracted from the shrinkage estimation covariance matrix.

[0113] In addition, the difference between the yield data of each portfolio and the expected yield can be calculated first, and then the average of the squares of these differences is calculated, and finally the standard deviation is obtained by taking the square root.

[0114] C2: For each portfolio, the expected yield of the portfolio is calculated according to the expected yield of the assets in the portfolio.

[0115] It can be understood that the expected yield of the portfolio is calculated according to the expected yield of the assets in the portfolio, specifically, the weighted average of the expected yields of each asset and their weights in the portfolio is calculated, thereby obtaining the expected yield of the portfolio.

[0116] C3: From all the expected yields of the portfolios, the assets with an expected yield less than a preset target yield are screened out and identified as target portfolios.

[0117] For example, the yield data of the portfolios is R1, R2,..., R n , and the preset target yield is T t, , and the assets with an expected yield less than the preset target yield are screened out from all the expected yields of the portfolios and identified as target portfolios.

[0118] C4: According to the expected yield of the target portfolio, the downside standard deviation is calculated.

[0119] It should be noted that the specific implementation process of calculating the downside standard deviation according to the expected yield of the target portfolio is as follows: the variance of the expected yield of the target portfolio is calculated using the variance calculation formula, and the square root of the variance is taken to obtain the downside standard deviation. The specific form of the variance calculation formula is: , m is the number of assets with an expected yield less than the preset target yield, R i is the expected yield of the target portfolio, T t is the preset target yield, is the variance, , is the downside standard deviation.

[0120] B3: Calculating the first difference between the expected yield of the portfolio and the risk-free yield, and calculating the ratio between the first difference and the standard deviation to obtain the yield-oriented value.

[0121] The first difference value between the expected return rate of the portfolio and the risk-free return rate is calculated, and a ratio between the first difference value and the standard deviation is calculated to obtain a specific form of the return-oriented value, as shown in formula (4).

[0122] (4).

[0123] In formula (4), Sharpe is the return-oriented value, μ is the expected return rate of the portfolio, T f is the risk-free return rate, and σ is the standard deviation.

[0124] B4: A second difference value between the expected return rate of the portfolio and a preset target return rate is calculated, and a ratio between the second difference value and the standard deviation is calculated to obtain a risk-oriented value.

[0125] The second difference value between the expected return rate of the portfolio and the preset target return rate is calculated, and a ratio between the second difference value and the standard deviation is calculated to obtain a specific form of the risk-oriented value, as shown in formula (5).

[0126] (5).

[0127] In formula (5), Sortino is the risk-oriented value, μ is the expected return rate of the portfolio, T t is the preset target return rate, is the downside standard deviation.

[0128] B5: A sum value between the return-oriented value and the risk-oriented value is calculated, and a ratio between the risk-oriented value and the sum value is calculated to obtain a weight.

[0129] The sum value between the return-oriented value and the risk-oriented value is calculated, and a specific form of a ratio between the risk-oriented value and the sum value is calculated, as shown in formula (6).

[0130] (6).

[0131] In formula (6), α is the weight.

[0132] It should be noted that the weight determines the priority of the return and the risk in the candidate scheme generated by the quantum calculation, and further affects the Sharpe / Sortino ratio of the final portfolio. Under a high α, the candidate scheme may sacrifice part of the downside protection in exchange for a higher expected return (Sharpe rises, and Sortino may decrease). Under a low α, the candidate scheme pays more attention to avoiding tail risk (Sortino rises, and Sharpe may decrease).

[0133] In addition, if the Sortino ratio significantly increases in searching for the optimal portfolio in a certain round, the alpha will decrease to avoid excessive bias towards risk aversion; if the Sharpe ratio is consistently higher than expected, the alpha will increase to strengthen the yield-oriented search.

[0134] S306: Based on the yield-oriented value, the risk-oriented value and the weight, a screening model is constructed.

[0135] The specific implementation process of constructing the screening model based on the yield-oriented value, the risk-oriented value and the weight is as follows: calculating the first product of the yield-oriented value and the weight, calculating the difference between 1 and the weight, calculating the second product of the difference and the risk-oriented value, calculating the sum of the first product and the second product, and obtaining the screening model. Its specific form can be seen in formula (7).

[0136] Fitness = a * Sharpe + (1-a) * Sortino (7).

[0137] In formula (7), Fitness is the fitness.

[0138] S104: From all the investment portfolios contained in the superposition state of the screening model, the investment portfolio with the highest fitness is selected as the optimal investment portfolio.

[0139] Among them, the optimal investment portfolio is an investment portfolio that achieves the maximum return under a given risk level by reasonably allocating different assets.

[0140] It can be understood that by embedding the screening model into the quantum state and adjusting the parameters of the quantum rotation gate (which are related to the alpha, the Sharpe ratio and the Sortino ratio), the quantum state can gradually tend to the state required by the screening model during the evolution process, that is, the investment portfolio state with better risk-return trade-off. In this way, the introduction of the screening model provides a clear direction for the search of quantum computing in the quantum state space. Quantum computing will explore each possible state in the quantum state space in the direction of maximizing (or minimizing, depending on the specific goal) the fitness. For example, if the current investment strategy focuses more on yield, by setting the alpha, the weight of the Sharpe ratio in the screening model will be larger, and quantum computing will tend to search for investment portfolio states that can improve the Sharpe ratio.

[0141] It should be noted that the optimal investment portfolio, the asset allocation ratio contained in the investment portfolio, the risk assessment result, the expected return rate and other information are generated into a candidate scheme and fed back.

[0142] Optionally, in another embodiment of the present application, the specific implementation of step S104 includes processes D1 to D3:

[0143] D1: Obtain the initial iteration number.

[0144] D2: Adjust the initial iteration number according to the fitness to obtain the adjusted iteration number.

[0145] It should be noted that the Sharpe ratio in the screening function directly affects the dynamic iteration formula. If the fitness is high, S is large, the initial iteration number T increases, and the fine search strategy is adopted. If the fitness is low, S is small, the initial iteration number T decreases, and the fast adjustment strategy is adopted. That is, when the fitness is large, it means that the current return-risk situation is good, and the initial iteration number needs to be increased. When the fitness is small, it means that the current return-risk situation is poor, and the initial iteration number needs to be reduced.

[0146] wherein the specific form of the dynamic iteration formula is shown in equation (8).

[0147] (8).

[0148] In equation (8), T is the initial iteration number, M represents the number of measurements per iteration; λ is the adaptive coefficient, which is calculated by λ=tanh(|ΔS|), where ΔS is the change in Sharpe ratio; S is the real-time Sharpe ratio.

[0149] D3: According to the adjusted iteration number of the screening model, the investment portfolio with the highest fitness is selected from the screening model superposition state as the optimal investment portfolio.

[0150] It can be understood that the specific implementation process of selecting the optimal investment portfolio from the superposition state according to the adjusted iteration number is as follows: after the iteration number is reached, the objective function in the iteration process is obtained, the maximum objective function is selected from all objective functions, and if the maximum objective function reaches the target value, the investment portfolio corresponding to the maximum objective function is determined as the optimal investment portfolio. Or, the change amplitude of the objective function is analyzed, and if the change amplitude is less than a preset threshold (for example, 0.01) or does not exist, the investment portfolio corresponding to the objective function without change is determined as the optimal investment portfolio.

[0151] From the above content, it can be seen that the dynamic iteration control and the embedding of the objective function are closely coordinated. The objective function determines the direction and target of the search, and the dynamic iteration control adjusts the iteration number and the search strategy according to the feedback of the objective function (such as the change of the real-time Sharpe ratio). For example, when the weight of the Sharpe ratio in the objective function is large, the dynamic iteration control will pay more attention to the change of the Sharpe ratio, and adjust the iteration number according to the change to better meet the optimization demand of the objective function. In this way, the dynamic iteration control helps to gradually approach the optimal investment portfolio in the quantum state space.

[0152] To better illustrate the above, see Figure 2 A schematic diagram of the generation process of a candidate solution is shown. Figure 2 The classical module, parameter optimizer and quantum module are contained in the Figure 2 The classical module is used to clean the asset data, and the cleaned asset data is input into the yield prediction model (i.e. Figure 2 in the prediction module) to obtain the expected yield rate; the risk calculation is performed on the cleaned asset data to obtain the risk assessment result (including the standard deviation and the covariance matrix). The parameter optimizer is used to calculate the weight and the adaptive coefficient, and in the screening model, the weight of the Sharpe ratio and the Sortino ratio is determined, and by adjusting the weight, the degree of emphasis on risk and yield can be changed. The adaptive coefficient is in the dynamic iteration control formula (i.e. in the dynamic iteration control formula), according to the Sharpe ratio (i.e.

[0153] the proportion of summer in the figure) and the time fluctuation of the portfolio, the iteration number is dynamically adjusted. The quantum module is used to select the optimal portfolio from the superposition state according to the fitness and the iteration number, and to generate a candidate solution according to the optimal portfolio.

[0154] As shown in Figure 3 , an architecture schematic diagram of a portfolio determination apparatus provided by an embodiment of the present application is shown, and the determination apparatus comprises an encoding unit 100, a transformation unit 200, a prediction unit 300 and a screening unit 400.

[0155] The encoding unit 100 is configured to obtain asset data to be recombined and encode the asset data to obtain a plurality of quantum bits; the screening model quantum bits indicate whether the asset data is included in the portfolio; wherein all portfolios of n assets are converted into quantum bits to obtain 2 n quantum bits.

[0156] The transformation unit 200 is configured to transform the plurality of screening model quantum bits by using quantum gates to obtain a superposition state; the screening model superposition state contains all portfolios; wherein the initial n ground states are acted on by n quantum gates to generate a superposition state containing 2 n states; the screening model ground state is the state of the quantum bit.

[0157] The prediction unit 300 is configured to input the portfolio into the corresponding screening model for prediction to obtain the fitness corresponding to the portfolio for each portfolio; the screening model is constructed based on the yield-oriented value and the risk-oriented value corresponding to the portfolio.

[0158] The prediction unit 300 includes:

[0159] Get subunits, used to obtain preset target rate of return and risk-free rate of return.

[0160] Determine subunits for determining standard deviation and downside standard deviation based on risk assessment results and expected rate of return of the asset.

[0161] The determination subunit is specifically used for: extracting the standard deviation from the asset data of the assets; for each investment portfolio, calculating the expected rate of return of the investment portfolio based on the expected rates of return of the assets in the investment portfolio; screening out the assets with a rate of return less than the preset target rate of return of the screening model from the expected rates of return of all investment portfolios, and marking them as target investment portfolios; and calculating the downside standard deviation based on the expected rate of return of the target investment portfolio of the screening model.

[0162] The first calculation subunit is used to calculate the first difference between the expected rate of return of the investment portfolio and the risk-free rate of return of the screening model, and calculate the ratio between the first difference of the screening model and the standard deviation of the screening model to obtain the return-oriented value.

[0163] The second calculation subunit is used to calculate the second difference between the expected rate of return of the investment portfolio and the preset target rate of return of the screening model, and calculate the ratio between the second difference of the screening model and the standard deviation of the screening model to obtain the risk-oriented value.

[0164] The third calculation subunit is used to calculate the sum of the screening model benefit-oriented value and the screening model risk-oriented value, and calculate the ratio of the screening model risk-oriented value to the screening model sum value to obtain the weight.

[0165] The fourth calculation subunit is used to construct a screening model based on the screening model return-oriented value, the screening model risk-oriented value and the screening model weight.

[0166] The screening unit 400 is used to screen out the investment portfolio with the highest fitness as the optimal investment portfolio from all investment portfolios included in the superposition state of the screening model.

[0167] The screening unit 400 is specifically used to: obtain the initial number of iterations; adjust the initial number of iterations of the screening model according to the fitness of the screening model to obtain the adjusted number of iterations; and screen out the investment portfolio with the highest fitness from the superposition state of the screening model as the optimal investment portfolio based on the adjusted number of iterations of the screening model.

[0168] In summary, the objective function provides a clear direction for screening superposition states. In large-scale investment portfolios, quantum computing can find the optimal investment portfolio more quickly.

[0169] Combine Figure 3As shown in the content, the determining apparatus further comprises:

[0170] An acquisition unit is configured to acquire asset information to be recombined.

[0171] A preprocessing unit is configured to preprocess the asset information to be recombined of the screening model to obtain preprocessed asset information.

[0172] A calculation unit is configured to perform risk calculation on the preprocessed asset information of the screening model by using a shrinkage estimation covariance matrix to obtain a risk evaluation result of the asset.

[0173] An input unit is configured to input the preprocessed asset information of the screening model into a prediction model to obtain an expected yield of the asset.

[0174] An output unit is configured to output the risk evaluation result of the asset and the expected yield of the asset as asset data to be recombined.

[0175] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts of each embodiment can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the system or system embodiments, since they are basically similar to the method embodiments, they are described more simply. The relevant parts can be referred to the part of the method embodiment. The system and system embodiments described above are only illustrative. The units described as separate components can be or can not be physically separated. The components shown as units can be or can not be physical units, that is, they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0176] The skilled person can further realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in the above description. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical scheme. Skilled persons can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0177] The foregoing description of the disclosed embodiments enables a person skilled in the art to make or use the application. Modifications of these embodiments will occur to persons of skill in the art, and that the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for determining an investment portfolio, characterized in that: include: Obtain the asset data to be recombined and encode it to obtain multiple quantum bits; The qubit indicates whether the asset data is included in the investment portfolio; wherein, all investment portfolios of n assets are converted into qubits, and 2 n qubits; The quantum gate is used to transform the plurality of quantum bits to obtain a superposition state; the superposition state includes all investment portfolios; wherein, n quantum gates act on the initial n basis states to generate a superposition state including 2 n The ground state is the state of the quantum bit; For each investment portfolio, the investment portfolio is input into a corresponding screening model for prediction to obtain the fitness of the investment portfolio; the screening model is constructed based on the return-oriented value and risk-oriented value corresponding to the investment portfolio; From all investment portfolios contained in the superposition state, the investment portfolio with the highest fitness is selected as the optimal investment portfolio.

2. The method according to claim 1, characterized in that Before obtaining the asset data to be recombined and encoding to obtain a plurality of quantum bits, the method further includes: Obtain information about assets to be reassembled; Preprocessing the asset information to be reassembled to obtain preprocessed asset information; Performing risk calculation on the pre-processed asset information using a shrinkage estimated covariance matrix to obtain an asset risk assessment result; Inputting the pre-processed asset information into a prediction model to obtain the expected rate of return of the asset; The risk assessment results of the assets and the expected rate of return of the assets are used as the asset data to be reassembled.

3. The method according to claim 1, characterized in that Constructing the screening model based on the return-oriented value and the risk-oriented value corresponding to the investment portfolio includes: Obtain preset target rate of return and risk-free rate of return; Determine the standard deviation and downside standard deviation based on the risk assessment results and expected rate of return of the asset; Calculating a first difference between the expected rate of return of the investment portfolio and the risk-free rate of return, and calculating a ratio between the first difference and the standard deviation to obtain a return-oriented value; Calculating a second difference between the expected rate of return of the investment portfolio and the preset target rate of return, and calculating a ratio between the second difference and the standard deviation to obtain a risk-oriented value; Calculating a sum of the benefit-oriented value and the risk-oriented value, and calculating a ratio of the risk-oriented value to the sum to obtain a weight; The screening model is constructed based on the benefit-oriented value, the risk-oriented value and the weight.

4. The method according to claim 3, characterized in that Determining the standard deviation and downside standard deviation based on the risk assessment results and expected rate of return of the asset includes: Extract standard deviation from asset data for an asset; For each investment portfolio, the expected rate of return of the investment portfolio is calculated based on the expected rates of return of the assets in the investment portfolio; Screening out assets with expected returns less than the preset target return from all investment portfolios and marking them as target investment portfolios; Based on the expected rate of return of the target investment portfolio, the downside standard deviation is calculated.

5. The method according to claim 1, wherein The step of selecting the investment portfolio with the highest fitness from all investment portfolios included in the superposition state as the optimal investment portfolio includes: Get the initial number of iterations; Adjusting the initial number of iterations according to the fitness to obtain an adjusted number of iterations; The investment portfolio with the highest fitness is selected from the superposition state according to the adjusted number of iterations as the optimal investment portfolio.

6. A device for determining an investment portfolio, characterized in that: include: An encoding unit, used to obtain the asset data to be recombined and encode it to obtain multiple quantum bits; The qubit indicates whether the asset data is included in the investment portfolio; wherein, all investment portfolios of n assets are converted into qubits, and 2 n qubits; A transformation unit is used to transform the plurality of quantum bits using quantum gates to obtain a superposition state; the superposition state includes all investment portfolios; wherein, n quantum gates act on the initial n basis states to generate a superposition state including 2 n The ground state is the state of the quantum bit; a prediction unit configured to input each investment portfolio into a corresponding screening model for prediction to obtain a fitness value corresponding to the investment portfolio; the screening model is constructed based on a return-oriented value and a risk-oriented value corresponding to the investment portfolio; The screening unit is used to screen out the investment portfolio with the highest fitness as the optimal investment portfolio from all investment portfolios included in the superposition state.

7. The device according to claim 6, characterized in that Also includes: An acquisition unit, used for acquiring asset information to be reassembled; a preprocessing unit, configured to preprocess the asset information to be reassembled to obtain preprocessed asset information; a calculation unit, configured to perform risk calculation on the pre-processed asset information using a shrinkage estimated covariance matrix to obtain a risk assessment result of the asset; An input unit, configured to input the pre-processed asset information into a prediction model to obtain an expected rate of return on the asset; As a unit, it is used to take the risk assessment results of assets and the expected rate of return of assets as the asset data to be recombined.

8. The device according to claim 6, characterized in that The prediction unit includes: Obtain subunits, used to obtain preset target rate of return and risk-free rate of return; Determine subunits for determining standard deviation and downside standard deviation based on risk assessment results and expected returns of assets; a first calculation subunit, configured to calculate a first difference between the expected rate of return of the investment portfolio and the risk-free rate of return, and calculate a ratio between the first difference and the standard deviation to obtain a return-oriented value; a second calculation subunit, configured to calculate a second difference between the expected rate of return of the investment portfolio and the preset target rate of return, and calculate a ratio between the second difference and the standard deviation to obtain a risk-oriented value; a third calculation subunit, configured to calculate a sum of the benefit-oriented value and the risk-oriented value, and calculate a ratio of the risk-oriented value to the sum to obtain a weight; The fourth calculation subunit is used to construct the screening model based on the return-oriented value, the risk-oriented value and the weight.

9. The device according to claim 8, characterized in that The determining subunit is specifically configured to: Extract standard deviation from asset data for an asset; For each investment portfolio, the expected rate of return of the investment portfolio is calculated based on the expected rates of return of the assets in the investment portfolio; Screening out assets with expected returns less than the preset target return from all investment portfolios and marking them as target investment portfolios; Based on the expected rate of return of the target investment portfolio, the downside standard deviation is calculated.

10. The device according to claim 6, characterized in that The screening unit is specifically used for: Get the initial number of iterations; Adjusting the initial number of iterations according to the fitness to obtain an adjusted number of iterations; The investment portfolio with the highest fitness is selected from the superposition state according to the adjusted number of iterations as the optimal investment portfolio.

Images