Investment portfolio recommendation method and device based on intelligent agent model, equipment and storage medium

Abstract

The invention discloses an intelligent agent model-based portfolio recommendation method, apparatus and device, and a storage medium, and relates to the technical field of finance, and the method comprises the steps of receiving a recommendation request instruction of a target user; training the intelligent agent model through a deep reinforcement learning algorithm based on historical data; performing feature analysis on the historical data through the intelligent agent model to obtain index data features of the historical data; based on the index data characteristics of the historical data, generating an investment portfolio under a simulation market scene from the real-time data through a Monte Carlo Markov chain model; and evaluating the investment portfolio under the simulation market scene to obtain a pre-recommended investment portfolio, and providing the pre-recommended investment portfolio to the target user. By combining the intelligent agent model with the Monte Carlo Markov chain model, the capability of capturing market complexity and dynamics during portfolio recommendation can be improved.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and in particular to an investment portfolio recommendation method, device, equipment and storage medium based on an intelligent agent model. Background Art

[0002] Portfolio recommendation is to obtain and recommend a portfolio with the highest return under controllable risk by selecting appropriate asset classes, allocating assets according to the characteristics of assets, determining the weights of asset classes to be held, and optimizing the weights of assets in the same asset class. Many financial applications require portfolio recommendations, but existing portfolio recommendation methods mainly rely on traditional financial theories and models, such as Modern Portfolio Theory (MPT) and Capital Asset Pricing Model (CAPM), which are difficult to capture the complexity and dynamics of the market.

[0003] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0004] The main purpose of this application is to provide an investment portfolio recommendation method, device, equipment and storage medium based on an intelligent agent model, aiming to solve the technical problem that it is difficult to capture the complexity and dynamics of the market when recommending investment portfolios.

[0005] To achieve the above objectives, the present application proposes an investment portfolio recommendation method based on an intelligent agent model, wherein the method comprises:

[0006] Receive a recommendation request instruction from a target user;

[0007] Train the intelligent agent model through deep reinforcement learning algorithm based on historical data;

[0008] Perform feature analysis on historical data through intelligent agent model to obtain the indicator data features of historical data;

[0009] Based on the indicator data characteristics of historical data, the real-time data is used through the Monte Carlo Markov chain model to generate an investment portfolio under a simulated market scenario;

[0010] The investment portfolio under the simulated market scenario is evaluated to obtain a pre-recommended investment portfolio, and the pre-recommended investment portfolio is provided to the target user.

[0011] In one embodiment, the step of training the intelligent agent model through a deep reinforcement learning algorithm based on historical data includes:

[0012] Initialize the policy network and value function network of the intelligent agent model based on historical data;

[0013] Calculating the advantage function and policy ratio of the deep reinforcement learning algorithm;

[0014] Based on the advantage function and the policy ratio, a clipping mechanism is used to define the loss function that minimizes the deep reinforcement learning algorithm.

[0015] Minimizing the loss function by gradient descent method to train and update the policy network of the intelligent agent model;

[0016] The value function network of the intelligent agent model is trained and updated by minimizing the mean square error.

[0017] In one embodiment, the step of generating an investment portfolio corresponding to the historical data in a simulated market scenario through a Monte Carlo Markov chain model includes:

[0018] Calculate rate of return based on historical data;

[0019] Determine a probability model according to the distribution characteristics of the yield;

[0020] Based on the probability model, random sampling is performed through the Monte Carlo Markov chain algorithm to generate investment portfolios in simulated market scenarios.

[0021] In one embodiment, the step of generating an investment portfolio in a simulated market scenario by random sampling through a Monte Carlo Markov chain algorithm based on the probability model includes:

[0022] Determine the state space based on the portfolio in historical data;

[0023] Random sampling is performed in the state space to construct a Markov chain and obtain an investment portfolio under a simulated market scenario.

[0024] In one embodiment, the step of constructing a Markov chain by random sampling in the state space to obtain an investment portfolio under a simulated market scenario includes:

[0025] determining a current state in the state space;

[0026] Based on the current state, generating a candidate state by random perturbation;

[0027] Obtaining an acceptance probability of the candidate state according to the current state and the candidate state;

[0028] Determining a next state of the current state based on the acceptance probability of the candidate state;

[0029] A Markov chain is constructed in the state space based on the current state and the next state to obtain an investment portfolio under a simulated market scenario.

[0030] In one embodiment, the step of determining the next state of the current state based on the acceptance probability of the candidate state includes:

[0031] When the acceptance probability of the candidate state is greater than or equal to a preset acceptance threshold, the candidate state is determined as the next state of the current state;

[0032] When the acceptance probability of the candidate state is less than a preset acceptance threshold, the current state is retained as the next state.

[0033] In one embodiment, the step of evaluating the investment portfolio under the simulated market scenario to obtain a pre-recommended investment portfolio and providing the pre-recommended investment portfolio to the target user comprises:

[0034] When the Markov chain is stably distributed, the investment portfolio on the Markov chain is analyzed, a pre-recommended investment portfolio is obtained according to a preset target, and the pre-recommended investment portfolio is recommended to the target user.

[0035] In addition, to achieve the above purpose, the present application also proposes an investment portfolio recommendation device based on an intelligent agent model, and the investment portfolio recommendation device based on an intelligent agent model includes:

[0036] An instruction receiving module, used for receiving a recommendation request instruction from a target user;

[0037] The feature generation module is used to perform feature analysis on historical data through an intelligent agent model to obtain the indicator data features of the historical data;

[0038] A portfolio generation module, for generating an investment portfolio in a simulated market scenario by using a Monte Carlo Markov chain model based on the indicator data characteristics of the historical data;

[0039] The optimal portfolio module is used to evaluate the investment portfolio under the simulated market scenario and obtain the optimal investment portfolio.

[0040] In addition, to achieve the above-mentioned purpose, the present application also proposes an investment portfolio recommendation device based on an intelligent agent model, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the investment portfolio recommendation method based on the intelligent agent model as described above.

[0041] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the investment portfolio recommendation method based on the intelligent agent model as described above are implemented.

[0042] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the investment portfolio recommendation method based on the intelligent agent model as described above.

[0043] One or more technical solutions proposed in this application have at least the following technical effects:

[0044] The embodiment of the present application proposes a method, device, equipment and storage medium for recommending an investment portfolio based on an intelligent agent model, receiving a recommendation request instruction from a target user; training the intelligent agent model through a deep reinforcement learning algorithm based on historical data; performing feature analysis on historical data through the intelligent agent model to obtain the index data characteristics of the historical data; based on the index data characteristics of the historical data, using the Monte Carlo Markov chain model to generate an investment portfolio in a simulated market scenario; evaluating the investment portfolio in the simulated market scenario to obtain a pre-recommended investment portfolio, and providing the pre-recommended investment portfolio to the target user. The combination of the intelligent agent model and the Monte Carlo Markov chain model can improve the ability to capture market complexity and dynamics when recommending investment portfolios. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0047] Figure 1 A flowchart of the first embodiment of the investment portfolio recommendation method based on the intelligent agent model of the present application is provided;

[0048] Figure 2 This is a schematic diagram of the module structure of an investment portfolio recommendation device based on an intelligent agent model according to an embodiment of the present application;

[0049] Figure 3 Schematic diagram of the device structure of the hardware operating environment involved in the investment portfolio recommendation method based on the intelligent agent model in the embodiment of the present application.

[0050] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0051] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0052] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0053] The main solution of the embodiment of the present application is: receiving a recommendation request instruction from a target user; training an intelligent agent model based on historical data through a deep reinforcement learning algorithm; performing feature analysis on historical data through an intelligent agent model to obtain index data features of the historical data; based on the index data features of the historical data, using the Monte Carlo Markov chain model to generate an investment portfolio in a simulated market scenario; evaluating the investment portfolio in the simulated market scenario to obtain a pre-recommended investment portfolio, and providing the pre-recommended investment portfolio to the target user. Training an intelligent agent model based on historical data through a deep reinforcement learning algorithm can improve the decision-making ability of the intelligent agent model to make investment portfolio recommendations.

[0054] Portfolio recommendation is to obtain and recommend a portfolio with the highest return under controllable risk by selecting appropriate asset classes, allocating assets according to the characteristics of the assets, determining the weights of the asset classes to be held, and optimizing the weights of assets in the same asset class. Many applications require portfolio recommendations, but existing portfolio recommendation methods mainly rely on traditional financial theories and models, such as Modern Portfolio Theory (MPT) and Capital Asset Pricing Model (CAPM), which are difficult to capture the complexity and dynamics of the market.

[0055] This application can improve the ability to capture market complexity and dynamics when recommending investment portfolios by combining an intelligent agent model with a Monte Carlo Markov chain model.

[0056] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, an investment portfolio recommendation device based on an intelligent agent model, etc. The following takes an investment portfolio recommendation device based on an intelligent agent model as an example to illustrate this embodiment and the following embodiments.

[0057] Based on this, the embodiment of the present application provides an investment portfolio recommendation method based on an intelligent agent model. Figure 1 , Figure 1 This is a flow chart of the first embodiment of the investment portfolio recommendation method based on the intelligent agent model of the present application.

[0058] In this embodiment, the investment portfolio recommendation method based on the intelligent agent model includes steps S10 to S50:

[0059] Step S10: receiving a recommendation request instruction from a target user.

[0060] It should be noted that the recommendation request instruction refers to a specific instruction issued by the target user to request the device to recommend an investment portfolio.

[0061] Specifically, the target user submits a recommendation request instruction on the device interface, the device receives the recommendation request instruction sent by the target user, verifies the identity of the target user, parses the information of the recommendation request instruction after the verification, and performs subsequent training and portfolio generation operations after the parsing is completed.

[0062] Step S20, training the intelligent agent model through a deep reinforcement learning algorithm based on historical data.

[0063] It should be noted that the deep reinforcement learning algorithm DRL (Deep Reinforcement Learning) is a combination of deep learning and reinforcement learning. It optimizes the reinforcement learning algorithm through deep learning techniques such as neural networks. The deep reinforcement learning algorithm is a machine learning method that aims to solve optimization problems with continuous states and action spaces. The deep reinforcement learning algorithm can be a deep Q network (DQN), a deep deterministic policy gradient (DDPG) or a proximal policy optimization (PPO). The deep Q network is mainly used for discrete action spaces and scenarios based on value functions; the deep deterministic policy gradient is used to process continuous action spaces and scenarios based on policy gradients; the proximal policy optimization uses a probabilistic strategy, which is suitable for scenarios that need to explore multiple possible actions. In one embodiment of the present application, the proximal policy optimization algorithm PPO is used as the deep reinforcement learning algorithm of the present application for explanation.

[0064] Specifically, the policy network and value function network of the intelligent agent model are initialized based on historical data; the policy ratio of the deep reinforcement learning algorithm is calculated; based on the policy ratio, the loss function of the deep reinforcement learning algorithm is defined through the clipping mechanism; the loss function is minimized by the gradient descent method, and the policy network of the intelligent agent model is trained and updated; the value function network of the intelligent agent model is trained and updated by minimizing the mean square error. In each training round, the AI ​​Agent makes decisions in a financial market simulation environment; the AI ​​Agent selects actions (such as buying or selling stocks) based on the current state (such as market indicators); the environment returns a new state and reward based on the action of the AI ​​Agent; the AI ​​Agent updates its strategy based on the reward, with the goal of maximizing the long-term cumulative reward. In the proximal policy optimization algorithm, the performance of the Agent is improved by optimizing the strategy of the policy network, the value function network is used to evaluate the value of the action selected by the policy network, and the advantage function is used to evaluate the quality of each action relative to the average action, and the parameters of the policy network are updated accordingly. During the training process, the training process is also optimized by adjusting hyperparameters such as the learning rate, discount factor, and trust region size.

[0065] Step S30, performing feature analysis on the historical data through an intelligent agent model to obtain indicator data features of the historical data.

[0066] It should be noted that the intelligent agent model refers to AI Agent, an autonomous computing system that can perceive its environment and take actions to achieve specific goals. It is an intelligent entity that can perceive the environment, conduct autonomous understanding, make decisions and perform actions. It is an important concept in the field of artificial intelligence and plays a key role in connecting human and machine intelligence and realizing automated decision-making and task execution.

[0067] In addition, it should be noted that historical data refers to data related to financial markets obtained from reliable data sources. Reliable data sources can be financial data providers (such as Bloomberg, Reuters) or public financial data APIs (such as Alpha Vantage, Yahoo Finance). The data types of historical data include but are not limited to stock prices, trading volumes, market indices, economic indicators (such as GDP growth rate, unemployment rate) and sentiment analysis data (extracted from news and social media). Historical data can be obtained by using Python's requests library or other HTTP client libraries to obtain data from APIs, or by using web crawler technology (such as Scrapy) to crawl data from public web pages.

[0068] In addition, it should be noted that the indicator data characteristics of historical data refer to quantitative indicators or attributes used to analyze financial markets and sentiment analysis data, including but not limited to financial indicator data characteristics such as moving average MA, relative strength index RSI, volatility in technical analysis, and sentiment analysis characteristics such as sentiment score in sentiment analysis. These indicator data characteristics together constitute an important basis for analyzing financial markets and public sentiment.

[0069] More specifically, the moving average is an important indicator in technical analysis, which is used to smooth price data and help identify trend direction and trend changes. The moving average is usually obtained by calculating the average value of asset prices over a period of time (such as 20 days, 50 days, etc.). The relative strength index is a technical analysis tool that measures the speed and magnitude of asset price changes. It evaluates the overbought or oversold state of asset prices by comparing the average magnitude of price increases with the average magnitude of price decreases over a period of time. Volatility measures the degree of change in asset prices over a certain period of time. In financial markets, high volatility usually means large price changes, while low volatility means small price changes. Volatility is an important parameter for measuring market risks and formulating investment strategies. In sentiment analysis data, the sentiment score is a quantitative indicator that measures the sentiment tendency of a text or sentence (such as positive, negative or neutral). Sentiment analysis technology analyzes text through natural language processing (NLP) methods, such as sentiment dictionary matching, machine learning or deep learning models, and outputs the corresponding sentiment score. This score reflects the intensity and direction of the emotions expressed in the text, which is of great significance for understanding public sentiment, market sentiment, and predicting asset price trends.

[0070] Specifically, after obtaining historical data from the data source, the historical data is preprocessed. Missing values ​​are processed by interpolation or direct deletion, outliers are identified and processed by the box plot method, and then normalized to scale the data to a specific range (such as between 0 and 1). Common normalization processes include Min-Max scaling and Z-score standardization. After completing the preprocessing of historical data, the indicator data features are extracted from the historical data through the intelligent agent model.

[0071] More specifically, in one embodiment of the present application, the step of obtaining a moving average includes: selecting a suitable window size (such as 20 days, 50 days, etc.); using a rolling window to calculate the average value of the stock price in the window to obtain a moving average. The step of obtaining a relative strength index includes: determining the window size of the relative strength index (such as 14 days); calculating the gains and losses of price changes; calculating relative strength, that is, the ratio of gains to losses; and finally calculating the RSI value based on RS. The step of obtaining volatility includes: calculating daily logarithmic returns; using a rolling window to calculate the standard deviation of the logarithmic return, and multiplying it by the number of time periods under the square root (such as annualized volatility needs to be multiplied by the square root of 252) to obtain historical volatility.

[0072] More specifically, in one embodiment of the present application, the step of obtaining a sentiment score includes: obtaining a sentiment analysis raw data set such as text comments and social media posts from a data source, and then cleaning the text, such as removing stop words and punctuation marks, performing word form restoration, etc., and converting the cleaned text into a numerical form that can be understood by a machine. Common methods include bag of words, TF-IDF, Word Embedding, etc., using a sentiment dictionary (such as GI, sentiWordNet, etc.) to perform sentiment polarity matching on words in the text and calculate the sentiment score. For example, for each sentence, traverse the words therein and calculate the sentence sentiment score according to the polarity value and weight in the sentiment dictionary.

[0073] Step S40, based on the indicator data characteristics of the historical data, the historical data is used to generate an investment portfolio under a simulated market scenario through a Monte Carlo Markov chain model.

[0074] It should be noted that the Monte Carlo Markov chain model is a statistical simulation method that combines the Monte Carlo method and the Markov chain. This method constructs a Markov chain so that its stationary distribution is consistent with the distribution to be sampled, then performs a random walk based on the Markov chain to generate a sample sequence, and finally uses these stationary distribution samples for approximate numerical calculations.

[0075] Specifically, the rate of return is calculated based on real-time data; a probability model is determined according to the distribution characteristics of the rate of return; and based on the probability model, random sampling is performed through the Monte Carlo Markov chain algorithm to generate an investment portfolio in a simulated market scenario.

[0076] Step S50, evaluating the investment portfolio under the simulated market scenario to obtain a pre-recommended investment portfolio, and providing the pre-recommended investment portfolio to the target user.

[0077] Specifically, when the Markov chain is stably distributed, the investment portfolio on the Markov chain is analyzed to obtain the optimal investment portfolio according to a preset target.

[0078] Through the above scheme, this embodiment combines the intelligent agent model with the Monte Carlo Markov chain model to improve the ability to capture the dynamic characteristics of the market when recommending investment portfolios, improve the ability to capture market complexity and dynamics, and provide reasonable suggestions for users; by training the intelligent agent model through a deep reinforcement learning algorithm based on historical data, the decision-making ability of the intelligent agent model in making investment portfolio recommendations can be improved.

[0079] Based on the above implementation scheme, in a feasible implementation manner, the training of the intelligent agent model by a deep reinforcement learning algorithm based on historical data includes steps S11 to S15:

[0080] Step S11, initializing the policy network and value function network of the intelligent agent model based on historical data.

[0081] It should be noted that the policy network of the intelligent agent model is responsible for selecting actions based on the current state. The input is the state, and the output is the probability distribution of each action (for discrete action space) or directly outputs the action (for continuous action space).

[0082] In addition, it should be noted that the value function network of the intelligent agent model is responsible for evaluating the value of the action, with the input being the state and the output being the estimated value of the state.

[0083] Specifically, the historical data of the financial market is collected from the data source and preprocessed, and then the collected historical data is used to initialize and build a financial market simulation environment, which can simulate the dynamic changes of the market and define the state space, action space and reward function. The state can include the current price, trading volume, technical indicators, etc.; the action can be buy, sell, hold, etc.; the reward function can be designed based on factors such as investment returns and risks. The network structure of the intelligent agent model is designed and initialized, and the network structure of the intelligent agent model includes a policy network and a value function network.

[0084] Step S12, calculating the advantage function and strategy ratio of the deep reinforcement learning algorithm.

[0085] It should be noted that the advantage function is used to evaluate the quality of each action relative to the average action, and update the parameters of the policy network accordingly.

[0086] In addition, it should be noted that the strategy ratio reflects the change in the probability of action selection after the strategy parameters are updated, which is used to measure the difference between the new and old strategies, and to limit the amplitude of the strategy update through the clipping mechanism, thereby improving the stability and efficiency of training.

[0087] Specifically, under the current strategy, a series of trajectories (such as the sequence of state, action, reward, and next state) are sampled by interacting with the environment, and these data are used to calculate the reward and advantage function of each state. The advantage function is used to guide the optimization of the strategy. The estimation formula of the advantage function is:

[0088]

[0089] Among them, δ t =r t +γV(s t+1 )-V(s t ), δ t+l is the TD residual (Temporal Difference Residual) at time step t+l, which represents the difference between the actual return and the expected return; V(s t+1 ) is the value function estimate of the state at time step t+1; V(s t ) is the value function estimate of the state at time step t; λ is the weight parameter of GAE, γ is the discount factor; A t is the advantage function value at time step t, which indicates the expected benefit of taking an action at time step t compared to the average action; r t is the immediate reward obtained at time step t.

[0090] For each state-action pair, calculate the ratio of the probability of selecting the action under the new strategy to the probability of selecting the action under the old strategy, that is, the strategy ratio. The calculation formula of the strategy ratio is:

[0091]

[0092] Among them, r(θ) is the current strategy and the old strategy taking action a at the state of time step t t The probability ratio of θ is the current strategy, which represents the strategy generated by the strategy network under the parameter θ; θold is the old strategy, which represents the strategy generated by the strategy network under the parameter θold, usually the strategy in the previous iteration or training step; t is the action taken at time step t; s t is the state at time step t; θ is the parameter of the current policy network; θold is the parameter of the old policy network.

[0093] Step S13, based on the advantage function and the strategy ratio, a loss function of the deep reinforcement learning algorithm is defined through a clipping mechanism to minimize the loss function.

[0094] It should be noted that the clipping mechanism is used to limit the range of policy changes.

[0095] Specifically, based on the advantage function and the strategy ratio, the loss function of the deep reinforcement learning algorithm is defined through the clipping mechanism. The calculation formula of the loss function is:

[0096]

[0097] Among them, L CLIP (θ) is the clipping loss function, which is used to measure the improvement of the current strategy relative to the old strategy and limit the magnitude of this improvement; is the expectation operator, which indicates the expected calculation of the sampled data at time step t; γ(θ) is the strategy ratio, which indicates the current strategy and the old strategy take action a at the state of time step t. t The probability ratio of is the estimated advantage function value at time step t; clip(γ(θ),1-η,1+η) is a clipping operation, which is used to limit the value of the policy ratio γ(θ) to a range. Specifically, if γ(θ) is greater than 1+η, it is set to 1+η; if γ(θ) is less than 1-η, it is set to 1-η. Otherwise, the original value of γ(θ) is maintained; η is a hyperparameter used to control the size of the clipping interval, usually set to 0.1 or 0.2. This parameter determines the strictness of the policy ratio being clipped. The loss function uses clipping operations to avoid excessive policy updates that lead to unstable training.

[0098] Step S14, minimizing the loss function by gradient descent method, and training and updating the policy network of the intelligent agent model.

[0099] Specifically, the policy loss function is minimized using a gradient descent method (such as the Adam optimizer) to update the parameters of the policy network.

[0100] Step S15, training and updating the value function network of the intelligent agent model by minimizing the mean square error.

[0101] Specifically, the sampled states and the corresponding estimated reward values ​​are used to update the parameters of the value function network by minimizing the mean square error (MSE).

[0102] The update is repeated multiple times, and each iteration uses a new strategy to sample data, calculate the loss, and update the policy network and value function network until the training termination conditions are met (such as reaching the maximum number of training rounds, stable performance, or meeting other preset conditions).

[0103] Through the above scheme, this embodiment can improve the decision-making ability of the intelligent agent model for investment portfolio recommendation by training the intelligent agent model through a deep reinforcement learning algorithm based on historical data.

[0104] Based on the above implementation scheme, in a feasible implementation manner, the method of generating an investment portfolio corresponding to the historical data in a simulated market scenario through a Monte Carlo Markov chain model includes steps S31 to S33:

[0105] Step S31, calculating the rate of return based on historical data.

[0106] It should be noted that the rate of return is the asset rate of return, which represents the ratio between the income obtained by an investment in a certain period and its original investment amount. The calculation of the rate of return helps investors evaluate the performance of their investments and compare the attractiveness of different investment opportunities. The rate of return can be calculated using log returns or simple returns. Log returns are more commonly used in simulations because they are time-additive.

[0107] Specifically, based on the collected asset price data, the asset yield for each time period (such as daily, weekly, and monthly) is calculated. The yield calculation formula is:

[0108]

[0109] Among them, r t is the return on assets; P t is the asset price at time t; P t-1 is the asset price at time t-1.

[0110] Step S32, determining a probability model according to the distribution characteristics of the yield rate.

[0111] Specifically, firstly, the distribution characteristics of the yield are obtained according to the yield. Draw a histogram or density plot of the yield according to the yield to observe its distribution characteristics. Generally speaking, the yield of financial assets presents a non-normal distribution, often showing thick tail and peak characteristics; calculate the mean, variance, skewness and kurtosis of the yield to describe the basic characteristics of the yield distribution; use the autocorrelation function (ACF) and partial autocorrelation function (PACF) to analyze the autocorrelation of the yield and the square of the yield to determine the time dependence of the yield series. Then, according to the distribution characteristics of the yield, select a suitable probability model. Commonly used probability models include normal distribution model (assuming that the yield follows a normal distribution, suitable for assets with a yield distribution close to normal), t distribution model (assuming that the yield follows a t distribution, suitable for asset returns with a thick tail distribution), mixed normal distribution model (using a mixed model of multiple normal distributions to capture the multi-peak characteristics of the yield), GARCH model (if there is volatility clustering in the yield (volatility is continuously high or low in a certain period), the generalized autoregressive conditional heteroskedasticity model (GARCH) can be used to describe the volatility changes).

[0112] Step S33, based on the probability model, random sampling is performed through the Monte Carlo Markov chain algorithm to generate an investment portfolio in a simulated market scenario.

[0113] Specifically, firstly, parameter estimation is performed based on the probability model. In one embodiment of the present application, the parameters of the probability model are estimated using the Bayesian method, and the prior distribution of the parameters is set. The Bayesian estimation combines historical data with prior knowledge and can provide uncertainty information of the parameters. In another embodiment of the present application, the probability model parameters are estimated using the maximum likelihood estimation method. For the normal distribution, the mean and standard deviation of the yield are estimated; for the GARCH model, the various parameters in the equation are estimated. Then random sampling is performed using the Monte Carlo Markov chain algorithm. Under the Bayesian framework, the posterior distribution of the parameters is constructed using historical yield data and the set prior distribution, and a suitable MCMC sampling algorithm, such as Metropolis-Hastings or Hamiltonian Monte Carlo (HMC), is selected for random sampling from the posterior distribution of the parameters to generate an investment portfolio under a simulated market scenario.

[0114] Based on the above implementation scheme, in a feasible implementation manner, the random sampling based on the probability model and the generation of an investment portfolio in a simulated market scenario by the Monte Carlo Markov chain algorithm include steps S331 to S332:

[0115] Step S331, determining the state space based on the investment portfolio in the historical data.

[0116] It should be noted that the state space refers to the set of all possible investment portfolios in the portfolio recommendation problem. Each investment portfolio is a state, which can be represented by an N-dimensional vector, where N is the number of investment targets (such as stocks, bonds, etc.). Each element (or component) in this vector represents the allocation ratio or weight of the corresponding investment target in the portfolio. For example, X = [x_1, x_2, ..., x_N], each x_i represents the allocation ratio of investment target i. Each point in the state space is a specific combination, such as [0.1, 0.2, 0.7] means 10%, 20%, and 70% investment in three targets respectively. When defining the state space, constraints must also be considered, such as the sum of investment ratios must be 1, or the allocation ratio of certain targets must be within a specific range. These constraints will define the boundaries of the state space, thereby limiting the sampling range of the Markov chain.

[0117] Specifically, all possible investment portfolios in the investment portfolio recommendation problem are obtained based on historical data, and the state space is determined.

[0118] Step S332, performing random sampling in the state space to construct a Markov chain, and obtaining an investment portfolio under a simulated market scenario.

[0119] Specifically, the current state in the state space is determined; based on the current state, a candidate state is generated through random perturbations; based on the current state and the candidate state, the acceptance probability of the candidate state is obtained; based on the acceptance probability of the candidate state, the next state of the current state is determined; based on the current state and the next state, a Markov chain is constructed in the state space to obtain an investment portfolio under a simulated market scenario.

[0120] Based on the above implementation scheme, in a feasible implementation manner, the random sampling in the state space to construct a Markov chain and obtain an investment portfolio under a simulated market scenario includes steps A321 to A325:

[0121] Step A321, determining the current state in the state space.

[0122] Specifically, we first randomly select an initial state X_0 from the state space, i.e. the initial investment portfolio. Selecting a suitable initial state can accelerate the convergence of the Markov chain to the optimal solution. Starting from the initial state as the current state, we subsequently generate candidate states through random perturbations, and then determine the next state of the current state based on the candidate states.

[0123] Step A322, based on the current state, generating a candidate state through random perturbation.

[0124] Specifically, starting from the current investment portfolio, that is, from the current state, a new investment portfolio, that is, the candidate state, is generated through random perturbations. This perturbation can be generated by random variables, indicating a small adjustment to the existing investment proportions. For example, based on the current portfolio X = [0.1, 0.2, 0.7], the proportion of x_1 is increased and the proportion of x_2 is reduced to form a new portfolio X' = [0.15, 0.15, 0.7].

[0125] Step A323, obtaining the acceptance probability of the candidate state according to the current state and the candidate state.

[0126] Specifically, the probability of transitioning to a candidate state is determined by an acceptance criterion, which is usually based on the value of an objective function. For portfolio recommendation problems, the objective function is usually a balance function between the return and risk of the portfolio (such as the Sharpe ratio), which is a key indicator used to measure portfolio performance. For an investment portfolio, the objective function usually does not only focus on the performance of a single dimension (such as return or risk), but takes into account multiple factors to achieve a balance. Let the objective function be f(X), then the transition probability P(X→X') is determined by the ratio of the objective function values ​​of the two states, f(X') / f(X). A commonly used strategy is the Metropolis-Hastings algorithm, and the calculation formula for the acceptance probability is:

[0127]

[0128] Among them, α is the acceptance probability; f(X) is the objective function value of the current state X, which is usually used to measure the balance between the return and risk of the investment portfolio; f(X') is the objective function value of the candidate state X'. If f(X')>f(X), that is, the objective function value of the candidate state is greater than the objective function value of the current state, then the candidate state is better, the acceptance probability is 1, and the candidate state is always accepted; if f(X')<f(X), that is, the objective function value of the new state is less than the objective function value of the current state, then the new state is not as good as the current state, the acceptance probability is f(X') / f(X), and the new state is accepted with a certain probability to prevent the algorithm from falling into a local optimum.

[0129] Step A324, determining the next state of the current state based on the acceptance probability of the candidate state.

[0130] Specifically, when the acceptance probability of the candidate state is greater than or equal to the preset acceptance threshold, the candidate state is determined as the next state of the current state; when the acceptance probability of the candidate state is less than the preset acceptance threshold, the current state is retained as the next state. The acceptance threshold is a random number u generated by a uniform distribution, and its value range is [0,1].

[0131] Step A325, constructing a Markov chain in the state space based on the current state and the next state to obtain an investment portfolio under a simulated market scenario.

[0132] Specifically, iterative sampling is performed starting from the initial state, and the next state is determined based on the current state. A Markov chain is constructed in the state space during the continuous sampling and iteration process. Each time a state value is generated, a set of investment portfolios is obtained.

[0133] Based on the above implementation scheme, in a feasible implementation manner, determining the next state of the current state based on the acceptance probability of the candidate state includes steps A241 to A242:

[0134] Step A241: When the acceptance probability of the candidate state is greater than or equal to a preset acceptance threshold, the candidate state is determined as the next state of the current state.

[0135] It should be noted that the preset acceptance threshold is a random number generated in a uniform distribution, and its value range is 0 to 1.

[0136] Specifically, when the acceptance probability of the candidate state is greater than or equal to a preset acceptance threshold, the candidate state is determined as the next state of the current state. In one embodiment of the present application, let the acceptance threshold be u, if u≤α, then the candidate state is accepted, and the candidate state is determined as the next state of the current state, and let X_(t+1)=X', where X_(t+1) is the next state and X' is the candidate state.

[0137] Step A242: When the acceptance probability of the candidate state is less than a preset acceptance threshold, the current state is retained as the next state.

[0138] Specifically, when the acceptance probability of the candidate state is less than a preset acceptance threshold, the current state is retained as the next state. In one embodiment of the present application, the acceptance threshold is set to u. If u>α, the candidate sample is rejected and the current state is retained as the next state, that is, X_(t+1)=X_t, where X_(t+1) is the next state and X_t is the current state.

[0139] Based on the above implementation scheme, in a feasible implementation mode, the investment portfolio under the simulated market scenario is evaluated to obtain a pre-recommended investment portfolio, and the pre-recommended investment portfolio is provided to the target user. Step S41:

[0140] Step S41, when the Markov chain is stably distributed, analyzing the investment portfolio on the Markov chain, obtaining a pre-recommended investment portfolio according to a preset target, and recommending the pre-recommended investment portfolio to the target user.

[0141] Specifically, as sampling continues, the Markov chain will gradually stabilize and reach a "stationary distribution". At this stage, the sampling results can be regarded as representing the probability distribution of the entire state space. The sampling results are statistically analyzed to identify high-probability areas. These areas usually contain optimal or near-optimal investment portfolios, and the optimal investment portfolio is provided to the user. There are many ways to determine the optimal combination. In one embodiment of the present application, a single objective is directly selected. If the goal is to find a single optimal combination, the combination with the highest objective function value (such as the Sharpe ratio) in the sampling results can be directly selected; in another embodiment of the present application, a comprehensive selection is made with multiple objectives, and multiple objectives include but are not limited to returns, risks, liquidity, etc., and a set of optimal combinations that meet all objective requirements is found through multi-objective optimization technology; in another embodiment of the present application, a probability weighted selection is made, and each combination in the sampling results is weighted according to the probability of occurrence to obtain a comprehensive optimal combination.

[0142] More specifically, the pre-recommended investment portfolio is executed in real time, and market data is acquired in real time. According to the acquired real-time data and market changes, the pre-recommended investment portfolio is dynamically adjusted through the intelligent agent model and the Monte Carlo Markov chain model to improve the real-time performance of investment portfolio recommendations.

[0143] For example, in order to help understand the implementation process of the investment portfolio recommendation method based on the intelligent agent model in this embodiment, this embodiment provides a specific scenario-based description of this solution:

[0144] When the target user is an ordinary investor who needs to make investment advice portfolio recommendations for personal financial management, a recommendation request instruction is sent on the interactive interface of the investment portfolio recommendation device based on the intelligent agent model to request the device to provide investment advice. After receiving the recommendation request instruction from the target user, the device analyzes the user's personal financial situation, investment goals, risk preferences and other information; then a large amount of historical data is collected from channels such as financial markets, investment institutions, and public databases, and the intelligent agent model AI Agent is trained through a deep reinforcement learning algorithm based on the historical data to learn the laws and patterns of the market. During the training process, the investment strategy is optimized by continuously adjusting the model parameters to achieve long-term stable returns; then the intelligent agent model is used to perform feature analysis on the historical data to extract the key indicator data features that affect investment decisions; then the Monte Carlo Markov chain model is constructed based on the indicator data features of the historical data. The model can simulate changes in market scenarios and generate investment portfolios under different scenarios. The Monte Carlo Markov chain model is used to generate multiple simulated market scenarios, each of which contains different market conditions and investment opportunities; then the indicators for evaluating the investment portfolio, such as rate of return, risk rate, Sharpe ratio, etc., are determined, and the investment portfolio under the simulated market scenario is evaluated using evaluation methods such as backtesting and cross-validation. By comparing the evaluation indicators of different investment portfolios, the investment portfolio with the best performance is screened out as the pre-recommended investment portfolio; finally, the pre-recommended investment portfolio is personalized according to the target user's financial status, investment goals, risk tolerance and other information, and the pre-recommended investment portfolio is displayed to the target user through the user interaction interface. At the same time, detailed investment analysis reports and investment recommendations are provided to help users better understand the composition and performance of the investment portfolio.

[0145] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the investment portfolio recommendation method based on the intelligent agent model of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0146] This application also provides an investment portfolio recommendation device based on an intelligent agent model, please refer to Figure 2 , the investment portfolio recommendation device based on the intelligent agent model includes:

[0147] An instruction receiving module 201 is used to receive a recommendation request instruction from a target user;

[0148] A model training module 202 is used to train the intelligent agent model through a deep reinforcement learning algorithm based on historical data;

[0149] The feature extraction module 203 is used to perform feature analysis on the historical data through the intelligent agent model to obtain the index data features of the historical data;

[0150] A portfolio generation module 204 is used to generate an investment portfolio under a simulated market scenario by using a Monte Carlo Markov chain model based on the indicator data characteristics of the historical data;

[0151] The portfolio recommendation module 205 is used to evaluate the investment portfolio under the simulated market scenario, obtain a pre-recommended investment portfolio, and provide the pre-recommended investment portfolio to the target user.

[0152] The investment portfolio recommendation device based on the intelligent agent model provided by the present application adopts the investment portfolio recommendation method based on the intelligent agent model in the above embodiment, which can solve the technical problem that it is difficult to capture the complexity and dynamics of the market when recommending investment portfolios. Compared with the prior art, the beneficial effects of the investment portfolio recommendation device based on the intelligent agent model provided by the present application are the same as the beneficial effects of the investment portfolio recommendation method based on the intelligent agent model provided by the above embodiment, and the other technical features of the investment portfolio recommendation device based on the intelligent agent model are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0153] The present application provides an investment portfolio recommendation device based on an intelligent agent model, and the investment portfolio recommendation device based on an intelligent agent model includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the investment portfolio recommendation method based on the intelligent agent model in the above-mentioned embodiment 1.

[0154] Reference below Figure 3 , which shows a schematic diagram of the structure of an investment portfolio recommendation device based on an intelligent agent model suitable for implementing the embodiment of the present application. The investment portfolio recommendation device based on an intelligent agent model in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The illustrated investment portfolio recommendation device based on the intelligent agent model is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0155] like Figure 3As shown, the investment portfolio recommendation device based on the intelligent agent model may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM) 1004. Various programs and data required for the operation of the investment portfolio recommendation device based on the intelligent agent model are also stored in the RAM 1004. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems may be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 may allow the portfolio recommendation device based on the intelligent agent model to communicate wirelessly or by wire with other devices to exchange data. Although the figure shows a portfolio recommendation device based on the intelligent agent model with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0156] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0157] The investment portfolio recommendation device based on the intelligent agent model provided by the present application adopts the investment portfolio recommendation method based on the intelligent agent model in the above embodiment, which can solve the technical problem that it is difficult to capture the complexity and dynamics of the market when recommending investment portfolios. Compared with the prior art, the beneficial effects of the investment portfolio recommendation device based on the intelligent agent model provided by the present application are the same as the beneficial effects of the investment portfolio recommendation method based on the intelligent agent model provided by the above embodiment, and the other technical features of the investment portfolio recommendation device based on the intelligent agent model are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0158] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0159] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0160] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the investment portfolio recommendation method based on the intelligent agent model in the above-mentioned embodiment.

[0161] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0162] The computer-readable storage medium may be included in the investment portfolio recommendation device based on the intelligent agent model; or may exist independently without being assembled into the investment portfolio recommendation device based on the intelligent agent model.

[0163] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the investment portfolio recommendation device based on the intelligent agent model, the investment portfolio recommendation device based on the intelligent agent model: receives a recommendation request instruction from a target user; trains the intelligent agent model through a deep reinforcement learning algorithm based on historical data; performs feature analysis on the historical data through the intelligent agent model to obtain indicator data features of the historical data; based on the indicator data features of the historical data, generates an investment portfolio under a simulated market scenario through a Monte Carlo Markov chain model for the historical data; evaluates the investment portfolio under the simulated market scenario to obtain a pre-recommended investment portfolio, and provides the pre-recommended investment portfolio to the target user.

[0164] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0165] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0166] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0167] The readable storage medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned investment portfolio recommendation method based on the intelligent agent model, and can solve the technical problem that it is difficult to capture the complexity and dynamics of the market when recommending investment portfolios. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the investment portfolio recommendation method based on the intelligent agent model provided in the above-mentioned embodiment, and will not be repeated here.

[0168] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned investment portfolio recommendation method based on the intelligent agent model when executed by a processor.

[0169] The computer program product provided by the present application can solve the technical problem that it is difficult to capture the complexity and dynamics of the market when recommending investment portfolios. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the investment portfolio recommendation method based on the intelligent agent model provided by the above embodiment, and will not be repeated here.

[0170] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A portfolio recommendation method based on an intelligent agent model, characterized in that: The method comprises: Receive a recommendation request instruction from a target user; Train the intelligent agent model through deep reinforcement learning algorithm based on historical data; Perform feature analysis on historical data through intelligent agent model to obtain the indicator data features of historical data; Based on the indicator data characteristics of historical data, the historical data is used to generate an investment portfolio under a simulated market scenario through the Monte Carlo Markov chain model; The investment portfolio under the simulated market scenario is evaluated to obtain a pre-recommended investment portfolio, and the pre-recommended investment portfolio is provided to the target user.

2. The method according to claim 1, characterized in that The step of training the intelligent agent model through a deep reinforcement learning algorithm based on historical data includes: Initialize the policy network and value function network of the intelligent agent model based on historical data; Calculating the advantage function and policy ratio of the deep reinforcement learning algorithm; Based on the advantage function and the policy ratio, a clipping mechanism is used to define the loss function that minimizes the deep reinforcement learning algorithm. Minimizing the loss function by gradient descent method to train and update the policy network of the intelligent agent model; The value function network of the intelligent agent model is trained and updated by minimizing the mean square error.

3. The method according to claim 1, characterized in that The step of generating an investment portfolio corresponding to the historical data in a simulated market scenario through the Monte Carlo Markov chain model includes: Calculate rate of return based on historical data; Determine a probability model according to the distribution characteristics of the yield; Based on the probability model, random sampling is performed through the Monte Carlo Markov chain algorithm to generate investment portfolios in simulated market scenarios.

4. The method according to claim 3, characterized in that The step of randomly sampling and generating an investment portfolio in a simulated market scenario based on the probability model by using a Monte Carlo Markov chain algorithm includes: Determine the state space based on the portfolio in historical data; Random sampling is performed in the state space to construct a Markov chain and obtain an investment portfolio under a simulated market scenario.

5. The method according to claim 4, characterized in that The step of constructing a Markov chain by random sampling in the state space to obtain an investment portfolio under a simulated market scenario comprises: determining a current state in the state space; Based on the current state, generating a candidate state by random perturbation; Obtaining an acceptance probability of the candidate state according to the current state and the candidate state; Determining a next state of the current state based on the acceptance probability of the candidate state; A Markov chain is constructed in the state space based on the current state and the next state to obtain an investment portfolio under a simulated market scenario.

6. The method according to claim 5, characterized in that The step of determining the next state of the current state based on the acceptance probability of the candidate state comprises: When the acceptance probability of the candidate state is greater than or equal to a preset acceptance threshold, the candidate state is determined as the next state of the current state; When the acceptance probability of the candidate state is less than a preset acceptance threshold, the current state is retained as the next state.

7. The method according to claim 1, characterized in that The step of evaluating the investment portfolio under the simulated market scenario to obtain a pre-recommended investment portfolio and providing the pre-recommended investment portfolio to the target user comprises: When the Markov chain is stably distributed, the investment portfolio on the Markov chain is analyzed, a pre-recommended investment portfolio is obtained according to a preset target, and the pre-recommended investment portfolio is recommended to the target user.

8. An investment portfolio recommendation device based on an intelligent agent model, characterized in that: The device comprises: An instruction receiving module, used for receiving a recommendation request instruction from a target user; Model training module, used to train intelligent agent models through deep reinforcement learning algorithms based on historical data; The feature extraction module is used to perform feature analysis on historical data through an intelligent agent model to obtain the indicator data features of the historical data; The portfolio generation module is used to generate investment portfolios in simulated market scenarios based on the indicator data characteristics of historical data and the Monte Carlo Markov chain model for real-time data; The portfolio recommendation module is used to evaluate the investment portfolio under the simulated market scenario, obtain a pre-recommended investment portfolio, and provide the pre-recommended investment portfolio to the target user.

9. An investment portfolio recommendation device based on an intelligent agent model, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the investment portfolio recommendation method based on the intelligent agent model as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the investment portfolio recommendation method based on the intelligent agent model as described in any one of claims 1 to 7 are implemented.

Images