Product Recommendation Method, Device, Computer Equipment and Storage Medium

Abstract

The present application relates to a product recommendation method, apparatus, computer device, storage medium, and computer program product, and relates to the field of artificial intelligence technology. The method includes: determining an initial product portfolio and target information of the initial product portfolio under multiple preset product goals according to the product to be recommended and the total resources of a first target user; screening out a candidate product portfolio that meets the preset target information from the initial product portfolio according to the target information of the initial product portfolio under multiple preset product goals; determining the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolio; determining the recommendation degree of the candidate product portfolio according to the similarity and the preference degree of the second target user for the candidate product portfolio; and determining a target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree. By adopting this method, the recommendation accuracy of the product portfolio can be improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular, to a product recommendation method, device, computer device, storage medium, and computer program product. Background Art

[0002] With the development of Internet technology, various types of products have emerged on the Internet, and users can select corresponding products according to their own needs.

[0003] However, when selecting a product combination, it is mainly through the recommendation of salespersons or other users; however, this recommendation method is subjective and considers fewer factors, resulting in a low accuracy rate of product combination recommendation. Summary of the Invention

[0004] Based on this, in view of the above technical problems, it is necessary to provide a product recommendation method, device, computer device, computer-readable storage medium, and computer program product that can improve the accuracy rate of product combination recommendation.

[0005] In a first aspect, this application provides a product recommendation method. The method includes:

[0006] Determine an initial product combination and the target information of the initial product combination under multiple preset product goals according to the total resources of the product to be recommended and the first target user; the initial product combination represents a selection combination of the product to be recommended under the constraint of the total resources, and each product in the initial product combination is associated with a corresponding predicted resource transfer ratio;

[0007] According to the target information of the initial product combination under multiple preset product goals, screen out candidate product combinations that meet the preset target information from the initial product combination;

[0008] Determine the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product combination; the second target user is a user other than the first target user among the preset users;

[0009] Determine the recommendation degree of the candidate product combination according to the similarity and the preference degree of the second target user for the candidate product combination;

[0010] Determine the target product combination recommended to the first target user from the candidate product combinations according to the recommendation degree.

[0011] In one of the embodiments, the determining an initial product combination and the target information of the initial product combination under multiple preset product goals according to the total resources of the product to be recommended and the first target user includes:

[0012] Obtain the product information of the product to be recommended;

[0013] Input the product information of the product to be recommended and the total resources of the first target user into the product portfolio prediction models under multiple preset product goals, to obtain an initial product portfolio, the predicted resource transfer ratio of each product in the initial product portfolio, and the target information of the initial product portfolio under the multiple preset product goals.

[0014] In one embodiment, the screening of the candidate product portfolio that meets the preset target information from the initial product portfolio according to the target information of the initial product portfolio under multiple preset product goals includes:

[0015] Take the target information of the initial product portfolio under multiple preset product goals as a particle to obtain multiple particles, and based on the multiple particles, obtain a current particle swarm and an external particle swarm;

[0016] According to the position information of the particles in the current particle swarm in the target space and the position information of the particles in the external particle swarm in the target space, confirm the distance between the particles in the current particle swarm and the particles in the external particle swarm;

[0017] Select the particles in the external particle swarm whose distance is greater than the average distance to obtain a candidate particle swarm;

[0018] According to the crowding distance of the particles in the candidate particle swarm, select the target particles in the candidate particle swarm whose crowding distance meets the preset crowding distance;

[0019] Confirm the initial product portfolio corresponding to the target particles as the candidate product portfolio that meets the preset target information.

[0020] In one embodiment, the obtaining of the current particle swarm and the external particle swarm based on the multiple particles includes:

[0021] Confirm the distance between each particle in the multiple particles and a reference point, and select the particle with the smallest distance from the multiple particles;

[0022] Select the particles in the preference region from the multiple particles; the preference region is a region set according to the reference point, the particle with the smallest distance, and a preference region expansion coefficient;

[0023] According to the Chebyshev dominance method, select the non-dominated particle swarm from the particles in the preference region as the external particle swarm, and use the particle swarm except the non-dominated particle swarm among the particles in the preference region as the current particle swarm.

[0024] In one embodiment, before determining the similarity between the first target user and the second target user according to the preference degrees of the candidate product portfolio for the first target user and the second target user, the method further includes:

[0025] Obtaining the historical resource transfer ratio of the preset user for the product to be recommended;

[0026] Constructing a product frequency matrix according to the historical resource transfer ratio of the preset user for the product to be recommended;

[0027] Determining the historical resource transfer ratio of each product in the candidate product portfolio for the first target user and the historical resource transfer ratio of each product in the candidate product portfolio for the second target user according to the product frequency matrix;

[0028] Performing a fusion process on the historical resource transfer ratios of each product in the candidate product portfolio for the first target user to obtain the preference degree of the first target user for the candidate product portfolio, and performing a fusion process on the historical resource transfer ratios of each product in the candidate product portfolio for the second target user to obtain the preference degree of the second target user for the candidate product portfolio.

[0029] In one embodiment, the step of determining the similarity between the first target user and the second target user according to the preference degrees of the candidate product portfolio for the first target user and the second target user includes:

[0030] Inputting the preference degrees of the first target user and the second target user for the candidate product portfolio into a Pearson similarity confirmation model to obtain the Pearson similarity between the first target user and the second target user;

[0031] Confirming the Pearson similarity between the first target user and the second target user as the similarity between the first target user and the second target user.

[0032] In one embodiment, the step of determining the recommendation degree of the candidate product portfolio according to the similarity and the preference degree of the second target user for the candidate product portfolio includes:

[0033] For each second target user, respectively performing a fusion process on the similarity between the first target user and the second target user and the preference degree of the second target user for the candidate product portfolio to obtain a plurality of fusion process results;

[0034] Performing a re-fusion process on the plurality of fusion process results to obtain the recommendation degree of the candidate product portfolio.

[0035] In one embodiment, determining the target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree includes:

[0036] Filtering out the target product portfolios from the candidate product portfolios whose recommendation degree meets the preset recommendation degree;

[0037] Recommending the target product portfolio to the first target user.

[0038] In a second aspect, the present application further provides a product recommendation device. The device includes:

[0039] A product portfolio module, configured to determine an initial product portfolio and the target information of the initial product portfolio under multiple preset product goals according to the products to be recommended and the total resources of the first target user; the initial product portfolio represents a selection combination of the products to be recommended under the constraint of the total resources, and each product in the initial product portfolio is associated with a corresponding predicted resource transfer ratio;

[0040] A combination screening module, configured to screen out candidate product portfolios that meet the preset target information from the initial product portfolio according to the target information of the initial product portfolio under multiple preset product goals;

[0041] A similarity calculation module, configured to determine the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolios; the second target user is a user other than the first target user among the preset users;

[0042] A recommendation degree calculation module, configured to determine the recommendation degree of the candidate product portfolio according to the similarity and the preference degree of the second target user for the candidate product portfolio;

[0043] A combination determination module, configured to determine the target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree.

[0044] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0045] Determine an initial product portfolio and the target information of the initial product portfolio under multiple preset product goals according to the total resources of the product to be recommended and the first target user; the initial product portfolio represents a selection combination of the products to be recommended under the constraint of the total resources, and each product in the initial product portfolio is associated with a corresponding predicted resource transfer ratio;

[0046] According to the target information of the initial product portfolio under multiple preset product goals, screen out candidate product portfolios that meet the preset target information from the initial product portfolio;

[0047] Determine the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolios; the second target user is a user other than the first target user among the preset users;

[0048] Determine the recommendation degree of the candidate product portfolio according to the similarity and the preference degree of the second target user for the candidate product portfolio;

[0049] Determine the target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree.

[0050] In a fourth aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, the following steps are implemented:

[0051] Determine an initial product portfolio and the target information of the initial product portfolio under multiple preset product goals according to the total resources of the product to be recommended and the first target user; the initial product portfolio represents a selection combination of the products to be recommended under the constraint of the total resources, and each product in the initial product portfolio is associated with a corresponding predicted resource transfer ratio;

[0052] According to the target information of the initial product portfolio under multiple preset product goals, screen out candidate product portfolios that meet the preset target information from the initial product portfolio;

[0053] Determine the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolios; the second target user is a user other than the first target user among the preset users;

[0054] Determine the recommendation degree of the candidate product portfolio according to the similarity and the preference degree of the second target user for the candidate product portfolio;

[0055] Determine a target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree.

[0056] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0057] Determine an initial product portfolio and the target information of the initial product portfolio under multiple preset product goals according to the products to be recommended and the total resources of the first target user; the initial product portfolio represents a selection combination of the products to be recommended under the constraint of the total resources, and each product in the initial product portfolio is associated with a corresponding predicted resource transfer ratio;

[0058] Filter out candidate product portfolios that meet the preset target information from the initial product portfolio according to the target information of the initial product portfolio under multiple preset product goals;

[0059] Determine the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolios; the second target user is a user other than the first target user among the preset users;

[0060] Determine the recommendation degree of the candidate product portfolio according to the similarity and the preference degree of the second target user for the candidate product portfolio;

[0061] Determine a target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree.

[0062] The above product recommendation method, device, computer device, storage medium and computer program product determine an initial product portfolio and the target information of the initial product portfolio under multiple preset product goals according to the product to be recommended and the total resources of the first target user; then, according to the target information of the initial product portfolio under multiple preset product goals, screen out candidate product portfolios that meet the preset target information from the initial product portfolio; then, determine the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolios; finally, determine the recommendation degree of the candidate product portfolios according to the similarity and the preference degree of the second target user for the candidate product portfolios, and determine the target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree. In this way, determining the initial product portfolio and the target information of the initial product portfolio under multiple preset product goals according to the product to be recommended and the total resources of the first target user can increase the diversity of the product portfolio and provide a more comprehensive product portfolio plan for customers; then, screening out candidate product portfolios that meet the preset target information from the initial product portfolio according to the target information of the initial product portfolio under multiple preset product goals is conducive to screening out product portfolios that match the user's needs as much as possible; then, determining the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolios, so that while recommending a suitable product portfolio plan for the user, the factors of the customer's own preferences are also fully considered; finally, determining the recommendation degree of the candidate product portfolios according to the similarity and the preference degree of the second target user for the candidate product portfolios, and determining the target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree; recommending the product portfolio to the first target user according to the objective index of the recommendation degree avoids the influence of subjective factors, and comprehensively considers multiple factors, while taking into account the diversity and practicability of the product portfolio plan, thereby improving the recommendation accuracy of the product portfolio. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is a schematic flowchart of the product recommendation method in an embodiment;

[0064] Figure 2 is a schematic flowchart of the steps of screening out candidate product portfolios that meet the preset target information in an embodiment;

[0065] Figure 3 is a schematic diagram of a preference area in an embodiment;

[0066] Figure 4 is a schematic flowchart of the product recommendation method in another embodiment;

[0067] Figure 5 is a schematic flowchart of the product recommendation method in yet another embodiment;

[0068] Figure 6 is a structural block diagram of a product recommendation device in an embodiment;

[0069] Figure 7 is an internal structure diagram of a computer device in an embodiment. Specific Embodiments

[0070] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0071] In one embodiment, as Figure 1 shown, a product recommendation method is provided. In this embodiment, it is exemplified that the method is applied to a terminal. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0072] Step S101: Determine an initial product portfolio and the target information of the initial product portfolio under multiple preset product objectives according to the product to be recommended and the total resources of the first target user; the initial product portfolio represents a selection combination of the products to be recommended under the constraint of the total resources, and each product in the initial product portfolio is associated with a corresponding predicted resource transfer ratio.

[0073] Among them, the products involved in the present application refer to financial products such as funds and stocks.

[0074] Among them, the first target user refers to the authorized user who currently needs to be recommended a product portfolio.

[0075] Among them, the initial product portfolio refers to a combination composed of multiple products selected from the products to be recommended under the constraint of the total resources; for example, the products to be recommended include I1, I2, I3, I4, I5, I6, I7, and the initial product portfolios are (I1, I2, I3), (I2, I4, I5, I7), (I1, I3, I4, I6, I7), etc.

[0076] Among them, the products to be recommended refer to the currently selectable products; the total resources refer to the existing total funds; the multiple preset product objectives refer to indicators such as the risk of product fluctuations, the total return, the risk coefficient, and the balance between return and risk; the target information refers to the variance of the initial product portfolio, the total net return of the initial product portfolio over a period of time, and the balance between return and risk of the initial product portfolio; the predicted resource transfer ratio refers to the investment ratio of the user in the product.

[0077] Specifically, the terminal receives a product portfolio recommendation request for a first target user. According to this product portfolio recommendation request, it obtains the products to be recommended, the product information of the products to be recommended, and the total resources of the first target user, and inputs the product information of the products to be recommended and the total resources of the first target user into the product portfolio prediction models under multiple preset product goals for fusion calculation, obtaining an initial product portfolio, the predicted resource transfer ratio of each product in the initial product portfolio, and the target information of the initial product portfolio under multiple preset product goals.

[0078] For example, the terminal receives a financial plan recommendation request for a first target user, and obtains the products to be recommended, namely products A, B, and C. The product information of the products to be recommended includes transaction fees, average expected return rate, minimum investment ratio, maximum investment ratio, etc., and the total resources of the first target user is T. Then it inputs the product information of the products to be recommended and the total resources of the first target user into the multi-objective financial investment model for fusion calculation, and obtains one of the financial recommendation plans: the investment ratio of product A is 30%, the investment ratio of product B is 20%, the investment ratio of product C is 50%, and the total investment amount is T.

[0079] Step S102: According to the target information of the initial product portfolio under multiple preset product goals, screen out candidate product portfolios that meet the preset target information from the initial product portfolio.

[0080] Among them, the preset target information refers to the expected variance of the initial product portfolio set in advance, the total expected net income of the initial product portfolio within a period of time, and the balance between the expected return and risk of the initial product portfolio; the candidate product portfolio refers to the product portfolio that meets the preset target information screened out from all the initial product portfolios.

[0081] Specifically, the terminal takes the target information of the initial product portfolio under multiple preset product goals as a particle, obtaining multiple particles; through the multi-objective particle swarm model, analyzes the position information of multiple particles in the target space, obtaining the position information analysis result; according to the position information analysis result, screens out a candidate particle swarm from multiple particles; through the multi-objective particle swarm model, analyzes the crowding distance of the particles in the candidate particle swarm, obtaining the crowding distance analysis result; according to the crowding distance analysis result, screens out the target particles from the candidate particle swarm; and determines the initial product portfolio corresponding to the target particles as the candidate product portfolio that meets the preset target information.

[0082] Step S103: Determine the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolio; the second target user is a user other than the first target user among the preset users.

[0083] Among them, the preference degree refers to the degree of interest of a user in a certain candidate product combination; the larger the preference degree, the more interested the user is in the candidate product combination.

[0084] Among them, the similarity is used to measure the similarity degree between the product selection preferences of the first target user and the second target user. The preset user refers to all users in the product recommendation system.

[0085] Specifically, the terminal calculates the preference degrees of the first target user and the second target user for the candidate product combination according to the historical product selection record information of the first target user and the second target user; inputs the preference degrees of the first target user and the second target user for the candidate product combination into the Pearson similarity confirmation model to obtain the Pearson similarity between the first target user and the second target user; and confirms the Pearson similarity between the first target user and the second target user as the similarity between the first target user and the second target user.

[0086] For example, the terminal inputs the preference degrees of the first target user and the second target user for the candidate product combination into the Pearson similarity confirmation model, and the obtained Pearson similarity between the first target user and the second target user is 0.5; and confirms the Pearson similarity 0.5 as the similarity between the first target user and the second target user.

[0087] Step S104, determine the recommendation degree of the candidate product combination according to the similarity and the preference degree of the second target user for the candidate product combination.

[0088] Among them, the recommendation degree refers to the size of the recommended value of the candidate product combination. The larger the recommended value, the more inclined the user is to purchase the candidate product combination.

[0089] Specifically, the terminal inputs the similarity and the preference degree of the second target user for the candidate product combination into the recommendation degree prediction model, and the recommendation degree prediction model performs a fusion process on the similarity and the preference degree of the second target user for the candidate product combination to obtain the recommendation degree of the candidate product combination.

[0090] Step S105, determine the target product combination recommended to the first target user from the candidate product combinations according to the recommendation degree.

[0091] Among them, the target product combination refers to the finally obtained product combination that the user is interested in and has a good investment prospect, such as a financial product portfolio recommendation plan.

[0092] Specifically, the terminal sorts the candidate product combinations in descending order of recommendation degree to obtain the sorted candidate product combinations; from the sorted candidate product combinations, the first N1 candidate product combinations are screened out as the target product combinations, and the target product combinations are recommended to the first target user; where N1 is a positive integer.

[0093] For example, the terminal screens out the target product combination from the candidate product combinations whose recommendation degree meets the preset recommendation degree of 90: the investment ratio of product A is 25%, the investment ratio of product B is 40%, the investment ratio of product C is 35%, and the total investment amount is T; the target product combination is recommended to the first target user.

[0094] In the above product recommendation method, according to the products to be recommended and the total resources of the first target user, the initial product combination and the target information of the initial product combination under multiple preset product goals are determined, which can increase the diversity of the product combination and provide a more comprehensive product combination plan for customers; then, according to the target information of the initial product combination under multiple preset product goals, the candidate product combinations that meet the preset target information are screened out from the initial product combination, which is conducive to screening out the product combinations that match the user's needs as much as possible; then, according to the preference degrees of the first target user and the second target user for the candidate product combinations, the similarity between the first target user and the second target user is determined, so that while recommending a suitable product combination plan for the user, the factors of the customer's own preferences are also fully considered; finally, according to the similarity and the preference degree of the second target user for the candidate product combinations, the recommendation degree of the candidate product combinations is determined, and according to the recommendation degree, the target product combinations recommended to the first target user are determined from the candidate product combinations; the product combination is recommended to the first target user according to this objective indicator of the recommendation degree, avoiding the influence of subjective factors, considering multiple factors comprehensively, taking into account both the diversity and practicality of the product combination plan, and improving the recommendation accuracy of the product combination.

[0095] In one embodiment, in the above step S101, according to the products to be recommended and the total resources of the first target user, determining the initial product combination and the target information of the initial product combination under multiple preset product goals specifically includes the following content: obtaining the product information of the products to be recommended; inputting the product information of the products to be recommended and the total resources of the first target user into the product combination prediction model under multiple preset product goals to obtain the initial product combination, the predicted resource transfer ratio of each product in the initial product combination, and the target information of the initial product combination under multiple preset product goals.

[0096] Among them, the product information of the products to be recommended includes: the total number of optional assets N, the number of assets invested K (K <= N), and the transaction rate P of asset i i , and the purchase amount of asset i does not exceed the given value bi , the average expected return rate r of asset i i , the minimum investment ratio n of asset i i , the maximum investment ratio m of asset i i , the actual investment ratio w of asset i i , the actual investment ratio w of asset j j , z i indicates whether to invest in asset i (z i ∈ {0, 1}), the transaction cost c(wi) of asset i, the net income R(wi) of asset i, the covariance σ between asset i and j ij ;

[0097]

[0098] R(w i ) = r i × M × w i - c(w i ) × (1 + r i )

[0099] The product portfolio prediction model (composed of Q, R, and G functions) under multiple preset product goals. The Q function represents the variance of the investment portfolio (w1,..., w N ), representing the risk of future fluctuations if deviating from the mean, and the smaller the better, represents the total pure income of the investment portfolio (w1,..., w N ) over a period of time, and the larger the better. Here, for the convenience of solution, it is changed to solve Min R. The G function comprehensively considers the return and risk of the investment portfolio, and the smaller the better.

[0100]

[0101]

[0102]

[0103] The constraint condition s.t is:

[0104] Specifically, the terminal receives a product portfolio instruction, obtains the product information of the product to be recommended according to the product portfolio instruction; according to the preset conditions of the product information of the product to be recommended, inputs the product information and the total resources of the first target user into the product portfolio prediction model under multiple preset product goals at the same time, and obtains the initial product portfolio, the predicted resource transfer ratio of each product in the initial product portfolio, and the target information of the initial product portfolio under multiple preset product goals.

[0105] In this embodiment, product information of the product to be recommended is obtained; and the product information of the product to be recommended and the total resources of the first target user are input into product portfolio prediction models under multiple preset product goals, so as to obtain an initial product portfolio, the predicted resource transfer ratio of each product in the initial product portfolio, and the target information of the initial product portfolio under multiple preset product goals; thus, the predicted investment ratio of each product in the initial product portfolio and the target information of the initial product portfolio under multiple constraint conditions can be accurately obtained.

[0106] In one embodiment, as Figure 2 shown, in step S102 above, according to the target information of the initial product portfolio under multiple preset product goals, candidate product portfolios that meet the preset target information are screened out from the initial product portfolio, which specifically includes the following steps:

[0107] Step S201: Take the target information of the initial product portfolio under multiple preset product goals as a particle, obtain multiple particles, and obtain a current particle swarm and an external particle swarm according to the multiple particles.

[0108] Step S202: Confirm the distance between the particles in the current particle swarm and the particles in the external particle swarm according to the position information of the particles in the current particle swarm in the target space and the position information of the particles in the external particle swarm in the target space.

[0109] Step S203: Screen out the particles in the external particle swarm with a distance greater than the average distance to obtain a candidate particle swarm.

[0110] Step S204: Screen out the target particles in the candidate particle swarm whose crowding distance meets the preset crowding distance according to the crowding distance of the particles in the candidate particle swarm.

[0111] Step S205: Confirm the initial product portfolio corresponding to the target particle as the candidate product portfolio that meets the preset target information.

[0112] Among them, each particle represents a solution (w1,..., w N ), w i is the actual investment ratio of the i-th asset, and the performance of these two particles (i.e., two solutions) in the Q, R, G objective functions will be compared, and the particle with excellent performance will be selected for the next iteration, and the solution set of the algorithm will converge towards the particle with relatively excellent performance; the target space refers to the preference region, as Figure 3 shown, z ref is the reference point, z * is the position of the *particle on f1, f2, where its achievement scalar function is the smallest and the distance from the reference point is the closest, and the concept of the preference region is formed by these two points and the expansion coefficient.

[0113] Specifically, the terminal sets a reference point according to the expected values of each preset product goal; uses the target information of the initial product portfolio under multiple preset product goals as a particle to obtain multiple particles; determines a preference region according to the position of the reference point and the position of the particle closest to the reference point among the multiple particles; determines the current particle swarm and the external particle swarm from the multiple particles according to the positions of the multiple particles and the preference region; determines the distance between the particles in the current particle swarm and the particles in the external particle swarm according to the distance between the particles and the reference point, that is, the position information of the particles in the current particle swarm in the target space and the position information of the particles in the external particle swarm in the target space; calculates the average value of the distances between one of the particles in the current particle swarm and each particle in the external particle swarm, uses the average value of the distances as the average distance, and filters out the particles with a distance greater than the average distance from the external particle swarm to obtain a candidate particle swarm; sorts the crowding distances of the particles in the candidate particle swarm in descending order according to the crowding distances of the particles, and filters out the target particles whose crowding distances meet the preset crowding distance from the candidate particle swarm; and determines the initial product portfolio corresponding to the target particles as the candidate product portfolio that meets the preset target information.

[0114] In this embodiment, by introducing the concepts of a reference point and a preference region, the distance between the particles in the current particle swarm and the particles in the external particle swarm is determined according to the distance between the particles and the reference point, that is, the position information of the particles in the current particle swarm in the target space and the position information of the particles in the external particle swarm in the target space; and target particles that meet the preset conditions are filtered out according to the distance, and the initial product portfolio corresponding to the target particles is determined as the candidate product portfolio that meets the preset target information; thereby effectively filtering out the candidate product portfolio that meets the preset target information, which is beneficial to providing a more accurate product portfolio solution for users.

[0115] In one embodiment, in the above step S201, to obtain the current particle swarm and the external particle swarm according to the multiple particles, the specific content is as follows: confirm the distance between each particle in the multiple particles and the reference point, and filter out the particle with the smallest distance from the multiple particles; filter out the particles in the preference region from the multiple particles; the preference region is a region set according to the reference point, the particle with the smallest distance, and the preference region expansion coefficient; according to the Chebyshev domination method, filter out the non-dominated particle swarm from the particles in the preference region as the external particle swarm, and use the particle swarm except the non-dominated particle swarm among the particles in the preference region as the current particle swarm.

[0116] Among them, the Chebyshev achievement scalar function can be expressed as:

[0117]

[0118] In the formula, m is the target dimension, is a given reference point on the target space, represents the value of the objective function of particle j in i, λ=(λ1,λ2,...,λ m ) is the weight vector, generally is the maximum value of the i-th objective, is the minimum value of the i-th objective, ρ is the augmentation coefficient, generally taking the value 10 -6 .

[0119] The preference region can be expressed as:

[0120] N(z ref ,δ)={z|s ∞ (z,z ref )≤s min +δ)

[0121] where s min =s ∞ (z,z ref ) is the minimum achievement scalar function value in the current population, z is the individual closest to the reference point, and δ is the expansion coefficient of the preference region.

[0122] The value-taking method of the expansion coefficient δ can be expressed as:

[0123]

[0124] where p(t) is the population in the t-th generation, is the maximum achievement scalar function value in the t-th generation population, is the minimum achievement scalar function value in the t-th generation population; the τ value is a fixed value set at the beginning of the algorithm, taking values in the interval [0,1]. The larger the τ value, the larger the preference region. However, the multi-objective particle swarm algorithm itself has the characteristic of premature convergence. If the τ value is taken to be small, the algorithm will search in a fixed preference region in the early stage, losing the diversity of solutions. If the τ value is large, the algorithm will be difficult to converge in the middle and late stages. Therefore, a dynamic τ value-taking is proposed, and the formula is:

[0125] τ=τ max -(τ max -τ min )×(current_t / max_iter) c

[0126] where τ max ,τ min are the maximum and minimum values of τ set by the decision maker in the search stage respectively, current_t is the current iteration number, max_iter is the maximum iteration number, and c is the descent speed;

[0127] Chebyshev domination method: Let x and y be any two different solutions in the feasible region S. If either of the following constraints is satisfied, it means that x dominates y, denoted as: xπ Chebyshev y

[0128] (1) xπ pareto y, and x, y ∈ N(z ref , δ);

[0129] (2) S ∞ (x, z ref ) < S ∞ (y, z ref ), and or Specifically, the terminal calculates the distance between each of multiple particles and a reference point, screens out the particle with the minimum distance from the multiple particles; determines a preference region based on the reference point, the particle with the minimum distance, and a preference region expansion coefficient; screens out the particles within the preference region from the multiple particles; and according to the Chebyshev domination method, screens out a non-dominated particle swarm from the particles within the preference region as an external particle swarm and saves it to an external region, and uses the particle swarm other than the non-dominated particle swarm among the particles within the preference region as the current particle swarm.

[0130] In this embodiment, by introducing the concepts of a reference point and a preference region, the distance between each of multiple particles and the reference point is confirmed, the particle with the minimum distance is screened out from the multiple particles, and finally, according to the Chebyshev domination method, the current particle swarm and the external particle swarm are screened out; thus, multiple particles can be accurately distinguished into the current particle swarm and the external particle swarm by using the distance as an index. In one embodiment, before determining the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product combination in step S103, the following steps are further included: obtaining the historical resource transfer ratio of a preset user for the product to be recommended; constructing a product frequency matrix according to the historical resource transfer ratio of the preset user for the product to be recommended; determining the historical resource transfer ratio of the first target user for each product in the candidate product combination and the historical resource transfer ratio of the second target user for each product in the candidate product combination according to the product frequency matrix; performing a fusion process on the historical resource transfer ratios of the first target user for each product in the candidate product combination to obtain the preference degree of the first target user for the candidate product combination, and performing a fusion process on the historical resource transfer ratios of the second target user for each product in the candidate product combination to obtain the preference degree of the second target user for the candidate product combination.

[0131] Among them, the historical resource transfer ratio refers to the sum of the actual investment ratios for a certain financial product; the product frequency matrix is The processing model for the fusion process is In the formula, k represents a wealth management product portfolio, which includes one or more wealth management products, and L ik represents the preference degree of user i for the k wealth management product portfolio, and R ij represents the sum of the actual investment ratios of customer i for each investment in wealth management product j; the preference degree can be represented by a user-wealth management product preference matrix This matrix is M×N* dimensional, where M represents M users and N* represents N* wealth management product portfolios.

[0132] Specifically, the terminal obtains the historical resource transfer ratio of the preset user for the product to be recommended; constructs a user-wealth management product frequency matrix according to the historical resource transfer ratio of the preset user for the product to be recommended; calculates the sum of the actual investment ratios of the first target user for each product in the candidate product portfolio and the sum of the actual investment ratios of the second target user for each product in the candidate product portfolio according to the user-wealth management product frequency matrix; accumulates and sums up the sum of the actual investment ratios of the first target user for each product in the candidate product portfolio to obtain the preference degree of the first target user for the candidate product portfolio, and accumulates and sums up the sum of the actual investment ratios of the second target user for each product in the candidate product portfolio in the same way to obtain the preference degree of the second target user for the candidate product portfolio.

[0133] In this embodiment, the historical investment ratios of the user for each product in the candidate product portfolio are determined by constructing a product frequency matrix, and the preference degree of the user for the candidate product portfolio is calculated based on the historical investment ratios; thus, the preference degree of the user for the candidate product portfolio can be accurately calculated according to this important information of the historical investment ratios, providing data support for calculating the similarity later.

[0134] In one embodiment, in the above step S103, according to the preference degrees of the first target user and the second target user for the candidate product portfolio, determining the similarity between the first target user and the second target user specifically further includes the following content: inputting the preference degrees of the first target user and the second target user for the candidate product portfolio into the Pearson similarity confirmation model to obtain the Pearson similarity between the first target user and the second target user; and confirming the Pearson similarity between the first target user and the second target user as the similarity between the first target user and the second target user.

[0135] Among them, the Pearson similarity confirmation model is:[[]]

[0136]

[0137] In the formula, Sim(U a ,U b ) is the Pearson similarity between users, and Lak represents the preference degree of user a for the k wealth management product portfolios, L bk represents the preference degree of user b for the k wealth management product portfolios, N * represents N * wealth management product portfolios.

[0138] Specifically, the terminal inputs the preference degrees of the first target user and the second target user for the candidate product portfolio obtained through fusion processing into the Pearson similarity confirmation model for calculation to obtain the Pearson similarity between the first target user and the second target user; and confirms the Pearson similarity between the first target user and the second target user as the similarity between the first target user and the second target user.

[0139] For example, the terminal inputs the preference degree 0.3 of the first target user for the candidate product portfolio and the preference degree 0.4 of the second target user for the candidate product portfolio into the Pearson similarity confirmation model for calculation to obtain the Pearson similarity 0.25 between the first target user and the second target user; and confirms the Pearson similarity 0.25 between the first target user and the second target user as the similarity between the first target user and the second target user.

[0140] In this embodiment, the Pearson similarity confirmation model is used to process the preference degrees to obtain the Pearson similarity between users, and this Pearson similarity is used as the similarity between users; thus, the similarity data between users can be accurately and effectively obtained based on the existing preference degree data.

[0141] In one embodiment, in the above step S104, according to the similarity and the preference degree of the second target user for the candidate product portfolio, determining the recommendation degree of the candidate product portfolio specifically includes the following content: for each second target user, respectively, fuse the similarity between the first target user and the second target user with the preference degree of the second target user for the candidate product portfolio to obtain multiple fusion processing results; and perform a second fusion processing on the multiple fusion processing results to obtain the recommendation degree of the candidate product portfolio.

[0142] Among them, the calculation model for the first fusion processing is In the formula, k represents the wealth management product portfolio, which includes one or more wealth management products, L ik represents the preference degree of user i for the k wealth management product portfolio, R ij represents the sum of the actual investment ratios of customer i for each investment in the j wealth management product; the calculation model for the second fusion processing is:

[0143]

[0144] In the formula, R ajrepresents the recommendation degree of the candidate product portfolio j for user a (i.e., the first target user), Sim(U a ,U i ) is the Pearson similarity between users, and L ij represents the preference degree of user i for the candidate product portfolio j.

[0145] Specifically, the terminal obtains the Pearson similarity Sim(U a ,U i ) calculated by the Pearson similarity confirmation model. For each second target user, the similarity Sim(U a ,U i ) between the first target user and the second target user is respectively input into Sim(U ij ,U a ,U i )×L ij for calculation to obtain multiple calculation results; then the multiple obtained calculation results are accumulated, and the obtained result is used as the recommendation degree of the candidate product portfolio.

[0146] In this embodiment, by performing multiple fusion processes on the similarity data and preference data, the recommendation degree of the candidate product portfolio is obtained; thus, the recommendation degree of the candidate product portfolio can be accurately and effectively obtained according to the existing similarity data and preference data, which is beneficial to providing a more accurate financial product portfolio for users.

[0147] In one embodiment, in the above step S205, according to the recommendation degree, the target product portfolio recommended to the first target user is determined from the candidate product portfolios, which specifically includes the following content: screening out the target product portfolios whose recommendation degrees meet the preset recommendation degree from the candidate product portfolios; and recommending the target product portfolios to the first target user.

[0148] Among them, the preset recommendation degree is a specific recommended value set in advance, such as 0.7.

[0149] Specifically, the terminal screens out the target product portfolios whose recommendation degrees are greater than the preset recommendation degree from the candidate product portfolios according to the obtained recommendation degree of the candidate product portfolio; and recommends the target product portfolios to the first target user.

[0150] Illustratively, the terminal screens out the target product portfolios whose recommendation degrees are greater than the preset recommendation degree of 0.7 from the candidate product portfolios according to the obtained recommendation degree of the candidate product portfolio; and recommends these target product portfolios to the first target user for the first target user to refer to.

[0151] In this embodiment, by screening out a target product portfolio whose recommendation degree meets a preset recommendation degree and recommending the target product portfolio to the current user, the best wealth management product portfolio can be recommended to the user through this important indicator of the recommendation degree.

[0152] In one embodiment, as Figure 4 shown, another product recommendation method is provided, which specifically includes the following steps:

[0153] Step S401, obtain the product information of the products to be recommended; input the product information of the products to be recommended and the total resources of the first target user into the product portfolio prediction models under multiple preset product targets to obtain an initial product portfolio, the predicted resource transfer ratio of each product in the initial product portfolio, and the target information of the initial product portfolio under multiple preset product targets.

[0154] Among them, the initial product portfolio represents the selection combination of the products to be recommended under the constraint of the total resources, and each product in the initial product portfolio is associated with a corresponding predicted resource transfer ratio.

[0155] Step S402, take the target information of the initial product portfolio under multiple preset product targets as a particle to obtain multiple particles; confirm the distance between each particle in the multiple particles and the reference point, and screen out the particle with the smallest distance from the multiple particles; screen out the particles in the preference region from the multiple particles; the preference region is a region set according to the reference point, the particle with the smallest distance, and the preference region expansion coefficient.

[0156] Step S403, according to the Chebyshev domination method, screen out the non-dominated particle swarm from the particles in the preference region as the external particle swarm, and take the particle swarm except the non-dominated particle swarm in the particles in the preference region as the current particle swarm.

[0157] Step S404, according to the position information of the particles in the current particle swarm in the target space and the position information of the particles in the external particle swarm in the target space, confirm the distance between the particles in the current particle swarm and the particles in the external particle swarm; screen out the particles with a distance greater than the average distance from the external particle swarm to obtain a candidate particle swarm.

[0158] Step S405, according to the crowding distance of the particles in the candidate particle swarm, screen out the target particles whose crowding distance meets the preset crowding distance from the candidate particle swarm; confirm the initial product portfolio corresponding to the target particles as the candidate product portfolio that meets the preset target information.

[0159] Step S406: Obtain the historical resource transfer ratio of a preset user for the product to be recommended; construct a product frequency matrix based on the historical resource transfer ratio of the preset user for the product to be recommended; determine the historical resource transfer ratio of each product in the candidate product combination for the first target user and the historical resource transfer ratio of each product in the candidate product combination for the second target user according to the product frequency matrix.

[0160] Step S407: Perform a fusion process on the historical resource transfer ratio of each product in the candidate product combination for the first target user to obtain the preference degree of the first target user for the candidate product combination, and perform a fusion process on the historical resource transfer ratio of each product in the candidate product combination for the second target user to obtain the preference degree of the second target user for the candidate product combination.

[0161] Step S408: Input the preference degrees of the first target user and the second target user for the candidate product combination into the Pearson similarity confirmation model to obtain the Pearson similarity between the first target user and the second target user; confirm the Pearson similarity between the first target user and the second target user as the similarity between the first target user and the second target user.

[0162] Step S409: For each second target user, respectively perform a fusion process on the similarity between the first target user and the second target user and the preference degree of the second target user for the candidate product combination to obtain multiple fusion results; perform a secondary fusion process on the multiple fusion results to obtain the recommendation degree of the candidate product combination.

[0163] Step S410: Screen out the target product combination whose recommendation degree meets the preset recommendation degree from the candidate product combinations; recommend the target product combination to the first target user.

[0164] The above product recommendation method increases the diversity of product combinations and provides a more comprehensive product combination plan for customers; screening out candidate product combinations that meet the preset target information from the initial product combinations is conducive to providing a more accurate product combination plan for users; while recommending a suitable product combination plan for users, fully considering the factors of the customer's own preferences, and recommending product combinations to the first target user based on the objective index of the recommendation degree, avoiding the influence of subjective factors, considering multiple factors comprehensively, taking into account both the diversity and practicality of the product combination plan, and improving the recommendation accuracy of the product combination.

[0165] To more clearly illustrate the product recommendation method provided in the embodiments of the present application, the following uses a specific embodiment to specifically describe the product recommendation method. In one embodiment, as Figure 5 shown, the present application also provides another product recommendation method, which specifically includes the following steps:

[0166] Step 1: Initialize the population, including the velocity, position, local optimal solution, and global optimal solution of the population;

[0167] Step 2: Calculate the values of the particles on the Q, R, and G functions, and determine whether the particles are in the preference region;

[0168] Step 3: According to the Chebyshev domination rule, judge the superiority and inferiority between particles, and save the non-dominated solution set to the external archive;

[0169] Step 4: Use the two-stage global optimal solution selection method to select the respective global optimal solutions for the particles, update the velocity and position, evaluate the particles, and update the local optimal solution;

[0170] Step 5: Update the preference region of the current generation, perform non-dominated sorting on the current generation population, and update the external archive;

[0171] Step 6: Determine whether the iteration times of the current generation are reached;

[0172] Step 7: Take the external archive (local optimal solution) of the last iteration, and the non-dominated solutions contained therein become the optimization objects of the recommendation algorithm;

[0173] Step 8: Construct a customer-financial product preference matrix, and calculate the similarity between each customer;

[0174] Step 9: Calculate the recommendation value for the financial product portfolio according to the similarity value, and select the final financial product portfolio plan according to the size of the recommendation value.

[0175] The above product recommendation method, by establishing an investment and financial management model and optimizing the multi-objective particle swarm optimization algorithm, solves the problems of the decline in the survival pressure of the multi-objective optimization particle swarm algorithm in the multi-objective space and the tendency to fall into local optima; balances the convergence and distribution of the multi-objective optimization particle swarm algorithm; applies the improved multi-objective optimization particle swarm algorithm to the investment and financial management model, and combines the recommendation algorithm to obtain a financial management plan that users are interested in and has a good investment prospect. A financial management recommendation method based on the improved multi-objective optimization particle swarm algorithm combined with the recommendation algorithm proposed in this application establishes an investment and financial management model, uses the improved multi-objective optimization particle swarm algorithm to solve the relatively optimal solution set, combines the recommendation algorithm, and selects a more interesting plan for users in the relatively optimal solution set, providing more professional, efficient, and personalized services for customers.

[0176] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are sequentially shown according to the indications of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same moment, but can be executed at different moments. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0177] Based on the same inventive concept, an embodiment of the present application further provides a product recommendation device for implementing the above-mentioned product recommendation method. The solution provided by this device to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the product recommendation device provided below can refer to the limitations on the product recommendation method in the above text, and will not be repeated here.

[0178] In one embodiment, as Figure 6 shown, a product recommendation device is provided, including: a product combination module 601, a combination screening module 602, a similarity calculation module 603, a recommendation degree calculation module 604, and a combination determination module 605, where:

[0179] The product combination module 601 is configured to determine an initial product combination and the target information of the initial product combination under multiple preset product goals according to the total resources of the product to be recommended and the first target user; the initial product combination represents a selection combination of the product to be recommended under the constraint of the total resources, and each product in the initial product combination is associated with a corresponding predicted resource transfer ratio.

[0180] The combination screening module 602 is configured to screen out candidate product combinations that meet the preset target information from the initial product combination according to the target information of the initial product combination under multiple preset product goals.

[0181] The similarity calculation module 603 is configured to determine the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product combination; the second target user is a user other than the first target user among the preset users.

[0182] The recommendation degree calculation module 604 is configured to determine the recommendation degree of the candidate product combination according to the similarity and the preference degree of the second target user for the candidate product combination.

[0183] A combination determination module 605 is configured to determine a target product combination recommended to a first target user from candidate product combinations according to the recommendation degree.

[0184] In one embodiment, the product combination module 601 is further configured to obtain product information of products to be recommended; input the product information of the products to be recommended and the total resources of the first target user into product combination prediction models under multiple preset product targets, and obtain an initial product combination, a predicted resource transfer ratio of each product in the initial product combination, and target information of the initial product combination under multiple preset product targets.

[0185] In one embodiment, the combination screening module 602 is further configured to use the target information of the initial product combination under multiple preset product targets as a particle to obtain multiple particles, and obtain a current particle swarm and an external particle swarm according to the multiple particles; confirm the distance between the particles in the current particle swarm and the particles in the external particle swarm according to the position information of the particles in the target space in the current particle swarm and the position information of the particles in the target space in the external particle swarm; screen out the particles with a distance greater than the average distance from the external particle swarm to obtain a candidate particle swarm; screen out the target particles with a crowding distance meeting a preset crowding distance from the candidate particle swarm according to the crowding distance of the particles in the candidate particle swarm; and confirm the initial product combination corresponding to the target particles as candidate product combinations meeting the preset target information.

[0186] In one embodiment, the combination screening module 602 is further configured to confirm the distance between each particle in the multiple particles and a reference point, and screen out the particle with the minimum distance from the multiple particles; screen out the particles in a preference area from the multiple particles; the preference area is an area set according to the reference point, the particle with the minimum distance, and a preference area expansion coefficient; screen out a non-dominated particle swarm from the particles in the preference area as the external particle swarm according to the Chebyshev domination method, and use the particle swarm except the non-dominated particle swarm among the particles in the preference area as the current particle swarm.

[0187] In one embodiment, the product recommendation device further includes a preference degree calculation module, configured to obtain the historical resource transfer ratio of a preset user for products to be recommended; construct a product frequency matrix according to the historical resource transfer ratio of the preset user for products to be recommended; determine the historical resource transfer ratio of the first target user for each product in the candidate product combination and the historical resource transfer ratio of the second target user for each product in the candidate product combination according to the product frequency matrix; perform a fusion process on the historical resource transfer ratio of the first target user for each product in the candidate product combination to obtain the preference degree of the first target user for the candidate product combination, and perform a fusion process on the historical resource transfer ratio of the second target user for each product in the candidate product combination to obtain the preference degree of the second target user for the candidate product combination.

[0188] In one embodiment, the similarity calculation module 603 is further configured to input the preference degrees of the first target user and the second target user for the candidate product combination into the Pearson similarity confirmation model to obtain the Pearson similarity between the first target user and the second target user; and confirm the Pearson similarity between the first target user and the second target user as the similarity between the first target user and the second target user.

[0189] In one embodiment, the recommendation degree calculation module 604 is further configured to, for each second target user, respectively fuse the similarity between the first target user and the second target user with the preference degree of the second target user for the candidate product combination to obtain a plurality of fusion processing results; and perform a secondary fusion processing on the plurality of fusion processing results to obtain the recommendation degree of the candidate product combination.

[0190] In one embodiment, the combination determination module 605 is further configured to determine, according to the recommendation degree, a target product combination recommended to the first target user from the candidate product combinations.

[0191] Each module in the above product recommendation device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above respective modules.

[0192] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 7 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as product information, target information, user preference degrees, and user similarities of products to be recommended. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, a product recommendation method is implemented.

[0193] Those skilled in the art can understand, Figure 7The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0194] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0195] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0196] In one embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0197] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0198] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., and are not limited thereto. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., and are not limited thereto.

[0199] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0200] The above-described embodiments only represent several implementation manners of the present application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A product recommendation method, characterized in that, The method includes: Determining an initial product portfolio and the target information of the initial product portfolio under multiple preset product goals according to the total resources of the product to be recommended and the first target user; wherein, the target information includes the variance of the initial product portfolio, the total net income of the initial product portfolio over a period of time, and the balance between the income and risk of the initial product portfolio; the initial product portfolio represents a selection combination of the products to be recommended under the constraint of the total resources, and each product in the initial product portfolio is associated with a corresponding predicted resource transfer ratio; Screening out candidate product portfolios that meet the preset target information from the initial product portfolio according to the target information of the initial product portfolio under multiple preset product goals; Determining the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolios; the second target user is a user other than the first target user among the preset users; Determining the recommendation degree of the candidate product portfolio according to the similarity and the preference degree of the second target user for the candidate product portfolio; Determining the target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree; The screening out candidate product portfolios that meet the preset target information from the initial product portfolio according to the target information of the initial product portfolio under multiple preset product goals includes: Regarding the target information of the initial product portfolio under multiple preset product goals as a particle, obtaining multiple particles, and obtaining a current particle swarm and an external particle swarm according to the multiple particles; Confirming the distance between the particles in the current particle swarm and the particles in the external particle swarm according to the position information of the particles in the current particle swarm in the target space and the position information of the particles in the external particle swarm in the target space; Screening out the particles in the external particle swarm with the distance greater than the average distance to obtain a candidate particle swarm; Screening out the target particles with the crowding distance meeting the preset crowding distance from the candidate particle swarm according to the crowding distance of the particles in the candidate particle swarm; Regarding the initial product portfolio corresponding to the target particles as the candidate product portfolio that meets the preset target information.

2. The method according to claim 1, wherein The determining an initial product portfolio and the target information of the initial product portfolio under multiple preset product goals according to the total resources of the product to be recommended and the first target user includes: Obtaining the product information of the product to be recommended; Inputting the product information of the product to be recommended and the total resources of the first target user into a product portfolio prediction model under multiple preset product goals to obtain an initial product portfolio, the predicted resource transfer ratio of each product in the initial product portfolio, and the target information of the initial product portfolio under the multiple preset product goals.

3. The method according to claim 1, wherein The obtaining a current particle swarm and an external particle swarm according to the multiple particles includes: Confirming the distance between each particle in the multiple particles and a reference point, and screening out the particle with the smallest distance from the multiple particles; Screen out the particles in the preference region from the multiple particles; the preference region is a region set according to the reference point, the particle with the minimum distance, and the preference region expansion coefficient; According to the Chebyshev domination method, screen out the non-dominated particle swarm from the particles in the preference region as the external particle swarm, and use the particle swarm other than the non-dominated particle swarm among the particles in the preference region as the current particle swarm.

4. The method according to claim 1, wherein Before determining the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product combination, it further includes: Obtain the historical resource transfer ratio of the preset user for the product to be recommended; Construct a product frequency matrix according to the historical resource transfer ratio of the preset user for the product to be recommended; According to the product frequency matrix, determine the historical resource transfer ratio of the first target user for each product in the candidate product combination, and the historical resource transfer ratio of the second target user for each product in the candidate product combination; Perform a fusion process on the historical resource transfer ratios of the first target user for each product in the candidate product combination to obtain the preference degree of the first target user for the candidate product combination, and perform a fusion process on the historical resource transfer ratios of the second target user for each product in the candidate product combination to obtain the preference degree of the second target user for the candidate product combination.

5. The method according to claim 4, characterized in that, The determining the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product combination includes: Input the preference degrees of the first target user and the second target user for the candidate product combination into the Pearson similarity confirmation model to obtain the Pearson similarity between the first target user and the second target user; Confirm the Pearson similarity between the first target user and the second target user as the similarity between the first target user and the second target user.

6. The method according to any one of claims 1 to 5, characterized in that The determining the recommendation degree of the candidate product combination according to the similarity and the preference degree of the second target user for the candidate product combination includes: For each second target user, respectively perform a fusion process on the similarity between the first target user and the second target user and the preference degree of the second target user for the candidate product combination to obtain a plurality of fusion process results; Perform a re-fusion process on the plurality of fusion process results to obtain the recommendation degree of the candidate product combination.

7. The method according to claim 6, wherein The determining the target product combination recommended to the first target user from the candidate product combination according to the recommendation degree includes: Screen out the target product combination whose recommendation degree meets the preset recommendation degree from the candidate product combination; Recommend the target product combination to the first target user.

8. A product recommendation device, characterized in that, The device includes: A product portfolio module, configured to determine an initial product portfolio and the target information of the initial product portfolio under multiple preset product goals according to the products to be recommended and the total resources of the first target user; wherein, the target information includes the variance of the initial product portfolio, the total net income of the initial product portfolio over a period of time, and the balance between the income and risk of the initial product portfolio; the initial product portfolio represents a selection combination of the products to be recommended under the constraint of the total resources, and each product in the initial product portfolio is associated with a corresponding predicted resource transfer ratio. A portfolio screening module, configured to screen out candidate product portfolios that meet the preset target information from the initial product portfolio according to the target information of the initial product portfolio under multiple preset product goals. A similarity calculation module, configured to determine the similarity between the first target user and the second target user according to the preference degrees of the first target user and the second target user for the candidate product portfolios; the second target user is a user other than the first target user among the preset users. A recommendation degree calculation module, configured to determine the recommendation degree of the candidate product portfolios according to the similarity and the preference degrees of the second target user for the candidate product portfolios. A portfolio determination module, configured to determine the target product portfolio recommended to the first target user from the candidate product portfolios according to the recommendation degree. Specifically, the portfolio screening module is configured to use the target information of the initial product portfolio under multiple preset product goals as a particle to obtain multiple particles, and obtain a current particle swarm and an external particle swarm according to the multiple particles; according to the position information of the particles in the current particle swarm in the target space and the position information of the particles in the external particle swarm in the target space, confirm the distance between the particles in the current particle swarm and the particles in the external particle swarm; screen out the particles with a distance greater than the average distance from the external particle swarm to obtain a candidate particle swarm; screen out the target particles whose crowding distance meets the preset crowding distance from the candidate particle swarm according to the crowding distance of the particles in the candidate particle swarm; and confirm the initial product portfolio corresponding to the target particles as the candidate product portfolio that meets the preset target information.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images