Resource recommendation method and device, equipment, storage medium and program product

Abstract

The embodiment of the invention provides a resource recommendation method and device, equipment, a storage medium and a program product, and relates to the field of big data. The method comprises the following steps: clustering stock users according to user information, generating a plurality of user clusters, and determining attributes of the user clusters; determining an evaluation level of each user cluster according to the attribute of each user cluster; determining a to-be-evaluated risk type corresponding to each user cluster from a plurality of risk types at least according to the attribute of each user cluster; based on the attribute of each user cluster, determining the risk level of each user cluster under the corresponding to-be-evaluated risk type; and according to the assessment level of each user cluster and the risk level of each user cluster under the corresponding to-be-assessed risk type, recommending resources to users of each user cluster. According to the resource recommendation method and device, resources are recommended to the users with the user cluster as the unit, and compared with a recommendation mode with a single user as granularity, the calculation amount is saved, and the resource recommendation efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of big data, and in particular to a resource recommendation method, apparatus, device, storage medium, and program product. Background Technology

[0002] In the financial industry, it is often necessary to recommend resources to users based on their potential needs. For example, recommending suitable financial products to users at a certain frequency, dynamically recommending loan amounts, and recommending services to users.

[0003] Because users' potential needs are diverse, recommendations are typically analyzed on a per-user basis to ensure accuracy. However, in scenarios with a large existing user base, this approach suffers from high computational demands and excessive resource consumption, making it difficult to meet the real-time recommendation needs of a large customer base and hindering the scalability of recommendation services. Summary of the Invention

[0004] This application provides a resource recommendation method, apparatus, device, storage medium, and program product. It introduces clustering to recommend resources on a per-user-cluster basis, reducing the computational load and improving recommendation efficiency. At the same time, to ensure recommendation accuracy, it adds evaluation levels and risk levels determined based on user-cluster attributes.

[0005] Firstly, this application provides a resource recommendation method, comprising: clustering existing users based on user information to generate multiple user clusters and determining the attributes of each user cluster; determining the evaluation level of each user cluster based on the attributes of each user cluster; determining the risk type to be evaluated for each user cluster from multiple risk types based at least on the attributes of each user cluster; determining the risk level of each user cluster under the corresponding risk type to be evaluated based on the attributes of each user cluster; and recommending resources to users of each user cluster based on the evaluation level of each user cluster and the risk level of each user cluster under the corresponding risk type to be evaluated.

[0006] Secondly, this application provides a resource recommendation device, comprising: a classification module, used to cluster existing users according to user information to generate multiple user clusters and determine the attributes of each user cluster; an evaluation level determination module, used to determine the evaluation level of each user cluster according to the attributes of each user cluster; a risk type determination module, used to determine the risk type to be evaluated corresponding to each user cluster from multiple risk types, at least according to the attributes of each user cluster; a risk level determination module, used to determine the risk level of each user cluster under the corresponding risk type to be evaluated based on the attributes of each user cluster; and a resource recommendation module, used to recommend resources to users of each user cluster according to the evaluation level of each user cluster and the risk level of each user cluster under the corresponding risk type to be evaluated.

[0007] Thirdly, this application provides an electronic device, including: a processor and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided in the first aspect above.

[0008] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method provided in the first aspect above.

[0009] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method provided in the first aspect above.

[0010] The resource recommendation method, apparatus, device, storage medium, and program product provided in this application implement a resource recommendation strategy based on user cluster attributes, using clustered user clusters as the granularity. First, in the clustering stage, existing users in the system are clustered to obtain multiple user clusters composed of users with high similarity. In the recommendation stage, the evaluation level and risk level of each user cluster are determined using its attributes. Based on the evaluation level and risk level, resources are recommended to users within each user cluster. By recommending resources at the user cluster granularity, the number of recommendations is reduced, the computational load of resource recommendation is lowered, and the efficiency of recommending resources to users is improved. Simultaneously, by combining the evaluation level and risk level of a user cluster, the methods and resources needed by that type of user are comprehensively determined and recommended to that type of user. This improves the efficiency of resource recommendation while ensuring the accuracy of the recommended resources and methods. Attached Figure Description

[0011] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0012] Figure 1 A flowchart illustrating a resource recommendation method provided in an embodiment of this application;

[0013] Figure 2 A schematic diagram illustrating a resource recommendation scenario provided in an embodiment of this application;

[0014] Figure 3 A flowchart illustrating another resource recommendation method provided in an embodiment of this application;

[0015] Figure 4 A flowchart illustrating yet another resource recommendation method provided in an embodiment of this application;

[0016] Figure 5 For this application Figure 4A schematic diagram of one implementation of the embodiment shown;

[0017] Figure 6 This is a schematic diagram of the structure of a resource recommendation device provided in an embodiment of this application;

[0018] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0019] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0021] 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 used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with the relevant laws, regulations, and standards of the relevant countries and regions, have taken necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation portals for users to choose to authorize or refuse.

[0022] Furthermore, the technical solution involved in this application, which involves big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.) and the use of artificial intelligence technology for automated decision-making, and makes decisions that have a significant impact on personal rights based on the results of automated decision-making, provides users with corresponding operation entry points for users to choose to agree to or reject the results of automated decision-making; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0023] It should be noted that the resource recommendation method, apparatus, equipment, storage medium, and program products provided in this application can be used in the field of big data, or in any field other than big data. The application fields of the resource recommendation method, apparatus, equipment, storage medium, and program products in this application are not limited.

[0024] This application can be specifically applied to scenarios where marketing plans are developed for users within a banking and financial system. Generally, a banking and financial system includes various services, such as deposits, loans, credit cards, and wealth management. Depending on the user's circumstances, different service needs may exist. For example, some retirees may have a potential need for deposits; some businesses may have a potential need for loans. The banking and financial system needs to recommend appropriate resources to users with different potential needs based on their information. For instance, for retirees, deposit products can be recommended to them at a certain time and frequency based on their information, such as deposit frequency and term; for businesses, a certain loan amount can be provided based on their information, such as stable returns and liquidity.

[0025] Currently, the method for recommending relevant resources to users often involves analyzing the user's potential needs based on their information within the banking and financial system, including static and behavioral data. Resources are then recommended to the user in a specific manner based on these potential needs. However, banking and financial systems typically contain a large number of users. Analyzing each user individually would result in a huge computational burden, consuming system resources and taking excessive time, leading to low efficiency in resource recommendations.

[0026] This application provides a resource recommendation method aimed at solving the aforementioned technical problems of existing technologies without compromising the accuracy of user resource recommendations. Specifically, users are categorized to generate multiple user clusters composed of highly similar users. Based on the attributes of these user clusters, the evaluation level and risk level of each cluster are analyzed. The evaluation level reflects the activity level of users within the cluster, and the risk level includes at least one risk type. Based on the evaluation level and risk level, a method for recommending resources to users is determined, and resources are recommended to users within the corresponding user cluster. This method achieves resource recommendations to highly similar users within the same user cluster, avoiding analysis of individual users, significantly reducing computational load, and improving the efficiency of resource recommendations. Furthermore, this method combines evaluation level and risk level, enabling precise marketing while maintaining controllable risk and improving the accuracy of resource recommendations.

[0027] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0028] Figure 1This is a flowchart illustrating a resource recommendation method provided in an embodiment of this application. The method provided in this embodiment can be executed by an electronic device with corresponding data processing capabilities, such as a computer, server, or other electronic device. Figure 1 As shown, the resource recommendation method provided in this embodiment includes the following steps:

[0029] Step S101: Based on user information, cluster existing users to generate multiple user clusters and determine the attributes of each user cluster.

[0030] Existing users are those recorded and stored in the banking and financial system. In one example, existing users can refer to users who have registered in the banking and financial system and have conducted business interactions, such as transactions and inquiries. Users include individual users and corporate users.

[0031] User information is multi-dimensional data collected by banking and financial systems in accordance with laws, regulations, and business needs. It includes basic information and user behavior information. Basic information refers to user-specific information that remains unchanged over a long period or is updated infrequently, including but not limited to the user's name and age. User behavior information is statistically derived from various operations, habits, and preferences generated by the user within the banking and financial system, including but not limited to the user's consumption index, income index, stability index, asset index, platform activity level, and loan index.

[0032] User behavior information consists of quantified metrics based on user actions, used to reflect user habits and preferences in different aspects. Specifically, user actions can be quantified, such as normalized to 0-100, to generate corresponding information in the user behavior data.

[0033] In one example, a piece of information can be related to multiple operations. After quantifying each operation, the information can be obtained by weighted summation based on pre-set weights for each operation. For instance, assuming the consumption index is related to the amount and frequency of a user's consumption, the amount and frequency over a period of time can first be normalized to 0-100. Then, based on the set weights of each parameter (e.g., amount accounts for 60%, frequency accounts for 40%), a weighted summation is performed to generate the consumption index. Other information items in user behavior data can be generated using the same method as the consumption index, and will not be listed here.

[0034] In real-world scenarios, banking and financial systems respond to user actions and submitted information, determine the user's basic information and behavioral information, and store this user information, including basic information and behavioral information, in the banking and financial system's data lake.

[0035] Based on the above scenario, user information of existing users is retrieved from the data lake. Based on this information, existing users are clustered into multiple user clusters. A user cluster refers to a set of existing users that share similar characteristics.

[0036] User cluster attributes refer to the combination of information that distinguishes each user cluster from other user clusters or the similarities between users within the same cluster, after clustering. These attributes describe the basic information and user behavior of the user cluster as a whole. User cluster attributes can correspond to various information items in the user information; for example, user cluster attributes may include one or more of the following: age, consumption index, income index, stability index, asset index, platform activity level, and loan index. Specifically, user cluster attributes can be at least one typical piece of information from the cluster center, or at least one statistical value corresponding to the user information of each user in the cluster, such as the mean or mode. For example, if users with similar platform activity levels are grouped into a user cluster, then the average platform activity level of the users in the cluster is calculated as the user cluster's platform activity attribute.

[0037] In one example, the attributes of a user cluster may also include other attributes, which can be determined based on the user information of each user in the user cluster. For example, the attributes of a user cluster may include investment preferences, which can be conservative, balanced, and aggressive.

[0038] For example, the attributes of a user cluster may include, but are not limited to, age group, consumption index, income index, and investment preferences.

[0039] In this embodiment, the clustering of existing users can be achieved using the Spectral Clustering algorithm. Spectral Clustering is a graph-based clustering algorithm. It treats existing users as vertices in a graph, uses the correlation between vertices as the edge weights connecting them, and divides the weighted graph into two or more optimal subgraphs, maximizing the distance between subgraphs and minimizing the proximity or similarity within subgraphs. Commonly used clustering algorithms, such as K-Means Clustering Algorithm, C-Means Clustering Algorithm, and neural networks, all require calculating the distance between two existing users. However, for high-dimensional data, this distance calculation often fails, failing to fully reveal the underlying structure of existing users. Spectral Clustering, by transforming the clustering problem into an optimal graph partitioning problem, is more intuitive, better suited to high-dimensional data, and offers higher efficiency and accuracy in clustering.

[0040] Specifically, clustering existing users using spectral clustering algorithms includes:

[0041] Based on the user information of each existing user, a graph is constructed and an adjacency matrix W is generated. The vertices of the graph correspond to each existing user, and the adjacency matrix W is used to represent the similarity between the vertices.

[0042] Given the adjacency matrix W, calculate the Laplace matrix L; the Laplace matrix L = DW, where D is a diagonal matrix representing the sum of elements in a row of the adjacency matrix W;

[0043] Calculate the first k smallest eigenvalues ​​of the Laplacian matrix L and extract the corresponding eigenvectors to construct an n*k dimensional eigenvector space; where n is the number of existing users and k is the number of clusters.

[0044] The extracted feature vectors are clustered using clustering algorithms (such as K-means clustering) to determine k user clusters.

[0045] In this embodiment, each element in the adjacency matrix W represents the edge weight between fixed points. The similarity between existing users is calculated using their user information, and this similarity constitutes the adjacency matrix W. The similarity can be calculated using existing Gaussian kernel functions, which will not be described in detail here.

[0046] Clustering algorithms (such as K-means clustering) are used to cluster the extracted feature vectors to determine k user clusters. Specifically, each row of the constructed n*k dimensional feature vector space is used as a sample, and a clustering algorithm is used to classify the existing users.

[0047] In one example, to improve the efficiency and accuracy of clustering, the adjacency matrix W can be reduced in dimensionality before generating the adjacency matrix W based on the edges in the graph.

[0048] Step S102: Determine the evaluation level of each user cluster based on the attributes of each user cluster.

[0049] The assessment rating is a classification of user groups based on their activity levels. Its purpose is to reflect the characteristics and potential needs of these user groups. User activity refers to the frequency and depth of a user's interaction with the banking and financial system within a certain period. Factors such as transaction frequency, product usage breadth, interactive behavior, and the scale of fund flows can reflect user activity.

[0050] User clusters can be divided into multiple levels based on their activity level. For example, the activity level of a user cluster can be quantified as a value from 0 to 100. If two evaluation levels are set, the first level can be a user cluster with an activity level of 0-50; the second level can be a user cluster with an activity level of 51-100.

[0051] The activity level of a user cluster can be determined by identifying at least one attribute among the user cluster's attributes that is related to the user cluster's activity level, and presetting the weight value of each attribute. The weight value reflects the contribution of the attribute to the activity level and is a configurable parameter. The activity level of the user cluster is obtained by weighted summation of at least one data point.

[0052] For example, the weights of platform activity and consumption index can be set to 40% and 60% respectively, and the quantified attributes of the user cluster can be weighted and summed, that is, user cluster activity = 40% * platform activity + 60% * consumption index.

[0053] Based on the activity level of a user cluster, the activity range within which it falls is determined, and the corresponding evaluation level is assigned to that user cluster. For example, if the calculated activity level is 60, then according to the above example, the evaluation level of this user cluster is Level 2. A higher evaluation level indicates higher user activity in the user cluster, which may require more resources from the banking and financial system to meet its needs.

[0054] In another example, to more accurately reflect the characteristics and potential needs of user clusters and avoid simple numerical classification, multiple evaluation levels can be set, including HP Support, Needs Exploration, Stable and Active, Deep Value, and Value Active, corresponding to levels one through five, respectively. For each evaluation level, different ranges of relevant attributes can be pre-set. When the attributes of a user cluster meet the range, it is determined as the evaluation level of that user cluster.

[0055] For example, for Level 1 (HP Support), the relevant attribute can be set to a consumption index. When the consumption index is less than a first threshold, this user group belongs to Level 1. For Level 2 (Demand Exploration), the relevant attribute can be set to platform activity. When platform activity is less than a first threshold, this user group belongs to Level 2. For Level 3 (Stable and Active), the relevant attribute can be set to platform activity. When platform activity is greater than or equal to a second threshold, this user group belongs to Level 3. For Level 4 (Deep Value), the relevant attribute can be set to a consumption index. When the consumption index is greater than or equal to a first threshold, this user group belongs to Level 4. For Level 5 (Value Active), the relevant attributes can be set to both consumption index and platform activity. When the consumption index is greater than or equal to a first threshold and platform activity is greater than or equal to a second threshold, this user group belongs to Level 5. For instance, a user group whose consumption amount in the past year exceeds a preset amount and whose platform activity is greater than or equal to a preset frequency can be set as Level 5, i.e., Value Active.

[0056] It should be noted that the above-described method for determining evaluation levels is merely exemplary and does not represent all possible methods. Evaluation levels can be categorized into other levels or classified in other ways, which are not listed here. Furthermore, the evaluation levels are used to analyze the potential demand for recommended resources among users within a user cluster and do not involve any value judgments about users.

[0057] Step S103: Determine the risk type to be assessed for each user cluster from multiple risk types, based at least on the attributes of each user cluster.

[0058] User actions can introduce multi-dimensional risks to the banking and financial system. Based on their causes, these risks can be categorized as credit risk, operational risk, and compliance risk. Credit risk refers to the risk of loss due to a user's inability to fulfill contractual obligations, such as loan defaults or excessive debt. Operational risk refers to the risk of loss due to user interaction with the system, such as frequent system access leading to resource consumption or user complaints. Compliance risk refers to the risk of loss due to user misconduct, such as abnormal operating environments. The risk types to be assessed refer to the types of risks that this user group may face.

[0059] For different risk types, the users who may pose that risk have typical characteristics. For example, user clusters that may lead to compliance risks are mostly enterprise-oriented and have high platform activity. Therefore, based on the attributes of each user cluster that are related to each risk type, the risk type to be assessed for each user cluster can be determined from multiple risk types.

[0060] For example, attributes related to credit risk can be pre-defined as income index, stability index, and loan index; attributes related to operational risk can be platform activity; and attributes related to compliance risk can be consumption index and income index. When abnormal attributes exist in a user cluster, the risk type related to the abnormal attribute is determined as the risk type to be assessed.

[0061] In some embodiments, in addition to the attributes of each user cluster, other information can be used to determine the type of risk to be assessed. For example, based on the users in each user cluster, abnormal operations and risk-generating operations of each user can be identified, and the risk type can be determined accordingly. For instance, if multiple users in a user cluster are identified to have abnormal operations such as large-amount transfers in the early morning or frequent large-amount transfers, there may be compliance risks, and compliance risks can be identified as the type of risk to be assessed.

[0062] Step S104: Based on the attributes of each user cluster, determine the risk level of each user cluster under the corresponding risk type to be assessed.

[0063] The risk level under the risk type to be assessed is a classification index used to characterize the probability of that risk type occurring. Multiple risk levels can be set based on the probability of occurrence, such as low risk, medium risk, and high risk.

[0064] The risk level can be determined by a pre-trained risk prediction model, which involves inputting the attributes of the user cluster into the risk prediction model and then outputting the risk level.

[0065] Alternatively, the risk level can be determined by statistically analyzing the risks generated by each user in the user cluster, based on the frequency and number of times the user in the cluster experiences the risk type to be assessed. For example, if no user in the user cluster experiences the risk within a certain time period, the risk level of that risk type is considered extremely low; if the number of times the risk occurs in the user cluster within a certain time period exceeds a preset high-risk threshold, the risk level of that risk type is considered high.

[0066] By analyzing the risk levels under multiple risk types, compared to analyzing only the overall risk level, this approach avoids ignoring the details of specific risk types. For example, the overall risk of a user cluster may be low, but the risk of a certain risk type may be extremely high. At the same time, it also avoids ignoring the synergistic effects between risks, making the risk assessment of user clusters more accurate.

[0067] Step S105: Based on the assessment level of each user cluster and the risk level of each user cluster under the corresponding risk type to be assessed, recommend resources to users of each user cluster.

[0068] Resources refer to the combination of financial products, services, and messages provided by the banking system. Financial products include deposit products, wealth management products, credit cards, etc. Services include service methods, such as whether abnormal transactions are monitored and triggered for review, loan limit ranges, etc. Messages include message content and message delivery methods. For example, message content may be account reminders or product recommendations, and message delivery methods may include telephone, SMS, and message delivery frequency.

[0069] In real-world scenarios, different users have different resource needs. For example, individual users apply for loans less frequently; users who prefer stable investments use financial products less often. Therefore, appropriate resource recommendation strategies can be matched to users based on their operational patterns. This application, through assessment levels and risk levels, can represent the general operational patterns of users in each user group, and recommend resources to users in each user group based on these general operational patterns.

[0070] Specifically, based on the combination of assessment levels and risk levels under the risk types to be assessed, a pre-defined correspondence between each combination and resources can be established. Based on the assessment levels of user clusters and the corresponding risk levels under the risk types to be assessed obtained above, resources corresponding to user clusters are determined from the correspondence, and resources are recommended to users within those user clusters.

[0071] For example, based on the above example, two assessment levels can be set (the first level can be a user group with an activity level of 0-60; the second level is a user group with an activity level of 61-100). If the user group's assessment level is the second level and all risk levels are low risk, it can be set as the first resource recommendation strategy, such as pre-approving a credit limit of 50,000. If the user group's assessment level is the second level and all risk levels are medium risk, it can be set as the second resource recommendation strategy, such as setting a credit limit of 30,000, but requiring supplementary income proof. If the user group's assessment level is the first level and all risk levels are medium-high risk or above, it can be set as the third resource recommendation strategy, such as not actively marketing and monitoring abnormal transactions.

[0072] Figure 2 This is a schematic diagram illustrating a resource recommendation scenario provided in an embodiment of this application. For example... Figure 2 As shown, multiple users are first clustered to obtain multiple user clusters, such as... Figure 2 The system identifies user clusters 1, 2, and 3. It determines the assessment ratings and risk levels of the risk types to be assessed for user clusters 1 through 3. Assuming user cluster 1 has an assessment rating of Level 2, the risk type to be assessed is credit risk, and the corresponding risk level is low, then user cluster 1 corresponds to the first resource recommendation strategy. User cluster 2 also has an assessment rating of Level 2, and the risk types to be assessed are credit risk and operational risk, with risk levels of low and medium respectively; therefore, user cluster 2 corresponds to the second resource recommendation strategy. User cluster 3 has an assessment rating of Level 1, the risk type to be assessed is operational risk, and the risk level of operational risk is high; therefore, user cluster 3 corresponds to the third resource recommendation strategy.

[0073] The resource recommendation method provided in this embodiment implements a resource recommendation strategy based on user cluster attributes, using clustered user clusters as the granularity. First, in the clustering stage, the existing users in the system are clustered to obtain multiple user clusters composed of users with high similarity. In the recommendation stage, the evaluation level and risk level of each user cluster are determined using its attributes. Based on the evaluation level and risk level, resources are recommended to users within each user cluster. By recommending resources at the user cluster level, the number of recommendations is reduced, the computational load of resource recommendation is lowered, and the efficiency of recommending resources to users is improved. Simultaneously, by combining the evaluation level and risk level of a user cluster, the appropriate methods and resources for that type of user are comprehensively determined and recommended. This improves the efficiency of resource recommendation while ensuring the accuracy of the recommended resources and methods.

[0074] Figure 3 This is a flowchart illustrating another resource recommendation method provided in an embodiment of this application. Figure 3 As shown, this implementation is in Figure 1 Based on the illustrated embodiment, steps S103-S105 have been refined. Specifically, the resource recommendation method provided in this embodiment includes the following steps:

[0075] Step S301: Based on user information, cluster existing users to generate multiple user clusters and determine the attributes of each user cluster.

[0076] Step S302: Determine the evaluation level of each user cluster based on the attributes of each user cluster.

[0077] Step S303: Based on the attribute thresholds corresponding to each risk type among multiple risk types and the attributes of each user cluster, determine the risk type to be evaluated for each user cluster from among multiple risk types.

[0078] Among them, at least one attribute of the user cluster exceeds the attribute threshold limit of the corresponding risk type to be assessed.

[0079] In this embodiment, multiple risk types include credit risk, operational risk, and compliance risk.

[0080] Specifically, the correspondence between each risk type and the attributes of the user cluster is determined, and attribute thresholds are set for the corresponding attributes. These thresholds are configurable parameters, allowing administrators to configure them based on historical risk data. The attributes of each user cluster are compared with the corresponding thresholds for each risk type. If at least one attribute exceeds the threshold for a given risk type, it indicates a higher probability of that type of risk occurring in that user cluster, and this risk type is identified as the risk type to be assessed.

[0081] For example, the attributes corresponding to credit risk are set as consumption index and income index, and the attribute thresholds are set as a first consumption index threshold and a first income index threshold. The range of the attribute thresholds is that the consumption index is less than the first consumption index threshold, and the income index is greater than or equal to the first income index threshold. The attribute corresponding to operational risk is set as platform activity, and the attribute threshold is set as a first platform activity threshold. The range of the attribute thresholds is that the platform activity is less than the first platform activity threshold. The attributes corresponding to compliance risk are set as consumption index, income index, and platform activity, and the attribute thresholds are set as a second consumption index threshold, a second income index threshold, and a second platform activity threshold. The range of the attribute thresholds is that the consumption index is less than the second consumption index threshold, or the income index is less than the second income index threshold, or the platform activity is less than the second platform activity threshold.

[0082] If there exists a user cluster whose consumption index is greater than or equal to the second consumption index threshold and the first consumption index threshold, whose income index is less than the second income index threshold but greater than the first income index threshold, and whose platform activity is less than the first platform activity threshold and the second platform activity threshold, then the attributes corresponding to this user cluster exceed the threshold limits for credit risk and compliance risk, and credit risk and compliance risk are identified as risk types to be assessed.

[0083] Step S304: For each risk type to be evaluated corresponding to each user cluster, obtain the model parameters corresponding to the risk type to be evaluated.

[0084] Step S305: Configure the initial prediction model based on the model parameters corresponding to the risk type to be assessed, and obtain the risk level prediction model corresponding to the risk type to be assessed.

[0085] Step S306: Input the attributes of the user cluster into the risk level prediction model corresponding to the risk type to be evaluated, and obtain the risk level of the user cluster under the corresponding risk type to be evaluated.

[0086] After identifying the type of risk to be assessed, it is necessary to evaluate that type of risk and determine its risk level. The attributes of the user cluster can be input into a risk level prediction model corresponding to the risk level to be assessed, thus obtaining the risk level of the user cluster under the corresponding risk type.

[0087] Specifically, model parameters for each risk type can be obtained through model training, or model parameters for each risk type can be pre-set and stored in the system. For each risk type to be assessed, the corresponding model parameters are retrieved from the system and configured into the initial prediction model to obtain the risk level prediction model corresponding to the risk type to be assessed. Finally, the attributes of the user cluster are input into this risk level prediction model to obtain the risk level of the user cluster under the risk type to be assessed.

[0088] The initial prediction model can be a statistical model, a machine learning model, etc. Statistical models include the normal distribution, the generalized autoregressive conditional heteroscedasticity model, etc. Machine learning models include support vector machines, artificial neural networks, and random forest methods, etc.

[0089] The preferred approach is to use a Hidden Markov Model (HMM). An HMM is a statistical model that can predict hidden states within input data that are not directly observable. Compared to other models that are sensitive to hypothetical data and heavily reliant on historical data, HMMs offer greater interpretability and efficiency.

[0090] Optionally, in order to obtain the model parameters of the initial prediction model, the method provided in this implementation also includes a model training method. Specifically, a sample set is obtained, which includes samples corresponding to multiple user clusters and risk labels corresponding to each sample. The samples corresponding to user clusters include user information of sample users within the user cluster and attributes of the user cluster. Among them, multiple user clusters are obtained by clustering existing users. Based on the samples in the sample set whose risk labels are of the same risk type, the initial prediction model is trained to obtain and store the model parameters corresponding to each risk type.

[0091] Risk labels refer to the occurrence of a particular risk by a sample user under each risk type, such as the frequency and duration. Based on the occurrence of the risk, the risk level of the sample user can be determined. For example, if a sample user frequently experiences credit risk issues within a year, then the risk level of that sample user under the credit risk type is high risk.

[0092] One way to obtain the sample set is to acquire multiple existing users in the data lake. These existing users can be clustered according to the methods described in the above embodiments to generate multiple user clusters, and the attributes of these user clusters can be obtained. The user information and risk tags of each existing user can also be obtained. A risk tag for an existing user can include a credit risk tag, an operational risk tag, and a compliance risk tag. Each sample corresponds to the user information of a sample user, the attributes of the user cluster to which the sample user belongs, and the corresponding risk tag.

[0093] After obtaining the sample set, the sample set is used as the training set to train the model. Specifically, samples with the same risk label are used as a training set to train the initial prediction model, obtain the model parameters corresponding to each risk type, and store them as the model parameters corresponding to that risk type.

[0094] In this embodiment, taking the initial training model as a Hidden Markov Model as an example, the model can be represented as λ=(S,O,π,P,Q). Wherein, O is the observation sequence, which is obtained by combining user information and the attributes of the user cluster to which the user belongs. It is the set of states associated with the hidden state obtained by direct observation; S is the hidden state, representing the risk level of each sample user under a certain risk type; P is the state transition matrix; π is the initial probability distribution vector; and Q is the observation vector probability matrix.

[0095] The state transition matrix P can be represented as:

[0096] 0≤p ij ≤1,

[0097] The observation vector probability matrix Q can be represented as:

[0098] Q = (q jk ) N×M ,0≤j≤1,1≤k≤M

[0099] The deviation between the observed sequence O and the Hidden Markov Model λ=(S,O,π,P,Q) is determined by calculating P(O|λ). That is, the probability of the observed sequence O occurring under the model λ=(S,O,π,P,Q) is calculated. The larger the calculated value, the closer the observed sequence is to the given model.

[0100] Given a specific model, the hidden state sequence S is estimated to more effectively match the observed sequence O. In other words, given the observed sequence O, the optimal possible hidden state sequence S is calculated, and the risk level can be determined based on the hidden state sequence. To make the determined hidden state sequence more accurate, the Viterbi algorithm can be used. The Viterbi algorithm is a dynamic programming algorithm used to find the Viterbi path (hidden state sequence S) most likely to produce the observed event sequence.

[0101] For example, if the number of risk levels optimized by the Hidden Markov Model is 7, then according to the order of the hidden state sequence S, the corresponding risk levels are extremely low risk, relatively low risk, low risk, medium risk, medium-high risk, relatively high risk, and high risk.

[0102] In some embodiments, the small sample size of certain user clusters may lead to class imbalance, resulting in misjudgment of risk. For example, the identification of students with no income and low-income groups may be inaccurate because they do not conform to the mainstream income-debt model and have a small sample size.

[0103] To address this issue, the SMOTE (Synthetic MinorityOversampling Technique) algorithm can be introduced in the above embodiments. This algorithm synthesizes new simulated sample users based on the user information of the sample users, thereby resolving the class imbalance problem.

[0104] Step S307: Weighted summation of the risk levels of the user cluster under the corresponding risk type to be assessed to determine the overall risk level of the user cluster.

[0105] If a user cluster may correspond to multiple risk types to be assessed, dimensionality reduction can be performed to avoid data processing difficulties caused by multidimensional data. Specifically, the risk levels under multiple risk types to be assessed can be weighted and summed to determine an overall risk level. The weight of each risk level is a configurable parameter that can be set according to actual conditions.

[0106] Step S308: Recommend resources to users in each user cluster based on the assessment level of each user cluster and the overall risk level.

[0107] Based on the assessment level and the overall risk level, a "value-risk dual-dimensional strategy" can be formed through cross-analysis of the assessment level and the overall risk level, generating a 4-quadrant strategy matrix. Resources are recommended to users in each user cluster based on the strategies determined in the 4-quadrant strategy matrix. Table 1 illustrates the 4-quadrant strategy matrix as an example.

[0108] Table 1

[0109] The assessment level is Level 2. The assessment level is Level 1. Low risk Resource Priority Strategy Standardization strategy High risk Risk hedging strategy Restricted service strategy

[0110] Risk hedging strategies can provide users with more resources and increase their priority, such as increasing the frequency of resource recommendations and recommending more products; standardization strategies can be pre-set, unified resource recommendation strategies; risk hedging strategies can provide users with more resources while monitoring them; restrictive service strategies can reduce the resources provided to users while monitoring their operations, such as manual review and operation verification.

[0111] In some embodiments, to facilitate subsequent data processing, the assessment level and risk level can be quantified. Weight values ​​for the assessment level and each risk level are pre-set, and the assessment level and each risk level are weighted and summed to obtain a comprehensive score for the user cluster. Based on the comprehensive score, the corresponding resource is determined, and the resource is recommended to users in the user cluster.

[0112] Optionally, resources are recommended to users in each user cluster based on the assessment level and the overall risk level. This includes: for each user cluster, determining multiple candidate recommended resources that match the overall risk level; determining target recommended resources that match the assessment level from the multiple candidate recommended resources based on the assessment level, and providing the target recommended resources to users in the user cluster.

[0113] Candidate recommended resources are determined based on the risk level of user clusters. For example, if the risk of a user cluster is high, more conservative resources are mainly considered, along with whether services to the user are restricted or the user is monitored. After determining the candidate recommended resources, resources that match the evaluation level are selected from multiple candidate recommended resources.

[0114] This method enables collaborative analysis of user activity and risk levels. Furthermore, it is simple to use and more efficient.

[0115] In this embodiment, the risk type to be assessed is determined by judging whether the attributes of the user cluster exceed the attribute threshold limit range of the corresponding risk type to be assessed. This judgment method is simple and effective. Furthermore, by determining the configuration parameters corresponding to the risk type to be assessed, the risk level prediction model is dynamically adjusted, which improves the applicability of the risk level prediction model. By modifying the configuration parameters, a single model can output multiple risk levels, improving the efficiency of obtaining multiple risk levels. At the same time, by weighted summation of the risk levels of different risk types, the influence of each risk type on the overall situation is clarified through the set weights. Without ignoring individual risk types, the data is dimensionality reduced, improving the execution efficiency of the method.

[0116] Optionally, in addition to the implementation of step S103 provided in the above embodiments, step S103 can also be implemented in the following way. Specifically, step S103 determines the risk type to be evaluated corresponding to each user cluster from multiple risk types based at least on the attributes of each user cluster, including: determining the risk type to be evaluated corresponding to each user cluster from multiple risk types based on the attributes and evaluation level of each user cluster.

[0117] Generally, determining whether a user possesses a certain risk type is based on multiple factors, meaning it involves judging through multiple attributes; judging with only a limited number of attributes will yield inaccurate results. To improve the accuracy and efficiency of determining the risk type to be assessed, it can be determined based on data and assessment levels for each user cluster.

[0118] Specifically, the assessment level can generally represent a certain group. Based on the types of risks that frequently occur in that group, the types of risks that may occur can be quickly determined. Next, based on the attributes of each user cluster, it is further determined whether the types of risks that may occur are the types of risks to be assessed.

[0119] Optionally, based on the attributes and assessment levels of each user cluster, the risk type to be assessed for each user cluster is determined from multiple risk types, including: determining candidate risk types from multiple risk types based on the assessment level; and determining the risk type to be assessed for each user cluster from the candidate risk types based on the attributes of the user clusters associated with each candidate risk type.

[0120] Specifically, historical data corresponding to the assessment level can be obtained. This historical data includes user and user behavior information for that assessment level. Based on the risk types that occurred in the historical data corresponding to the assessment level, these risk types are identified as candidate risk types. Furthermore, based on whether at least one attribute corresponding to a candidate risk type exceeds the attribute threshold limit of the corresponding candidate risk type, it is determined whether the candidate risk type is the risk type to be assessed.

[0121] This approach avoids the need to determine whether the attributes of user clusters exceed the attribute threshold range for each risk type, reducing computational load and improving the efficiency of determining the risk type to be evaluated.

[0122] Optionally, based on the attributes and assessment levels of each user cluster, the risk type to be assessed for each user cluster is determined from multiple risk types, including: obtaining the initial attribute thresholds corresponding to each risk type from multiple risk types; adjusting the initial attribute thresholds based on the assessment level; and determining the risk type to be assessed for each user cluster from multiple risk types based on the comparison results between the adjusted initial attribute thresholds corresponding to each risk type and the attributes of the user cluster.

[0123] In real-world scenarios, users with different assessment levels may experience different types of risks. For example, the threshold values ​​for credit risk attributes differ between student groups and working users.

[0124] In this embodiment, initial attribute thresholds corresponding to each risk type can be preset. Before determining the risk type to be evaluated, the initial attribute thresholds can be adjusted according to the evaluation level, for example, linearly adjusting the initial attribute thresholds according to the evaluation level. The risk type to be evaluated is determined based on the comparison results between the adjusted initial attribute thresholds corresponding to each risk type and the attributes of the user cluster. The method of determining the risk type to be evaluated based on the comparison results between the adjusted initial attribute thresholds corresponding to each risk type and the attributes of the user cluster is the same as that in the above embodiment, only the value of the attribute thresholds is different, and will not be elaborated here.

[0125] For example, the initial consumption index threshold for credit risk can be set as the first consumption index threshold, and the consumption index threshold for user groups with higher activity can be linearly increased according to the assessment level.

[0126] By dynamically adjusting the initial attribute thresholds, the applicability of the method for determining the type of risk to be assessed is improved, as well as the flexibility and accuracy of the determination method.

[0127] Figure 4 This is a flowchart illustrating yet another resource recommendation method provided in an embodiment of this application. For example... Figure 4 As shown, in order to better understand the resource recommendation method provided in this embodiment, this embodiment will provide a detailed description of the method.

[0128] Step S401: Based on user information, cluster existing users to generate multiple user clusters and determine the attributes of each user cluster.

[0129] In this method, existing users are clustered using spectral clustering to generate multiple user clusters, and the attributes of the user clusters are the information corresponding to the cluster centers.

[0130] Step S402: Determine the evaluation level of each user cluster based on the attributes of each user cluster.

[0131] Based on the attributes of each user cluster, the activity level of the user cluster is determined, and an evaluation level is determined based on the activity level. The activity level of a user cluster can be calculated by weighted summing of the consumption index of each user cluster and the platform activity level.

[0132] Step S403: Obtain the initial attribute threshold corresponding to each risk type among multiple risk types; and adjust the initial attribute threshold based on the assessment level.

[0133] Step S404: Based on the comparison results between the initial attribute thresholds corresponding to each risk type and the attributes of the user clusters after adjustment, determine the risk type to be evaluated for each user cluster from multiple risk types.

[0134] Step S405: For each risk type to be evaluated corresponding to each user cluster, obtain the model parameters corresponding to the risk type to be evaluated.

[0135] Step S406: Configure the initial prediction model based on the model parameters corresponding to the risk type to be assessed, and obtain the risk level prediction model corresponding to the risk type to be assessed.

[0136] Step S407: Input the attributes of the user cluster into the risk level prediction model corresponding to the risk type to be evaluated, and obtain the risk level of the user cluster under the corresponding risk type to be evaluated.

[0137] The initial prediction model is a Hidden Markov Model (HMM), and the configuration parameters are obtained by training the HMM based on the sample set.

[0138] Step S408: Weighted summation of the risk levels of the user cluster under the corresponding risk type to be assessed to determine the overall risk level of the user cluster.

[0139] Step S409: For each user cluster, determine multiple candidate recommendation resources that match the overall risk level.

[0140] Step S410: Based on the evaluation level, determine the target recommendation resource that matches the evaluation level from multiple candidate recommendation resources, and provide the target recommendation resource to users in the user cluster.

[0141] Figure 5 For this application Figure 4 A structural schematic diagram of one implementation of the illustrated embodiment. (See diagram below.) Figure 5 As shown, the method also includes a training method for the initial prediction model. Specifically, user information is used as input, and spectral clustering (the specific implementation of which is described in the above embodiment) is used to obtain multiple user clusters and their attributes. The user information and the attributes of the user clusters are then used as a training set to train a Hidden Markov Model, generating risk prediction models for each risk type. An evaluation level is determined based on the attributes of the multiple user clusters. The attributes of the multiple user clusters are then input into the risk level prediction model corresponding to the risk type to be evaluated to obtain the risk level of the risk type. Based on the evaluation level and the risk level of the risk type to be evaluated, a final resource recommendation strategy is obtained. Resources are then provided to users of the user clusters according to this resource recommendation strategy.

[0142] The resource recommendation method provided in this application is similar in technical effect to the method provided in any of the above embodiments of this application, and will not be described again here.

[0143] Figure 6 This is a schematic diagram of the structure of a resource recommendation device provided in an embodiment of this application, as shown below. Figure 6 As shown, the resource recommendation device provided in this embodiment includes: a classification module 601, an assessment level determination module 602, a risk type determination module 603, a risk level determination module 604, and a resource recommendation module 605.

[0144] The classification module 601 is used to cluster existing users based on user information, generate multiple user clusters, and determine the attributes of each user cluster; the assessment level determination module 602 is used to determine the assessment level of each user cluster based on the attributes of each user cluster; the risk type determination module 603 is used to determine the risk type to be assessed for each user cluster from multiple risk types, based at least on the attributes of each user cluster; the risk level determination module 604 is used to determine the risk level of each user cluster under the corresponding risk type to be assessed based on the attributes of each user cluster; and the resource recommendation module 605 is used to recommend resources to users of each user cluster based on the assessment level of each user cluster and the risk level of each user cluster under the corresponding risk type to be assessed.

[0145] Optionally, the risk type determination module 603 is specifically used to: determine the risk type to be evaluated for each user cluster from the multiple risk types based on the attribute thresholds corresponding to each risk type and the attributes of each user cluster, wherein at least one attribute of the user cluster exceeds the range limited by the attribute threshold of the corresponding risk type to be evaluated.

[0146] Optionally, the risk type determination module 603 is specifically used to: determine the risk type to be evaluated for each user cluster from multiple risk types based on the attributes and evaluation level of each user cluster.

[0147] Correspondingly, the risk type determination module 603 is specifically used to: determine candidate risk types from multiple risk types based on the assessment level; and determine the risk type to be assessed for each user cluster from the candidate risk types based on the attributes of the user clusters associated with each candidate risk type.

[0148] Correspondingly, the risk type determination module 603 is specifically used to: obtain the initial attribute threshold corresponding to each risk type among multiple risk types; adjust the initial attribute threshold based on the assessment level; and determine the risk type to be assessed for each user cluster from multiple risk types based on the comparison results between the adjusted initial attribute threshold corresponding to each risk type and the attributes of the user cluster.

[0149] Optionally, the risk level determination module 604 is specifically used for: obtaining the model parameters corresponding to each risk type for each user cluster; configuring an initial prediction model based on the model parameters corresponding to the risk type to obtain the risk level prediction model corresponding to the risk type; and inputting the attributes of the user cluster into the risk level prediction model corresponding to the risk type to obtain the risk level of the risk type corresponding to the user class.

[0150] Correspondingly, the resource recommendation device also includes a model training module, wherein: the model training module is used to obtain a sample set, the sample set includes samples corresponding to multiple user clusters and risk labels corresponding to each sample, the samples corresponding to user clusters include user information of sample users within the user cluster and attributes of the user cluster; wherein, multiple user clusters are obtained by clustering existing users; based on samples in the sample set whose risk labels are of the same risk type, an initial prediction model is trained to obtain and store the model parameters corresponding to the risk type.

[0151] Optionally, the resource recommendation module 605 is specifically used to: weight and sum the risk levels of user clusters under the corresponding risk types to be assessed to determine the overall risk level of the user clusters; and recommend resources to users of each user cluster based on the assessment level of each user cluster and the overall risk level.

[0152] Optionally, the resource recommendation module 605 is specifically used to: weight and sum the risk levels of user clusters under the corresponding risk types to be assessed to determine the overall risk level of the user cluster; for each user cluster, determine multiple candidate recommended resources that match the overall risk level based on the overall risk level; and determine the target recommended resource that matches the assessment level from the multiple candidate recommended resources based on the assessment level, and provide the target recommended resource to the users of the user cluster.

[0153] The resource recommendation device provided in this application embodiment can be used to execute the technical solution of the resource recommendation method provided in any of the above embodiments of this application. Its implementation principle and technical effect are similar, and will not be described again here.

[0154] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 7 As shown, the electronic device of this embodiment may include: at least one processor 701; and a memory 702 communicatively connected to at least one processor; wherein the memory 702 stores instructions that can be executed by at least one processor 701, and the instructions are executed by at least one processor 701 to cause the electronic device to perform the method as described in any of the above embodiments.

[0155] Optionally, the memory 702 can be either standalone or integrated with the processor 701. When the memory 702 is set up independently, the device also includes a bus for connecting the memory 702 and the processor 701.

[0156] The implementation principle and technical effects of the electronic device provided in this embodiment can be found in the foregoing embodiments, and will not be repeated here.

[0157] This application also provides a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are executed by a processor, the methods provided in any of the foregoing embodiments can be implemented.

[0158] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the method provided in any of the foregoing embodiments.

[0159] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0160] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0161] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components can be combined, or integrated into another system, or some features can be ignored or not executed.

[0162] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0163] When an integrated unit / module is implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. Unless otherwise specified, the processor can be any suitable hardware processor, such as a CPU (Central Processing Unit), GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), DSP (Digital Signal Processor), and ASIC (Application Specific Integrated Circuit), etc. Unless otherwise specified, the storage unit can be any suitable magnetic or magneto-optical storage medium, such as resistive random access memory (RRAM), dynamic random access memory (DRAM), static random access memory (SRAM), enhanced dynamic random access memory (EDRAM), high-bandwidth memory (HBM), hybrid memory cube (HMC), etc.

[0164] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0165] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0166] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0167] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A resource recommendation method, characterized in that, include: Based on user information, existing users are clustered to obtain multiple user clusters, and the attributes of each user cluster are determined. The evaluation level of each user cluster is determined based on its attributes. Based at least on the attributes of each user cluster, determine the risk type to be assessed for each user cluster from multiple risk types; Based on the attributes of each user cluster, determine the risk level of each user cluster under the corresponding risk type to be assessed; Based on the evaluation level of each user cluster and the risk level of each user cluster under the corresponding risk type to be evaluated, resources are recommended to users of each user cluster.

2. The method according to claim 1, characterized in that, The step of determining the risk type to be assessed for each user cluster from multiple risk types, at least based on the attributes of each user cluster, includes: Based on the attribute thresholds corresponding to each of the multiple risk types and the attributes of each user cluster, the risk type to be evaluated corresponding to each user cluster is determined from the multiple risk types, wherein at least one attribute of the user cluster exceeds the range limited by the attribute threshold of the corresponding risk type to be evaluated.

3. The method according to claim 1, characterized in that, The step of determining the risk type to be assessed for each user cluster from multiple risk types, at least based on the attributes of each user cluster, includes: Based on the attributes of each user cluster and the assessment level, the risk type to be assessed for each user cluster is determined from the plurality of risk types.

4. The method according to claim 3, characterized in that, The step of determining the risk type to be assessed for each user cluster from the plurality of risk types based on the attributes of each user cluster and the assessment level includes: Based on the assessment level, candidate risk types are determined from the plurality of risk types; Based on the attributes of the user clusters associated with each of the candidate risk types, determine the risk type to be evaluated for each user cluster from the candidate risk types.

5. The method according to claim 3, characterized in that, The step of determining the risk type to be assessed for each user cluster from the plurality of risk types based on the attributes of each user cluster and the assessment level includes: Obtain the initial attribute threshold corresponding to each of the multiple risk types; and adjust the initial attribute threshold based on the assessment level; Based on the comparison results between the adjusted initial attribute thresholds corresponding to each of the risk types and the attributes of the user clusters, the risk type to be evaluated for each user cluster is determined from the plurality of risk types.

6. The method according to claim 1, characterized in that, The step of determining the risk level of each user cluster under the corresponding risk type to be assessed based on the attributes of each user cluster includes: For each risk type to be evaluated corresponding to each user cluster, obtain the model parameters corresponding to the risk type to be evaluated; Based on the model parameters corresponding to the risk type to be assessed, an initial prediction model is configured to obtain the risk level prediction model corresponding to the risk type to be assessed. The attributes of the user cluster are input into the risk level prediction model corresponding to the risk type to be evaluated, and the risk level of the user cluster under the corresponding risk type to be evaluated is obtained.

7. The method according to claim 6, characterized in that, The method further includes: Obtain a sample set, which includes samples corresponding to multiple user clusters and risk labels corresponding to each sample. The samples corresponding to the user clusters include user information of the sample users within the user clusters and the attributes of the user clusters. The multiple user clusters are obtained by clustering the existing users. Based on samples in the sample set that have the same risk type, the initial prediction model is trained to obtain and store the model parameters corresponding to each risk type.

8. The method according to any one of claims 1 to 7, characterized in that, The step of recommending resources to users in each user cluster based on the evaluation level of each user cluster and the risk level of each user cluster under the corresponding risk type to be evaluated includes: The overall risk level of the user cluster is determined by weighted summation of the risk levels of the user cluster under the corresponding risk types to be evaluated. Resources are recommended to users in each user cluster based on the assessment level of each user cluster and the overall risk level.

9. The method according to claim 8, characterized in that, The step of recommending resources to users in each user cluster based on the assessment level of each user cluster and the overall risk level includes: For each user cluster, based on the overall risk level, multiple candidate recommendation resources matching the overall risk level are determined; Based on the evaluation level, a target recommended resource matching the evaluation level is determined from multiple candidate recommended resources, and the target recommended resource is provided to users of the user cluster.

10. A resource recommendation device, characterized in that, include: The classification module is used to cluster existing users based on user information, generate multiple user clusters, and determine the attributes of each user cluster. The evaluation level determination module is used to determine the evaluation level of each user cluster based on the attributes of each user cluster. The risk type determination module is used to determine the risk type to be evaluated for each user cluster from multiple risk types, based at least on the attributes of each user cluster. The risk level determination module is used to determine the risk level of each user cluster under the corresponding risk type to be assessed based on the attributes of each user cluster. The resource recommendation module is used to recommend resources to users of each user cluster based on the evaluation level of each user cluster and the risk level of each user cluster under the corresponding risk type to be evaluated.

11. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 9.

13. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 9.

Images