Product recommendation method and device based on customer demand, equipment and medium

Abstract

The application relates to the technical field of intelligent decision-making, can be applied to business system platforms such as medical health and financial technology, and discloses a product recommendation method, device and equipment based on customer demand and a medium, multi-dimensional data of a customer and clause explanations of products are acquired, and a customer portrait is constructed according to the multi-dimensional data; demand analysis is carried out on the multi-dimensional data according to the customer portrait, and a customer demand graph is obtained; product features are extracted from the clause explanations, and product features are obtained; the matching degree between the customer demand graph and the product features is analyzed; and a product recommendation scheme is generated according to the matching degree and preset customer risk preferences. By analyzing the demand of multi-dimensional data, generating a product recommendation scheme according to the matching degree and customer risk preferences, the product recommendation accuracy is improved, and the product recommendation scheme is in line with the core demand of the customer.

Description

Technical Field

[0001] This invention relates to the field of intelligent decision-making technology, and in particular to a method, apparatus, equipment, and medium for product recommendation based on customer needs. Background Technology

[0002] In product marketing scenarios, personalized product recommendations have become a core means to improve customer conversion rates and optimize service experiences, and this intelligent recommendation model has deeply penetrated multiple fields. For example, in the healthcare field, based on multi-dimensional information such as patients' health records, medical records, genetic data, and lifestyle habits, personalized health management plans, treatment suggestions, rehabilitation services, and suitable medical products are accurately matched. As another example, in the fintech field, relying on data such as customers' financial status, risk preferences, consumption behavior, and life cycle events, customized recommendations of financial solutions such as insurance products, wealth management products, and credit services are provided to users, achieving an efficient match between needs and services.

[0003] However, existing product recommendation methods generally have the following shortcomings: On the one hand, existing methods often rely on single-dimensional customer data or historical interaction records for product recommendations, lacking the ability to dynamically perceive the needs of customers throughout their entire life cycle, resulting in recommended products failing to meet customers' true demands; on the other hand, product feature extraction mostly stays at the level of basic attributes (such as premiums and coverage periods), failing to fully explore the core information in the terms and conditions, resulting in insufficiently accurate product feature descriptions, which in turn affects the accuracy of matching customer needs with products.

[0004] Therefore, in the face of the growing demand for product recommendations, current product recommendation methods based on customer needs urgently need to be improved to address the problem of insufficient accuracy in recommending products using existing methods. Summary of the Invention

[0005] This invention provides a product recommendation method, apparatus, equipment, and medium based on customer needs, which addresses how to improve the accuracy of product recommendations while also meeting the core needs of customers.

[0006] Firstly, a product recommendation method based on customer needs is provided, including: Obtain multi-dimensional customer data and product terms and conditions, and construct customer profiles based on the multi-dimensional data; Based on the customer profile, a demand analysis is performed on the multi-dimensional data to obtain a customer demand map. Product features are extracted from the aforementioned terms and conditions to obtain product features; Analyze the matching degree between the customer demand map and the product features; A product recommendation scheme is generated based on the matching degree and the preset customer risk preferences.

[0007] Secondly, a product recommendation device based on customer needs is provided, including: The customer profile building module is used to obtain multi-dimensional customer data and product terms and conditions, and to build a customer profile based on the multi-dimensional data. The demand mapping analysis module is used to perform demand analysis on the multi-dimensional data based on the customer profile to obtain a customer demand mapping. The product feature extraction module is used to extract product features from the terms and conditions to obtain product features. The demand-product matching degree analysis module is used to analyze the matching degree between the customer demand map and the product features; The recommendation scheme generation module is used to generate product recommendation schemes based on the matching degree and preset customer risk preferences.

[0008] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described product recommendation method based on customer needs.

[0009] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned product recommendation method based on customer needs.

[0010] In the aforementioned solution implemented by a product recommendation method, apparatus, computer equipment, and storage medium based on customer needs, a customer profile is constructed through multi-dimensional data, transforming scattered information into an interpretable customer profile to solve the problem of ambiguous customer needs. By analyzing needs through multi-dimensional data, abstract needs are transformed into a structured demand graph, improving the accuracy of product recommendations. By extracting product features from the terms and conditions, redundancy and semantic ambiguity in the terms and conditions are eliminated, thereby making the product information matchable. By calculating similarity, subjective judgment is replaced to avoid blind recommendations. A product recommendation scheme is generated based on the matching degree and customer risk preferences, ensuring that the recommended products meet the core needs of customers while also being suitable for their risk tolerance. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1This is a schematic diagram of an application environment for a product recommendation method based on customer needs according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a product recommendation method based on customer needs in one embodiment of the present invention; Figure 3 This is a schematic diagram of a product recommendation device based on customer needs according to one embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device according to one embodiment of the present invention; Figure 5 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] The user personal information involved in these embodiments of the invention is all authorized (with knowledge and consent) by the relevant parties or fully authorized by all parties, and the executing entity may obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with the relevant laws and regulations of the relevant countries and regions, and do not violate public order and good morals.

[0015] This invention provides a product recommendation method based on customer needs, which can be applied to, for example... Figure 1In this application environment, the client communicates with the server via a network. The server can obtain multi-dimensional customer data and product terms and conditions, and construct a customer profile based on the multi-dimensional data; perform demand analysis on the multi-dimensional data based on the customer profile to obtain a customer demand map; extract product features from the terms and conditions to obtain product features; analyze the matching degree between the customer demand map and the product features; generate a product recommendation plan based on the matching degree and preset customer risk preferences, and feed the product recommendation plan back to the client. This invention provides a product recommendation device based on customer needs. For product recommendation plan business, by analyzing the demand of multi-dimensional data, it generates a product recommendation plan based on the matching degree and customer risk preferences, improving the accuracy of product recommendations while meeting the core needs of customers. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.

[0016] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a product recommendation method based on customer needs, provided by an embodiment of the present invention, includes the following steps: S1. Obtain multi-dimensional customer data and product terms and conditions, and construct a customer profile based on the multi-dimensional data.

[0017] In this embodiment of the invention, the multi-dimensional data includes basic customer information, health status, financial data, preference data (such as past insurance purchase records and consultation records), and external related data (such as credit data and insurance orders); the terms and conditions description refers to the core terms and key explanatory information of the product itself. For example, the terms and conditions description of an insurance product includes the explanation of coverage, the explanation of exclusions, and the explanation of underwriting rules.

[0018] In the medical setting, multi-dimensional customer data (including basic information, past medical history, physical examination reports, and medical records) and terms and conditions of medical services / devices (such as applicable diseases and contraindications for treatment items) are obtained. Based on the data, customer profiles covering characteristics such as health status, treatment needs, and medical insurance compatibility are constructed to provide a basis for accurate triage and personalized treatment plan recommendations.

[0019] In fintech scenarios, multi-dimensional customer data (including basic information, financial status, risk preferences, credit records, etc.) and terms and conditions of financial products (such as the coverage, underwriting rules, and premium amounts of insurance products) are acquired. Based on this multi-dimensional data, customer profiles covering characteristics such as financial capacity and risk tolerance are constructed to support personalized financial product recommendations.

[0020] In this embodiment of the invention, constructing a customer profile based on the multi-dimensional data includes: By linking and fusing the multi-dimensional data, the customer's overall data is obtained; Core behavioral features are extracted from the customer's overall data to generate core behavioral features; Construct a high-dimensional feature matrix based on the core behavioral characteristics; The high-dimensional feature matrix is ​​subjected to strong correlation feature filtering to obtain retained features; Customer profiles are constructed based on the retained features.

[0021] In this invention, during the knowledge graph association and fusion process of multi-dimensional data, the multi-dimensional data is preprocessed, including data cleaning (such as removing missing values ​​and outliers), unifying field formats, and structured transformation (such as OCR technology to parse medical invoices and voice semantic analysis of call recordings), to obtain preprocessed data.

[0022] Subsequently, information such as "health records" is extracted from the preprocessed data as entities, information such as "age and health indicator values" is extracted from the preprocessed data as attributes, and information such as "customer-held-insurance orders" is extracted from the preprocessed data to identify entity relationships. A knowledge graph is constructed based on entities, attributes, and entity relationships, and multi-dimensional data is cross-domain linked and integrated based on the entity relationships in the knowledge graph to eliminate data silos and form a unified data view across the entire domain, namely, customer full-domain data.

[0023] Furthermore, in the process of extracting core behavioral characteristics from customer's full-domain data, information such as "purchase frequency and insurance amount preference" is first extracted from multi-dimensional data to form purchase behavior characteristics; information such as "physical examination frequency and health indicator compliance" is extracted from multi-dimensional data to form health management behavior characteristics; and purchase behavior characteristics and health management behavior characteristics are combined to form core behavioral characteristics.

[0024] Furthermore, in the process of constructing a high-dimensional feature matrix based on core behavioral features, the core behavioral features are split into multiple feature dimensions. The unique identifier of the customer (such as ID card number) is used as the row index and the feature dimension is used as the column index. The quantified value of each customer in the feature dimension is used as the matrix element to construct the high-dimensional feature matrix.

[0025] In this invention, during the process of screening strongly correlated features in a high-dimensional feature matrix, the Pearson correlation coefficient can be used to calculate the correlation coefficient between feature dimension columns in the high-dimensional feature matrix. Redundant feature dimensions with correlation coefficients higher than a set threshold (such as 0.85) are eliminated. The cross-validation method is used to determine whether the retained feature dimensions can cover the core needs of customers (such as security preferences and budget constraints). The feature dimensions that pass the validation are used as retained features.

[0026] Furthermore, when building a customer profile, the retained features are quantified, and the retained feature labels are mapped according to the quantified values ​​and preset thresholds. All mapped labels are then aggregated to form a customer profile.

[0027] For example, retain the following characteristics: percentage of financial investment in the past year, quantitative value: 75%, preset threshold rule: ≥60% = "investment tendency: aggressive", mapping label: investment tendency: aggressive; Retained characteristics: Risk tolerance, quantified value: 1, preset threshold rule: 1 = "Risk preference: Conservative", mapping label: Risk preference: Conservative; The summarized customer profile (customer mobile number 13098939020) is as follows: "Investment Tendency": "Aggressive", "Risk Preference": "Conservative".

[0028] S2. Based on the customer profile, perform demand analysis on the multi-dimensional data to obtain a customer demand map.

[0029] In the healthcare context, based on customer profiles that include features such as basic customer information, past medical history, and physical examination data, demand analysis is conducted on the aforementioned multi-dimensional health-related data to uncover customers' core treatment needs (such as chronic disease management and serious disease screening), potential health needs (such as rehabilitation care and health intervention), and demand constraints (such as medical insurance compatibility and medical access convenience requirements), thus obtaining a customer demand map.

[0030] In fintech scenarios, based on customer profiles that include characteristics such as basic customer information, financial status, and asset size, demand analysis is conducted on the aforementioned multi-dimensional financial data to identify customers' core financial needs (such as critical illness protection and wealth appreciation), potential needs (such as retirement planning and children's education fund reserves), and constraints (such as budget limits and risk tolerance thresholds), thus obtaining a customer demand map.

[0031] In this embodiment of the invention, the step of performing demand analysis on the multi-dimensional data based on the customer profile to obtain a customer demand map includes: The multi-dimensional data is classified and labeled based on the profile features of the customer profile to form classified and labeled data; The classification and labeling data are spatiotemporally correlated and encoded to obtain spatiotemporally correlated data; Perform demand feature analysis on spatiotemporal correlated data to generate demand features; A customer demand map is obtained by constructing a graph based on the aforementioned demand characteristics.

[0032] In this embodiment of the invention, the portrait features refer to the retained features, and the spatiotemporal correlation data refers to data that integrates timestamps and spatial / scene tags on the basis of classification and labeling data to establish spatiotemporal correlation, making the demand analysis more in line with the time sequence pattern.

[0033] In this embodiment of the invention, during the process of classifying and labeling multi-dimensional data, mapping rules are established by using the portrait features (i.e. retained features) in the customer profile and the corresponding mapping labels to form a labeling rule library. Then, the multi-dimensional data is compared with the labeling rule library, and the rule engine matches the corresponding label for each piece of data in the multi-dimensional data. All data that successfully match the label are summarized to form classification and labeling data.

[0034] Furthermore, precise timestamps are added to the categorized and labeled data, and the data is uniformly formatted according to a four-level time granularity of "year-month-day-hour" to obtain timestamped labeled data. The timestamped labeled data is then matched with a preset customer lifecycle event database. For example, "submitted insurance application on 2024-10-15" matches the "insurance decision" event to obtain data bound to lifecycle event tags. Subsequently, a ternary coding rule is used to perform spatiotemporal correlation coding on the data bound to lifecycle event tags to obtain spatiotemporal correlated data.

[0035] Furthermore, in the process of demand feature analysis of spatiotemporal related data, the textual information in the spatiotemporal related data is semantically parsed through a large model. The parsed content is used as potential demand features, the matching labels of the classified label data are used as label features, the frequency of the label features in the spatiotemporal related data is counted, and the label features that appear more than a preset threshold (e.g., ≥5 times) are used as core features. The potential demand features and the core features are combined to form demand features.

[0036] Furthermore, when constructing the demand feature graph, multi-dimensional data corresponding to the demand features are extracted, and the logical relationships between the extracted multi-dimensional data are analyzed. Based on the logical relationships, core entities and their corresponding attribute content are extracted from the extracted multi-dimensional data (the graph construction process is the same as step S1, and will not be elaborated here). According to the core entity, attribute, and logical relationship framework, the corresponding content (corresponding multi-dimensional data) is filled in to form the customer demand graph.

[0037] S3. Extract product features from the aforementioned terms and conditions to obtain product features.

[0038] In the medical setting, product features are extracted from medical service plans, drug instructions, or medical device terms of use to obtain product features that include core information such as the applicable population, covered diseases / treatment items, service response time, contraindications, and data privacy compliance level.

[0039] In fintech scenarios, product features are extracted from insurance product terms, wealth management product prospectuses, or credit service agreements to obtain product features containing core information such as coverage / wealth management limits, compensation / repayment rules, exclusions, payment / term requirements, and risk levels.

[0040] In this embodiment of the invention, the step of extracting product features from the terms and conditions to obtain product features includes: The aforementioned clause descriptions are preprocessed to obtain the standard clause text; The product feature dimensions are determined based on the standard clause text, and the standard clause text is matched based on the product feature dimensions to generate a set of matching texts. The core semantics of the matched text set are extracted to obtain the core semantics of the text, and preliminary product features are generated based on the core semantics of the text. Calculate the semantic similarity between the preliminary product features, and perform redundancy filtering on the preliminary product features based on the semantic similarity to obtain the product features.

[0041] In this embodiment of the invention, during the preprocessing of the insurance product terms and conditions, the clause description text can be split using a sentence segmentation and word segmentation algorithm in natural language processing technology. Redundant modifiers and meaningless characters in the clause description after sentence and word segmentation are filtered out using regular expressions. The remaining clause description is then tagged with parts of speech, and the field format of the remaining clause description is standardized to obtain standardized clause text.

[0042] In this invention, during the process of determining product feature dimensions based on standard clause text, information such as "scope of coverage, exclusions, and compensation ratio" is extracted from the standard clause text to form product feature dimensions. Subsequently, corresponding keywords are matched for each product feature dimension to form a standardized mapping library. A text matching algorithm is used to compare the standard clause text with the standardized mapping library to identify text fragments containing keywords of the product feature dimensions. These text fragments are then compiled into a set of matching texts.

[0043] Furthermore, in the process of extracting core semantics from the matching text set, dependency parsing techniques can be used to mine the core semantics of the text fragments of each product feature dimension in the matching text set, and the mined core semantics can be transformed in the format of "product feature dimension + core semantics" and the transformed core semantics can be used as the initial product features.

[0044] Furthermore, in the process of calculating the semantic similarity between preliminary product features, each preliminary product feature is first converted into a semantic vector. The cosine similarity algorithm is used to calculate the angle between any two semantic vectors to obtain the semantic similarity. Finally, the semantic similarity between two preliminary product features is compared with a preset similarity threshold. If it exceeds the preset similarity threshold (e.g., 0.8), it is determined that there are redundant features between the two preliminary product features. Features that have no substantial semantics or are weakly related to the core attributes of the product are removed from the two preliminary product features to obtain the product features.

[0045] In this embodiment of the invention, by determining the product feature dimensions, core feature dimensions are extracted from standard clauses, and corresponding text fragments are filtered according to the dimensions. This achieves dimensional classification and focus of clause information, avoiding the blindness of feature extraction. By extracting the core semantics of the matching text set, redundant expressions in the text are stripped away, and core semantics are mined, abstract clause information is transformed into quantifiable and interpretable product attributes. By calculating semantic similarity, synonymous and near-synonymous redundant expressions in the initial product features are eliminated, and core and unique product features are retained. This avoids matching interference caused by feature duplication and improves the efficiency and accuracy of subsequent matching calculations.

[0046] S4. Analyze the matching degree between the customer demand map and the product features.

[0047] In this embodiment of the invention, analyzing the matching degree between the customer demand map and the product features includes: The customer demand map and the product features are dimensionally aligned to obtain a standardized matching dataset. The standardized matching dataset is initially screened by a rule engine to form a preliminary matching candidate set; The fuzzy matching degree is calculated for the preliminary matching candidate set to determine the basic matching score; The basic matching score is optimized through hierarchical analysis to obtain the matching degree.

[0048] In this embodiment of the invention, the standardized matching dataset refers to a comparison dataset formed by aligning the customer demand map and product features according to a unified dimensional system and standardizing the terminology. The standardized matching dataset includes the aligned customer demand map and the aligned product features. The preliminary matching candidate set refers to the set of products that meet the core requirements selected from the standardized matching dataset.

[0049] In this invention, during the dimensional alignment of the customer demand graph and product features, a normalization algorithm can be used to map the customer demand graph and product features to the same feature space. A preset field mapping rule is used to perform semantic association matching on the mapped customer demand graph and product features. Based on the matching results, the field in the product feature corresponding to each field in the customer demand graph is determined, resulting in a standardized matching dataset.

[0050] Furthermore, each field in the customer demand graph is concatenated with the corresponding field in the product features to form a "demand-product" field combination. Based on a preset rule engine library, the "demand-product" field combination is validated sequentially to verify whether it meets the demand feature fields in the customer demand graph. The validated "demand-product" field combinations are used to generate a preliminary matching candidate set.

[0051] Furthermore, in the process of calculating the fuzzy matching degree of the preliminary matching candidate set, a deep learning model is introduced to extract deep semantic features from the "demand" field in the "demand-product" field combination and extract deep semantic features from the "product" field in the "demand-product" field combination. The basic matching score of the deep semantic features of the two is calculated by cosine similarity, that is, the basic matching score.

[0052] Furthermore, when optimizing the hierarchical analysis of the basic matching score, a multi-dimensional correction criterion is first introduced, and a correction coefficient is assigned to each criterion. At the same time, weights are assigned to each criterion and the basic matching score. Then, the weights of each criterion and the corresponding correction coefficients, as well as the weights of the basic matching score and the basic matching score, are fused together to calculate the matching degree.

[0053] For example, the multi-dimensional correction criteria include product compliance, customer historical purchase preferences, product supply stability, and product compliance status (the product meets regulatory requirements and has no compliance risks, with a product compliance correction coefficient of 1.05); customer historical purchase preferences status (the customer has purchased annuity insurance multiple times in the past 3 years, with a customer historical purchase preference correction coefficient of 1.03); and product stability status (the product has no recent sell-out risk and sufficient supply, with a product stability correction coefficient of 1.02 and a matching degree of 0.8715).

[0054] In this embodiment of the invention, by aligning the customer demand graph with the product feature dimension, the description dimensions of demand and product are unified, semantic differences are eliminated, and the problem of incompatibility and inability to directly compare the "demand-product" dimensions is solved. A rule engine is used to initially screen standardized matching datasets, quickly filtering obviously mismatched combinations, significantly narrowing the matching range, reducing subsequent computation, improving matching efficiency, and avoiding wasted computing power. Fuzzy matching degree calculation is performed on the initial matching candidate set to replace subjective judgment, making the matching results quantifiable and comparable. Through hierarchical analysis and optimization of the basic matching score, combined with multi-dimensional influencing factors, the basic score is corrected to compensate for the limitations of a single quantitative indicator, making the final matching degree more closely aligned with the customer's actual needs and business scenarios, thus improving the accuracy and practicality of the matching results.

[0055] S5. Generate a product recommendation scheme based on the matching degree and the preset customer risk preferences.

[0056] In this embodiment of the invention, generating a product recommendation scheme based on the matching degree and a preset customer risk preference includes: The preset customer risk preferences are converted into weight coefficients, and the weight coefficients and the matching degree are weighted and fused to obtain comprehensive evaluation data; A multi-objective optimization model is constructed based on the comprehensive evaluation data; The preset optimization objectives are solved using the multi-objective optimization model to generate a set of optimal solutions for the objectives. The optimal solution set is prioritized based on the customer's risk preference to obtain a product recommendation scheme.

[0057] In this embodiment of the invention, in the process of converting preset customer risk preferences into weighting coefficients, the types of customer risk preferences are first extracted, and the types are converted into weighting coefficients according to the preset conversion rule base.

[0058] In this embodiment of the invention, the step of converting the preset customer risk preference into a weighting coefficient includes: The preference type is determined based on the preset customer risk preferences; The initial weight coefficients are assigned to the preference types to obtain the initial weight coefficients; The initial weight coefficients are adjusted based on the multi-dimensional data to obtain the weight coefficients.

[0059] In this embodiment of the invention, a standardized risk preference type system is first preset, and core definition rules for each type are set; then, the collected preset customer risk preferences are compared with the type definition rules to match the corresponding preference type.

[0060] For example, a customer's risk preference is "only wanting to protect their funds, not accepting any loss of principal, and hoping to obtain a long-term stable cash flow through insurance," which perfectly matches the conservative type definition and is matched as conservative; a customer's risk preference is "hoping that insurance has both basic protection functions and can obtain certain dividends according to market conditions, and can accept small fluctuations in returns," which meets the core requirements of the stable type and is matched as stable.

[0061] Furthermore, in the process of assigning initial weight coefficients to preference types, firstly, a dimension importance ranking standard is set according to the preference type, and at the same time, dimensions are extracted from customer risk preferences. According to the dimension importance ranking standard, numerical ranges are set for dimensions of different priorities. Within the corresponding range, an initial weight value is assigned to each dimension, and the initial weight coefficients of the corresponding preference type are output.

[0062] For example, if the preference type is conservative, the dimensions extracted from the client's risk preference are risk level acceptance, liquidity needs, investment term suitability, and matching degree. The importance ranking criteria for these dimensions are: risk level acceptance (highest priority) > liquidity needs (second highest priority) > investment term suitability (medium priority) > matching degree (lowest priority). The numerical range for the highest priority is 0.35-0.5, the range for the second highest priority is 0.2-0.3, the range for the medium priority is 0.15-0.25, and the range for the lowest priority is 0.05-0.15. Therefore, the initial weight values ​​for risk level acceptance are 0.4, liquidity needs are 0.3, investment term suitability is 0.15, and matching degree is 0.15.

[0063] Furthermore, in the process of adjusting the initial weight coefficients based on multi-dimensional data, customer attribute data that is strongly correlated with the initial weight coefficients and can be quantified is first selected from the multi-dimensional data. According to the customer attribute data-weight adjustment rules, the initial weight values ​​of each dimension are adjusted to obtain the weight coefficients.

[0064] For example, for conservative clients: extract attribute data such as age 62, assets 500,000, and investment experience 1 year from multi-dimensional data, and adjust the initial weight coefficients according to the client attribute data-weight adjustment rules: increase risk level acceptance (0.4→0.48), liquidity needs (0.3→0.32), and investment period suitability (0.15→0.17), and decrease matching degree (0.15→0.03), and output the weight coefficients.

[0065] Furthermore, in the process of weighted fusion of weight coefficients and matching degree, the customer-product risk matching dimension score, customer-product liquidity matching dimension score, customer-product investment term matching dimension score, and customer-product overall matching dimension score are extracted from the matching degree. The above dimension scores are then weighted and calculated with weight coefficients to obtain comprehensive evaluation data.

[0066] Furthermore, in the process of constructing a multi-objective optimization model based on comprehensive evaluation data, binary decision variables are first defined. Based on the customer demand map and the preset optimization objectives, a multi-objective function is constructed with the binary decision variables as the core parameters. Subsequently, the preset business constraints (set according to the comprehensive evaluation data), such as budget constraints, recommendation quantity constraints, product mutual exclusion constraints, and core demand coverage constraints, are transformed into quantitative constraints. Finally, the binary decision variables, the three multi-objective functions, and the multiple quantitative constraints are integrated to form a multi-objective optimization model.

[0067] For example, assuming a client has a budget of 50,000 yuan, two core needs (critical illness + medical insurance), and three potential products (A: critical illness insurance, comprehensive assessment score of 90, low risk, cost 20,000 yuan; B: medical insurance, comprehensive assessment score of 85, low risk, cost 15,000 yuan; C: participating critical illness insurance, comprehensive assessment score of 88, medium risk, cost 30,000 yuan), and A and C are mutually exclusive (both belong to the critical illness category), then 1-2 products should be recommended. Define binary decision variables: (Recommendation A=1 / Not Recommended=0) (Recommendation B=1 / Not Recommended=0) (Recommendation C=1 / Not Recommended=0); Constructing a multi-objective function: 1. Maximizing the overall evaluation score ( ) 2. Minimize portfolio risk ( Transform into maximization ; 3. Maximum demand coverage ( The number of types of critical illnesses / medical treatments covered, such as Covering two categories ); Set quantitative constraints: 1. Budget constraint 2. ; 2. Quantity Constraint: 1 ; 3. Mutual Exclusion Constraint: ; 4. Coverage constraint: ; Integration to form a model: Satisfying the above constraints .

[0068] Furthermore, in the process of solving the preset optimization objective according to the multi-objective optimization model, the inheritance algorithm can be used to traverse all possible product combinations that meet the optimization objective conditions within the feasible domain defined by the quantitative constraints. The optimal product combination that shows complementary advantages in different objectives can be selected from the possible product combinations. The optimal product combination is then verified to ensure that it meets the multiple quantitative constraints. The product combinations that pass the verification are taken as the optimal solution set for the objective.

[0069] Furthermore, in the process of prioritizing the optimal solution set according to the customer's risk preferences, a specific ranking rule is first set according to the customer's risk preference type (e.g., conservative customers are ranked according to "minimize overall risk > maximize core needs coverage > maximize comprehensive suitability"). The product portfolios in the optimal solution set are ranked according to the specific ranking rule. Then, information such as recommendation reasons and product investment allocation suggestions are added to each priority portfolio to generate a product recommendation plan.

[0070] In this embodiment of the invention, by weighting and fusing weight coefficients and matching degree, the assessment data is made more in line with personalized needs, taking into account both customer risk preferences and demand adaptability. By constructing a multi-objective optimization model, the recommendation requirements are transformed into a standardized mathematical framework, taking into account the balance of multiple objectives and business constraints, ensuring that the recommendation is scientific and compliant. By solving the optimal solution set of the multi-objective optimization model, the optimal product combination at the algorithm level and in compliance with constraints is selected, providing a high-quality candidate set.

[0071] As can be seen, in the above solution, for the product recommendation solution business, multi-dimensional customer data and product terms and conditions are obtained, and a customer profile is constructed based on the multi-dimensional data; demand analysis is performed on the multi-dimensional data based on the customer profile to obtain a customer demand map; product features are extracted from the terms and conditions to obtain product features; the matching degree between the customer demand map and the product features is analyzed; and a product recommendation solution is generated based on the matching degree and preset customer risk preferences. By analyzing the demand from multi-dimensional data and generating a product recommendation solution based on the matching degree and customer risk preferences, the accuracy of product recommendations is improved while also meeting the core needs of customers.

[0072] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0073] In one embodiment, a product recommendation device based on customer needs is provided, which corresponds one-to-one with the product recommendation method based on customer needs described in the above embodiments. For example... Figure 3As shown, this product recommendation device based on customer needs includes a customer profile construction module 101, a demand graph analysis module 102, a product feature extraction module 103, a demand-product matching degree analysis module 104, and a recommendation scheme generation module 105. Detailed descriptions of each functional module are as follows: The customer profile building module 101 is used to obtain multi-dimensional customer data and product terms and conditions, and to build a customer profile based on the multi-dimensional data. The demand mapping analysis module 102 is used to perform demand analysis on the multi-dimensional data based on the customer profile to obtain a customer demand mapping. Product feature extraction module 103 is used to extract product features from the terms and conditions to obtain product features; The demand-product matching degree analysis module 104 is used to analyze the matching degree between the customer demand map and the product features; The recommendation scheme generation module 105 is used to generate product recommendation schemes based on the matching degree and preset customer risk preferences.

[0074] In one embodiment, the customer profile building module 101, when building a customer profile based on the multi-dimensional data, is used to: By linking and fusing the multi-dimensional data, the customer's overall data is obtained; Core behavioral features are extracted from the customer's overall data to generate core behavioral features; Construct a high-dimensional feature matrix based on the core behavioral characteristics; The high-dimensional feature matrix is ​​subjected to strong correlation feature filtering to obtain retained features; Customer profiles are constructed based on the retained features.

[0075] In one embodiment, when the demand mapping analysis module 102 performs demand analysis on the multi-dimensional data based on the customer profile to obtain a customer demand mapping, it is used to: The multi-dimensional data is classified and labeled based on the profile features of the customer profile to form classified and labeled data; The classification and labeling data are spatiotemporally correlated and encoded to obtain spatiotemporally correlated data; Perform demand feature analysis on spatiotemporal correlated data to generate demand features; A customer demand map is obtained by constructing a graph based on the aforementioned demand characteristics.

[0076] In one embodiment, when the product feature extraction module 103 extracts product features from the terms and conditions to obtain product features, it is used to: The aforementioned clause descriptions are preprocessed to obtain the standard clause text; The product feature dimensions are determined based on the standard clause text, and the standard clause text is matched based on the product feature dimensions to generate a set of matching texts. The core semantics of the matched text set are extracted to obtain the core semantics of the text, and preliminary product features are generated based on the core semantics of the text. Calculate the semantic similarity between the preliminary product features, and perform redundancy filtering on the preliminary product features based on the semantic similarity to obtain the product features.

[0077] In one embodiment, the demand-product matching degree analysis module 104, when analyzing the matching degree between the customer demand map and the product features, is used to: The customer demand map and the product features are dimensionally aligned to obtain a standardized matching dataset. The standardized matching dataset is initially screened by a rule engine to form a preliminary matching candidate set; The fuzzy matching degree is calculated for the preliminary matching candidate set to determine the basic matching score; The basic matching score is optimized through hierarchical analysis to obtain the matching degree.

[0078] In one embodiment, when generating a product recommendation scheme based on the matching degree and preset customer risk preferences, the recommendation scheme generation module 105 is used to: The preset customer risk preferences are converted into weight coefficients, and the weight coefficients and the matching degree are weighted and fused to obtain comprehensive evaluation data; A multi-objective optimization model is constructed based on the comprehensive evaluation data; The preset optimization objectives are solved using the multi-objective optimization model to generate a set of optimal solutions for the objectives. The optimal solution set is prioritized based on the customer's risk preference to obtain a product recommendation scheme.

[0079] In one embodiment, the recommendation generation module 105, when converting preset customer risk preferences into weighting coefficients, is used to: The preference type is determined based on the preset customer risk preferences; The initial weight coefficients are assigned to the preference types to obtain the initial weight coefficients; The initial weight coefficients are adjusted based on the multi-dimensional data to obtain the weight coefficients.

[0080] This invention provides a product recommendation device based on customer needs. For product recommendation solutions, it acquires multi-dimensional customer data and product terms and conditions, and constructs a customer profile based on the multi-dimensional data. It then performs demand analysis on the multi-dimensional data based on the customer profile to obtain a customer demand map. Next, it extracts product features from the terms and conditions to obtain product features. Finally, it analyzes the matching degree between the customer demand map and the product features. Based on the matching degree and preset customer risk preferences, it generates a product recommendation solution. By analyzing the demand from multi-dimensional data and generating a product recommendation solution based on the matching degree and customer risk preferences, it improves the accuracy of product recommendations while also aligning with core customer needs.

[0081] For specific limitations regarding a product recommendation device based on customer needs, please refer to the limitations of a product recommendation method based on customer needs described above, which will not be repeated here. Each module in the aforementioned product recommendation device based on customer needs can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0082] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements a server-side function or step of a product recommendation method based on customer needs.

[0083] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When executed by the processor, the computer program implements client-side functions or steps of a product recommendation method based on customer needs.

[0084] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain multi-dimensional customer data and product terms and conditions, and construct customer profiles based on the multi-dimensional data; Based on the customer profile, a demand analysis is performed on the multi-dimensional data to obtain a customer demand map. Product features are extracted from the aforementioned terms and conditions to obtain product features; Analyze the matching degree between the customer demand map and the product features; A product recommendation scheme is generated based on the matching degree and the preset customer risk preferences.

[0085] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Obtain multi-dimensional customer data and product terms and conditions, and construct customer profiles based on the multi-dimensional data; Based on the customer profile, a demand analysis is performed on the multi-dimensional data to obtain a customer demand map. Product features are extracted from the aforementioned terms and conditions to obtain product features; Analyze the matching degree between the customer demand map and the product features; A product recommendation scheme is generated based on the matching degree and the preset customer risk preferences.

[0086] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0087] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0088] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0089] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. If any software tools or components other than those of our company appear in the embodiments, they are merely illustrative examples and do not represent actual use. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A product recommendation method based on customer needs, characterized in that, include: Obtain multi-dimensional customer data and product terms and conditions, and construct customer profiles based on the multi-dimensional data; Based on the customer profile, a demand analysis is performed on the multi-dimensional data to obtain a customer demand map. Product features are extracted from the aforementioned terms and conditions to obtain product features; Analyze the matching degree between the customer demand map and the product features; A product recommendation scheme is generated based on the matching degree and the preset customer risk preferences.

2. The product recommendation method based on customer needs as described in claim 1, characterized in that, The process of constructing a customer profile based on the multi-dimensional data includes: By linking and fusing the multi-dimensional data, the customer's overall data is obtained; Core behavioral features are extracted from the customer's overall data to generate core behavioral features; Construct a high-dimensional feature matrix based on the core behavioral characteristics; The high-dimensional feature matrix is ​​subjected to strong correlation feature filtering to obtain retained features; Customer profiles are constructed based on the retained features.

3. The product recommendation method based on customer needs as described in claim 1, characterized in that, The step of performing demand analysis on the multi-dimensional data based on the customer profile to obtain a customer demand map includes: The multi-dimensional data is classified and labeled based on the profile features of the customer profile to form classified and labeled data; The classification and labeling data are spatiotemporally correlated and encoded to obtain spatiotemporally correlated data; Perform demand feature analysis on spatiotemporal correlation data to generate demand features; A customer demand map is obtained by constructing a graph based on the aforementioned demand characteristics.

4. The product recommendation method based on customer needs as described in claim 1, characterized in that, The process of extracting product features from the aforementioned terms and conditions to obtain product features includes: The aforementioned clause descriptions are preprocessed to obtain the standard clause text; The product feature dimensions are determined based on the standard clause text, and the standard clause text is matched based on the product feature dimensions to generate a set of matching texts. The core semantics of the matched text set are extracted to obtain the core semantics of the text, and preliminary product features are generated based on the core semantics of the text. Calculate the semantic similarity between the preliminary product features, and perform redundancy filtering on the preliminary product features based on the semantic similarity to obtain the product features.

5. The product recommendation method based on customer needs as described in claim 1, characterized in that, The analysis of the matching degree between the customer demand map and the product features includes: The customer demand map and the product features are dimensionally aligned to obtain a standardized matching dataset. The standardized matching dataset is initially screened by a rule engine to form a preliminary matching candidate set; The fuzzy matching degree is calculated for the preliminary matching candidate set to determine the basic matching score; The basic matching score is optimized through hierarchical analysis to obtain the matching degree.

6. The product recommendation method based on customer needs as described in claim 1, characterized in that, The step of generating a product recommendation scheme based on the matching degree and preset customer risk preferences includes: The preset customer risk preferences are converted into weight coefficients, and the weight coefficients and the matching degree are weighted and fused to obtain comprehensive evaluation data; A multi-objective optimization model is constructed based on the comprehensive evaluation data; The preset optimization objectives are solved using the multi-objective optimization model to generate a set of optimal solutions for the objectives. The optimal solution set is prioritized based on the customer's risk preference to obtain a product recommendation scheme.

7. The product recommendation method based on customer needs as described in claim 6, characterized in that, The step of converting preset customer risk preferences into weighting coefficients includes: The preference type is determined based on the preset customer risk preferences; The initial weight coefficients are assigned to the preference types to obtain the initial weight coefficients; The initial weight coefficients are adjusted based on the multi-dimensional data to obtain the weight coefficients.

8. A product recommendation device based on customer needs, characterized in that, include: The customer profile building module is used to obtain multi-dimensional customer data and product terms and conditions, and to build a customer profile based on the multi-dimensional data. The demand mapping analysis module is used to perform demand analysis on the multi-dimensional data based on the customer profile to obtain a customer demand mapping. The product feature extraction module is used to extract product features from the terms and conditions to obtain product features. The demand-product matching degree analysis module is used to analyze the matching degree between the customer demand map and the product features; The recommendation scheme generation module is used to generate product recommendation schemes based on the matching degree and preset customer risk preferences.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the product recommendation method based on customer needs as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the product recommendation method based on customer needs as described in any one of claims 1 to 7.

Images