Method, system and device for intelligently and automatically recommending car loan scheme and medium

Abstract

The invention relates to the field of automobile financial consumption, and provides a method, a system, equipment and a medium for intelligently and automatically recommending an automobile loan scheme. According to the method, a two-dimensional portrait is constructed by obtaining user data and fund party data, the user data is processed by feature crossing and a portrait algorithm to generate a user portrait, and the fund party data is processed by normalization and feature engineering to generate a fund party portrait. The method is advantaged in that the user and fund party portraits are matched based on the machine learning model, the matching degree is calculated, the personalized car loan scheme is generated according to matching degree sorting, interest rate, deadline and approval requirements are considered, accurate matching of the user and the fund party is realized, and recommendation success rate and user satisfaction are substantially improved. Meanwhile, the method supports dynamic adjustment of the model, a matching algorithm is optimized in real time through performance comparison of the challenger model and the champion model, efficient extension and multi-fund-party access are achieved, and the method has industrial popularization value.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automobile financial consumption, and in particular to a method, system, device and medium for intelligently and automatically recommending a car loan scheme. BACKGROUND

[0002] In the existing field of car loan scheme recommendation, although there are some attempts to improve the accuracy of recommendation by analyzing user data, such as the reference patent CN202211043233.0 discloses a car loan scheme recommendation method, device, electronic equipment and computer storage medium, which collects user data and uses a deep learning algorithm to construct a car loan recommendation model to improve the relevance of service recommendation.

[0003] However, this method mainly focuses on user services and lacks deep mining of specific needs of fund parties in the car loan field. When recommending a scheme to the user, the user's qualifications may not match the requirements of the fund party, resulting in the user being unable to implement the car loan according to the recommended scheme. In addition, the reference patent CN202210783935.6 proposes a car loan financial product scheme recommendation method and device, storage medium and electronic equipment, which provides a car loan recommendation scheme for users by analyzing the historical loan contracts of the vehicle, solving the problem that the system cannot recommend a car loan without user data. Although this method can solve the problem of missing user data, in the car loan field, especially in matching user and fund party needs, the algorithm model may not fully consider the lending needs of the fund party, such as different fund parties preferring different user types and different fund parties being affected differently by market interest rate fluctuations. This leads to a lack of dynamic response mechanism for market interest rate fluctuations and changes in fund party strategies. Existing solutions often only focus on a single dimension (user qualifications or vehicle information) and fail to build a comprehensive evaluation system for user qualifications, vehicle conditions, fund party preferences and market environment, making it difficult to cope with complex and changing market environments and diverse user needs. SUMMARY

[0004] In view of the above shortcomings of the prior art, the present application aims to provide a method and system for intelligently and automatically recommending a car loan scheme, which builds a model for comprehensive analysis of multi-dimensional data of users and fund parties and uses machine learning algorithms to achieve accurate matching to improve the accuracy, efficiency and effectiveness of car loan scheme recommendation.

[0005] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] In a first aspect, a method for intelligently and automatically recommending a car loan scheme is provided, comprising the following steps:

[0007] Obtaining user data and fund party data, the user data including basic information, credit score, asset status and consumption behavior of the user, and the fund party data including fund attribute, risk preference and business rule of the fund party;

[0008] Constructing a user portrait according to the user data, and constructing a fund party portrait according to the fund party data;

[0009] Inputting the user portrait and the fund party portrait into a car loan scheme matching model to obtain a matching degree of the user and the fund party, the car loan scheme matching model being a machine learning model trained based on a training sample set, the training sample set including historical user portraits, historical fund party portraits and historical matching result labels;

[0010] According to the matching degree of the user and the fund party, generating a recommended car loan scheme and pushing the recommended car loan scheme to the user.

[0011] In a possible design, the user portrait is constructed according to the user data, including:

[0012] Generating a multi-dimensional feature vector according to the user data, and performing feature cross to obtain a user feature vector;

[0013] Processing the user feature vector through a preset portrait construction algorithm to generate the user portrait.

[0014] In a possible design, the fund party portrait is constructed according to the fund party data, including:

[0015] Performing normalization processing on the fund party data to obtain a fund party feature vector;

[0016] Processing the fund party feature vector through a preset portrait construction algorithm to generate the fund party portrait.

[0017] In a possible design, the method further includes the following steps:

[0018] Dynamically adjusting the car loan scheme matching model.

[0019] In a possible design, the dynamically adjusting the car loan scheme matching model includes:

[0020] Taking the deployed car loan scheme matching model as a champion model, and taking other matching models generated in the training process as challenger models;

[0021] Calculating evaluation indexes of the challenger models and the champion model through the user data and the fund party data;

[0022] If at least one evaluation index of the challenger model exceeds that of the champion model, replacing the champion model with the challenger model.

[0023] In one possible design, the training method of the car loan scheme matching model includes the following steps:

[0024] An initial model of the car loan scheme matching model is constructed;

[0025] The training sample set is input into the initial model, and the optimal model parameter of the initial model is selected through loss parameter, cross-validation and hyperparameter optimization;

[0026] The initial model is simplified according to the optimal model parameter, and the car loan scheme matching model is obtained.

[0027] In one possible design, the generation method of the recommended car loan scheme includes:

[0028] The fund parties are sorted according to the matching degrees;

[0029] The recommended car loan scheme corresponding to the fund party with the highest matching degree is selected as the recommended car loan scheme matched with the user.

[0030] In a second aspect, a system for intelligently and automatically recommending a car loan scheme is provided, including:

[0031] A data acquisition module is configured to acquire user data and fund party data;

[0032] A portrait construction module is configured to construct a user portrait according to the user data and construct a fund party portrait according to the fund party data;

[0033] A matching module is configured to input the user portrait and the fund party portrait into a car loan scheme matching model to obtain a matching degree between the user and the fund party;

[0034] A recommendation module is configured to generate a recommended car loan scheme according to the matching degree and push the recommended car loan scheme to the user.

[0035] In a third aspect, the present application provides a computer device, including a memory, a processor and a transceiver which are sequentially communicatively connected, wherein the memory is configured to store a computer program, the transceiver is configured to receive and send messages, and the processor is configured to read the computer program and execute the method for intelligently and automatically recommending a car loan scheme provided in the first aspect.

[0036] In a fourth aspect, the present application provides a computer readable storage medium, and the computer readable storage medium stores instructions, when the instructions are run on a computer, the method for intelligently and automatically recommending a car loan scheme provided in the first aspect is executed.

[0037] The above-mentioned scheme has the following beneficial effects:

[0038] The application realizes the optimal matching of the supply and demand sides by calculating the matching degree between the portraits of the user and the fund party, and providing the fund party with the user meeting the risk preference of the fund party, so as to improve the success rate of providing the user with the car loan recommendation scheme, improve the satisfaction of the user and the fund party, and improve the actual feasibility of the recommendation scheme. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0040] Figure 1 A flowchart of a method for intelligently and automatically recommending a car loan scheme provided by the present application;

[0041] Figure 2 A flowchart of a training method of a car loan scheme matching model provided by the present application;

[0042] Figure 3 A system structure diagram of a method for intelligently and automatically recommending a car loan scheme provided by the present application;

[0043] Figure 4 A structure diagram of a computer device provided by the present application. DETAILED DESCRIPTION

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor. It should be noted that the description of these embodiment modes is used to help understand the present application, but does not constitute a limitation on the present application.

[0045] It should be understood that although the terms first and second, etc. may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another. For example, the first object can be called the second object, and similarly, the second object can be called the first object, without departing from the scope of the example embodiments of the present application.

[0046] It should be understood that, for the term "and / or" that can appear in the present text, it is only a description of the association relationship of the associated object, which means that there can be three relationships, for example, A and / or B can represent: A exists alone, B exists alone, or A and B exist at the same time, and so on; for example, A, B and / or C, can represent any one of A, B and C or any combination thereof; for the term " / and" that can appear in the present text, it is another description of the relationship of another associated object, which means that there can be two relationships, for example, A / and B, can represent: A exists alone or A and B exist at the same time; in addition, for the character " / " that can appear in the present text, it generally represents that the associated objects before and after are an "or" relationship.

[0047] Referring to Figure 1 , the first aspect of the present application provides a method for intelligently and automatically recommending a car loan scheme, which can be executed by a computer device with certain computing resources, such as a server, a personal computer (PC, which refers to a multi-purpose computer suitable for personal use in size, price and performance; desktop, notebook computer to small notebook computer and tablet computer and ultra-book, etc.), smart phone, personal digital assistant (PDA) or wearable device, etc. As shown in the figure, Figure 1 The method for intelligently and automatically recommending a car loan scheme includes but is not limited to the following steps:

[0048] S1, obtaining user data and fund data.

[0049] User data refers to information directly related to individual or enterprise users, which is used to understand user needs, provide personalized services, conduct business interactions and risk management, and is multi-dimensional data. Specifically, users can be divided into individual users or enterprise users, and the user data of individual users can include,

[0050] Personal information: including name, ID number, contact information (phone, email, address), date of birth, gender, etc.

[0051] Account information: account number, password, login record, security authentication information (such as verification code, biometric data).

[0052] Financial information: income level, asset condition, debt situation, credit score, investment preference.

[0053] Transaction records: historical transaction data (purchase, payment, refund, service usage records), consumption habits, behavior tracks.

[0054] Preference and behavior data: product preferences, service choices, interaction history (such as customer service communication records), satisfaction feedback.

[0055] Device and location information: device ID, IP address, geographic location (for contextualized services or risk control).

[0056] Sensitive information: health data (such as insurance users), professional information, educational background, etc.

[0057] User data of enterprise users can include:

[0058] Basic information of enterprises: enterprise name, registration information, business license, legal person information, industry type.

[0059] Business data: business scope, revenue data, supply chain information, partner list.

[0060] Financial data: corporate account information, financial statements, tax records, credit rating.

[0061] Transactions and contracts: procurement records, cooperation contracts, service agreements, payment terms.

[0062] Operational data: frequency of using platform functions by users, business pain points, technical support needs.

[0063] Funding party data refers to information related to funding providers (such as banks, investment institutions, financial institutions) for funding management, risk assessment, compliance review, and transaction settlement, which is also multi-dimensional data. Specifically, funding party data usually includes the following categories:

[0064] Fund source and structure data:

[0065] Fund type (own funds, loan funds, investment funds), fund source institution information (name, qualification).

[0066] Fund size, cost (interest rate, fee), term, liquidity requirement.

[0067] Fund use restrictions (such as designated investment fields, prohibition of investment in high-risk industries).

[0068] Fund flow and transaction data:

[0069] Fund transfer records (lending, repayment, investment, dividend), counterparty information.

[0070] Settlement path of funds (bank account, payment channel, clearing institution), monitoring of fund flow.

[0071] Income and risk data: expected yield, actual yield, risk exposure (such as leverage ratio, default probability).

[0072] Risk assessment and compliance data:

[0073] Credit rating, financial condition, historical default record of the capital party.

[0074] Compliance information: regulatory qualifications (license, record information), review data for abnormal situations.

[0075] Investment strategy and restrictions: investment geographical restrictions, industry preferences, risk control indicators (such as maximum loss threshold).

[0076] S2, constructing a user portrait according to user data, and constructing a capital party portrait according to capital party data.

[0077] The user portrait is obtained by collecting, integrating and analyzing multi-source data of the user, and converting the abstract user into a specific and quantifiable feature model. The core goal is to describe the individual or group characteristics of the user in a labeled manner, and the capital party portrait is a multi-dimensional feature description of financial institutions, investment institutions and capital providers, focusing on analyzing their capital attributes, investment preferences, risk tolerance and compliance requirements. Preferably, according to the user data, the user portrait is constructed, and the capital party portrait is constructed according to the capital party data, which includes the following steps:

[0078] S201, generating a multi-dimensional feature vector according to user data, and performing feature cross to obtain a user feature vector.

[0079] In the step S201, the original data of each dimension of the user is converted into a numerical feature vector. Each feature represents a dimension or sub-dimension, forming an initial feature matrix. The core of feature cross is to combine different features to capture the interaction between features, generate new composite features, and improve the model's expression ability for complex relationships. Further, if the user data generates a large number of new features after feature cross, resulting in dimension disaster, feature importance screening, PCA / factor analysis or regularization can be used for optimization.

[0080] S202, processing the user feature vector by a preset portrait construction algorithm to generate a user portrait.

[0081] In the step S202, the preset portrait algorithm can be a clustering analysis algorithm or a neural network algorithm to generate a user portrait, and the user portrait is labeled, including but not limited to labels such as risk level, preference type, etc.

[0082] S203, normalizing the capital party data to obtain a capital party feature vector.

[0083] The purpose of normalization is to convert different scales and dimensional fund data into a unified range, eliminate the magnitude difference between features, and thus improve the effectiveness and stability of subsequent algorithms (such as clustering, classification, and similarity calculation). According to different fund data, different normalization methods can be selected:

[0084] For continuous numerical features, such as fund size, interest rate, and other feature data, Min-Max normalization can be used to scale the data to the range [0, 1]; standardization (Z-score) can be used to convert the feature to a distribution with a mean of 0 and a standard deviation of 1, or logarithmic transformation can be used to process long-tailed distribution data (such as when the fund size difference is extremely large).

[0085] For discrete or categorical features, such as industry preference, institutional type, and other features, one-hot encoding can be used to convert categories into binary vectors, such as "industry = technology" encoded as [0, 1, 0, 0].

[0086] S204, processing the fund feature vector through a preset portrait construction algorithm to generate a fund portrait.

[0087] S3, inputting the user portrait and fund portrait into a car loan scheme matching model to obtain the matching degree of the user and the fund.

[0088] In the step S3, the car loan scheme matching model is a machine learning model trained based on a training sample set, and the training sample set includes historical user portraits, historical fund portraits, and historical matching result labels. Specifically, the machine learning model includes a logistic regression model, a decision tree model, and a neural network model. The historical user portrait includes the feature vector of the past car loan applicant, such as credit score, income level, debt ratio, car purchase budget, repayment ability, risk preference, etc. The historical fund portrait includes the feature vector of the corresponding fund (such as bank, financial institution), such as interest rate range, risk tolerance, approval condition, fund liquidity, compliance requirement, etc. The historical matching result label can be a binary label or a continuous score label, which is labeled based on the actual loan success or failure, user satisfaction, and default rate.

[0089] Referring to Figure 2 , S301-S303 are the training process of the car loan scheme matching model. Among them:

[0090] S301, constructing an initial model of the car loan scheme matching model.

[0091] In step S301, the initial model of the car loan scheme matching model can be selected according to the label of the training sample set, for example, a classification model (such as logistic regression, support vector machine, random forest) is suitable for a binary matching result label; a regression model (such as linear regression, neural network) is suitable for a continuous matching degree score label; a recommendation system algorithm (such as matrix decomposition, collaborative filtering) is suitable for user-fund interaction history optimized matching; a deep learning model (such as DNN, Transformer): suitable for processing a large amount of high-dimensional, nonlinear relationship, but requires a large amount of data support.

[0092] After determining the type of the initial model, the model weight and bias are randomly initialized or a pre-trained model is used.

[0093] S302, input the training sample set into the initial model, and select the optimal model parameter of the initial model through loss parameter, cross-validation and hyperparameter optimization.

[0094] In step S302, the training sample set can be divided into a training set, a validation set and a test set (such as 80% training, 10% validation and 10% testing). The preprocessed user portrait feature vector and the fund portrait feature vector are spliced into the model input, and the label is the historical matching result (such as a binary matching result or a matching degree score). The model calculates the predicted result (such as matching probability) according to the input.

[0095] The loss function is used to measure the difference between the prediction and the true label.

[0096] The model parameters are optimized by gradient descent method, and after minimizing the loss function, the parameters (such as weight regularization coefficient) in the loss function are adjusted, the model performance is evaluated on the validation set, and the parameters are adjusted to improve the generalization ability and balance the model complexity and the generalization ability.

[0097] Cross-validation is used to evaluate the stability of the model and avoid overfitting.

[0098] The training set is divided into K folds, one fold is selected as the validation set in turn, the remaining K-1 folds are used to train the model, and the process is repeated K times. The average performance index (such as average accuracy, average F1-score) is calculated as a robust estimate of the model performance, so as to determine the optimal combination of hyperparameters.

[0099] Hyperparameter optimization is used to select the optimal hyperparameters (such as learning rate, regularization strength, number of neural network layers) of the model. Specifically, grid search, random search or Bayesian optimization can be used for hyperparameter optimization to find the optimal combination of hyperparameters.

[0100] S303, simplify the initial model according to the optimal model parameter to obtain the car loan scheme matching model.

[0101] In step S303, the initial model is simplified to improve the inference speed of the model, reduce the complexity of the model, further prevent overfitting, and improve the explainability. A simpler model (such as a decision tree model) is used to replace the deep model.

[0102] S4, according to the matching degree of the user and the fund party, a recommended car loan scheme is generated and pushed to the user.

[0103] In step S4, first, based on the matching degree score output by the model, the fund party meeting the conditions is screened:

[0104] The matching degree threshold is set, and the low matching degree fund party is filtered according to the matching degree threshold, and the options with scores lower than the threshold are excluded to reduce the user's selection burden. For example, a matching degree of ≥80 points is considered as a recommended candidate, and if there are multiple high matching degree fund parties, multiple alternative schemes are generated for the user to choose from.

[0105] Combined with the user's demand and the fund party's conditions, specific scheme details are generated. The scheme details can include the loan amount, the interest rate range based on the fund party's portrait, the selected optimal interest rate, the matching user's expected term and the fund party's available term, equal principal and interest, equal principal or other customized repayment methods, the fund party's approval requirements (such as required materials, credit score threshold) and additional terms. For example, the matching degree of user A and fund party B is 90 points, fund party B provides a loan product with an interest rate of 7%, a term of 3 years, and a loan amount of up to 300,000. The scheme for user A is as follows: loan 200,000, annual interest rate 7%, monthly payment ≈6,000 yuan, 3 years to pay off, need to submit income proof and credit report.

[0106] The fund party's scheme can be the fund party's preset scheme, or a personalized scheme adjusted according to the user's demand.

[0107] Further, the system can perform secondary scoring on multiple alternative schemes that meet the conditions after screening, considering additional factors including interest rate discount level, approval speed, user historical preferences, etc.

[0108] For example, after calculating the user's matching degree, the system screens out three high matching degree fund parties A, B, and C.

[0109] Generate corresponding schemes A (interest rate 7%, 3 years), B (interest rate 7.5%, 2 years), and C (interest rate 6.5%, high down payment required).

[0110] The system preferentially recommends scheme C (because the interest rate is the lowest and the matching degree is high), and marks it as "best recommendation".

[0111] Push a notification to the user's APP: "You have 1 car loan scheme recommendation, interest rate 6.5%, click to apply immediately".

[0112] User views details, can switch to scheme B.

[0113] S5, dynamically adjusting the car loan scheme matching model.

[0114] In step S5, the car loan scheme matching model is dynamically adjusted to ensure that the model can continuously adapt to market changes, user behavior updates or the access of new funders. For example, when the central bank adjusts the benchmark interest rate, the model automatically updates the funders' interest rate parameter range and recalculates the matching degree; or, a new funder C (prefer high-risk users) is added, and the model learns its preference features through a small number of cases and adjusts the matching logic; or, it is found that users are more concerned about the approval speed than the interest rate in the near future, and the model weight adjustment improves the priority of the approval process complexity feature. Specifically, in the present application, the car loan scheme matching model can be dynamically adjusted by the following methods:

[0115] The deployed car loan scheme matching model is used as the champion model, and other matching models generated during the training process are used as challenger models.

[0116] The evaluation indicators of the challenger model and the champion model are calculated through user data and funder data.

[0117] If at least one evaluation indicator of the challenger model exceeds that of the champion model, the challenger model replaces the champion model.

[0118] Wherein, the champion model refers to the main model currently deployed in the production line, which is responsible for real-time processing of user requests and generating car loan scheme recommendations; the challenger model refers to the candidate model generated during the training process (such as different algorithms, parameter configurations or feature combinations), which needs to be evaluated to determine whether its performance is better than that of the champion model. Before determining whether the challenger model can replace the champion model, the evaluation indicator system of the model needs to be determined. Exemplarily, the core indicators can be selected according to the business objectives, the auxiliary indicators of the model's explainability, robustness and fairness are considered, and the thresholds of the indicators are set. The core indicators include: matching accuracy (such as the actual approval rate of the recommended scheme), user satisfaction (such as the acceptance rate or feedback score of the user to the scheme), business income related indicators (such as the matching success rate of the funder, the loan conversion rate, the risk default rate), efficiency indicators (such as inference delay, resource consumption). The replacement decision of the auxiliary indicators can be determined based on the core indicators. According to the threshold setting, the baseline of each indicator is set, for example, when the accuracy needs to be improved by ≥2%, the replacement is triggered.

[0119] For model evaluation, offline evaluation or online evaluation can be adopted, the main difference being the evaluation flow used, specifically, offline evaluation uses historical data set (test set is divided) or simulation data to evaluate the challenger model. Online evaluation shunts part of the real-time flow to the challenger model through shadow mode, and compares the actual effect. In order to ensure the stability of the model, the evaluation period can be set to trigger evaluation at fixed periods (such as daily, weekly), or based on data volume threshold (such as after adding X million new user data).

[0120] The indicators are calculated for the challenger model and the champion model, for example:

[0121] The same user and fund data are input, and the accuracy of the output matching degree score is compared; or, the approval pass rate, user click rate, etc. of the recommended scheme are counted.

[0122] If a key indicator of the challenger model significantly exceeds the champion model (such as accuracy improvement ≥ threshold), replacement is triggered; or, there are multiple indicators, a weighted scoring mechanism can be used to comprehensively evaluate whether the challenger model is overall better. The weight of each indicator is determined by business demand, if the comprehensive score exceeds the champion model, replacement is performed. It should be noted that the challenger model should not be inferior to the champion model in key risk indicators (such as default rate), to avoid risk increase caused by replacement.

[0123] The present application proposes a method and system for intelligently and automatically recommending car loan schemes, by constructing a dynamic matching model of users and fund parties, precise recommendation and efficient resource allocation are achieved. Compared with traditional recommendation methods, the specific technical effects of the present application are as follows:

[0124] By constructing dynamic user portraits and fund party portraits, the car loan scheme matching model can deeply analyze user credit, asset and consumption behavior characteristics, and fund party risk preference and business rules, and the matching accuracy is improved by more than 20% compared with traditional methods. The real-time matching recommendation system shortens the response time to milliseconds, supports high-concurrency scenarios, and improves the efficiency of car loan scheme approval by 3-5 times. Based on the matching degree ranking recommendation mechanism, high-matching-degree fund party schemes are preferentially pushed, which can effectively improve the user scheme acceptance rate and loan conversion rate, at the same time, reduce the workload of manual screening, and reduce the operating cost by 30%.

[0125] Support fund strategy real-time update (such as interest rate adjustment, approval rule change), fund portrait dynamic refresh cycle is shortened to hours. The built-in compliance verification module ensures that the recommended scheme meets the regulatory requirements (such as interest rate cap, risk control index), and the compliance review pass rate is increased to 98%. The modular system architecture of the application supports rapid expansion to other financial products (such as housing loans, consumer loans), and the development cost is reduced. For example, when expanding from car loans to business loans, only industry data needs to be supplemented and model parameters need to be fine-tuned. At the same time, it is compatible with multiple fund access, supports small and medium-sized banks and financial institutions to quickly build a car loan recommendation system, and the deployment time is reduced by 60%.

[0126] The car loan scheme matching model can provide "feature weight-matching logic-fund sorting" reasoning chain visualization, and business personnel can intuitively understand the recommendation basis, and further support personalized scheme explanation, and enhance user trust.

[0127] Therefore, based on the foregoing steps S1-S5, the method for intelligently and automatically recommending a car loan scheme, through the intelligent matching model, dynamic portrait construction and modular system design, realizes precision, efficiency and intelligence in the car loan recommendation scenario, significantly improves the matching efficiency and satisfaction of financial institutions and users, and has wide industry promotion value.

[0128] As shown in Figure 3 The second aspect of the embodiment provides a virtual system for implementing the method for intelligently and automatically recommending a car loan scheme according to the first aspect, comprising a data acquisition module, a portrait construction module, a matching module and a recommendation module.

[0129] The data acquisition module is configured to acquire user data and fund data.

[0130] The portrait construction module is configured to construct a user portrait according to the user data, and construct a fund portrait according to the fund data.

[0131] The matching module is configured to input the user portrait and the fund portrait into a car loan scheme matching model to obtain a matching degree of the user and the fund.

[0132] The recommendation module is configured to generate a recommended car loan scheme according to the matching degree and push it to the user.

[0133] The working process, working details and technical effects of the foregoing device provided by the second aspect of the embodiment can be referred to the method for intelligently and automatically recommending a car loan scheme according to the first aspect, which will not be repeated here.

[0134] As shown in Figure 4As shown, the third aspect of the present embodiment provides a computer device for performing the method of intelligently and automatically recommending a car loan scheme according to the first aspect, which comprises a storage module, a processing module and a transceiver module connected in sequence, wherein the storage module is configured to store a computer program, the transceiver module is configured to transceive messages, and the processing module is configured to read the computer program and perform the method of constructing a financial risk control and collection behavior scorecard according to the first aspect. Specifically, the storage module can include, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a flash memory, a first-in first-out memory (FIFO) and / or a first-in last-out memory (FILO), etc.; and the processing module can use a microprocessor of the STM32F105 series, but is not limited thereto. In addition, the computer device can further include a power module, a display screen and other necessary components, but is not limited thereto.

[0135] The working process, working details and technical effects of the aforementioned computer device provided by the third aspect of the present embodiment can be referred to the method of intelligently and automatically recommending a car loan scheme according to the first aspect, which will not be repeated here.

[0136] The fourth aspect of the present embodiment provides a computer readable storage medium storing instructions of the method of intelligently and automatically recommending a car loan scheme according to the first aspect, i.e., the computer readable storage medium stores instructions, and when the instructions are run on a computer, the method of constructing a financial risk control and collection behavior scorecard according to the first aspect is performed. The computer readable storage medium refers to a carrier for storing data, which can include, but is not limited to, floppy disks, optical disks, hard disks, flash memories, USB flash drives and / or memory sticks, etc., and the computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices.

[0137] The working process, working details and technical effects of the aforementioned computer readable storage medium provided by the fourth aspect of the present embodiment can be referred to the method of intelligently and automatically recommending a car loan scheme according to the first aspect, which will not be repeated here.

[0138] Finally, it should be noted that: the above only describes the preferred embodiments of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for intelligently and automatically recommending a car loan scheme, characterized in that, The method comprises the following steps: obtaining user data and funding party data, wherein the user data comprises basic information, credit score, asset status and consumption behavior of the user, and the funding party data comprises funding attribute, risk preference and business rules of the funding party; constructing a user portrait according to the user data and constructing a funding party portrait according to the funding party data; inputting the user portrait and the funding party portrait into a car loan scheme matching model to obtain a matching degree of the user and the funding party, wherein the car loan scheme matching model is a machine learning model trained based on a training sample set, and the training sample set comprises historical user portraits, historical funding party portraits and historical matching result labels; generating a recommended car loan scheme according to the matching degree of the user and the funding party and pushing the recommended car loan scheme to the user.

2. The method of claim 1, wherein, The method of constructing the user portrait according to the user data comprises: generating a multi-dimensional feature vector according to the user data and performing feature cross processing to obtain a user feature vector; processing the user feature vector through a preset portrait construction algorithm to generate the user portrait.

3. The method of claim 1, wherein, The method of constructing the funding party portrait according to the funding party data comprises: performing normalization processing on the funding party data to obtain a funding party feature vector; processing the funding party feature vector through a preset portrait construction algorithm to generate the funding party portrait.

4. The method of claim 1, wherein, The method further comprises the following steps: dynamically adjusting the car loan scheme matching model.

5. The method of claim 4, wherein, The method of dynamically adjusting the car loan scheme matching model comprises: taking the deployed car loan scheme matching model as a champion model and taking other matching models generated in the training process as challenger models; calculating evaluation indexes of the challenger models and the champion model through the user data and the funding party data; if at least one of the evaluation indexes of the challenger model exceeds that of the champion model, replacing the champion model with the challenger model.

6. The method of claim 1, wherein, The training method of the car loan scheme matching model comprises the following steps: constructing an initial model of the car loan scheme matching model; inputting the training sample set into the initial model, selecting optimal model parameters of the initial model through loss parameters, cross-validation and hyperparameter optimization; simplifying the initial model according to the optimal model parameters to obtain the car loan scheme matching model.

7. The method of claim 1, wherein, The method of generating the recommended car loan scheme comprises: sorting the funding parties according to the matching degrees; selecting a recommended car loan scheme corresponding to a funding party with the highest matching degree as a recommended car loan scheme matched with the user.

8. A system for intelligently recommending a car loan scheme, characterized in that, The method comprises: a data acquisition module configured to acquire user data and funding party data; a portrait construction module configured to construct a user portrait according to the user data and construct a funding party portrait according to the funding party data; a matching module configured to input the user portrait and the funding party portrait into a car loan scheme matching model to obtain a matching degree of the user and the funding party; a recommendation module configured to generate a recommended car loan scheme according to the matching degree and push the recommended car loan scheme to the user.

9. A computer device, comprising: The application relates to a computer device comprising a memory, a processor and a transceiver connected in sequence, wherein the memory is used for storing a computer program, the transceiver is used for transmitting and receiving messages, and the processor is used for reading the computer program and executing the intelligent automatic recommendation vehicle loan scheme method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The application relates to a computer device comprising a memory, a processor and a transceiver connected in sequence, wherein the memory is used for storing a computer program, the transceiver is used for transmitting and receiving messages, and the processor is used for reading the computer program and executing the intelligent automatic recommendation vehicle loan scheme method as claimed in any one of claims 1 to 7.

Images