Intelligent order reporting methods, devices, equipment and storage media for commercial vehicles

Abstract

This application provides a method, apparatus, device, and storage medium for intelligent commercial vehicle order processing, relating to the fields of intelligent finance and insurance. The method generates a recommendation strategy by acquiring multi-source business data, including user input information, credit information, and policy information. The recommendation strategy includes credit risk constraints obtained through credit risk assessment using machine learning algorithms combined with multi-source business data, and policy constraints obtained by parsing policy information. Following the recommendation strategy, user input information is used to match commercial vehicle product combinations, and then recommendation information is generated based on these combinations. By integrating multi-source business data and combining AI algorithms, this method can output personalized commercial vehicle product recommendations and complete credit limit assessments, thereby improving the efficiency and personalization of commercial vehicle order processing.

Description

Technical Field

[0001] This application relates to the field of intelligent financial technology, and in particular to a method, apparatus, device and storage medium for intelligent order placement for commercial vehicles. Background Technology

[0002] In the financial and insurance sectors, commercial vehicles are automobiles used for passenger and freight transportation or leasing, generating direct or indirect revenue from freight or rental fees. Commercial vehicles can include trucks, buses, special-purpose vehicles, and rental vehicles. Different commercial vehicles can have different profit characteristics, corresponding to different commercial vehicle service invoices. These service invoices directly affect the determination of vehicle insurance premiums, loan risk assessments, and the rights and obligations stipulated in insurance contracts.

[0003] With the rapid development of the commercial vehicle market, users face increasingly more choices and complex decision-making processes when selecting commercial vehicles and related financial products. To address this, a recommendation system can be built to suggest suitable commercial vehicle-related products during the user's decision-making process. This system can rely on manual analysis or a simple rule engine. After receiving user input regarding their commercial vehicle business needs, it analyzes user preferences based on the input information and then recommends relevant products accordingly.

[0004] However, order recommendation systems that rely on manual analysis and rule engines suffer from low output efficiency and struggle to meet personalized needs. Furthermore, because these systems rely on limited input information, they cannot fully utilize multi-dimensional data to provide accurate product recommendations, thus degrading the user experience. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method, apparatus, device, and storage medium for intelligent order placement in commercial vehicles to solve the problems of low output efficiency and low personalization in order placement recommendation systems.

[0006] According to a first aspect of this application, a method for intelligent order reporting for commercial vehicles is provided, the method comprising: Acquire multi-source business data, including user input information, credit information, and policy information; A recommendation strategy is generated based on the multi-source business data. The recommendation strategy includes credit risk constraints and policy constraints. The credit risk constraints are constraints obtained by using machine learning algorithms to perform credit risk assessment in conjunction with the multi-source business data. The policy constraints are constraints obtained by parsing the policy information. According to the recommendation strategy, the user input information is used to match a commercial vehicle product portfolio, which includes at least one of commercial vehicle loan products, commercial vehicle insurance products, and commercial vehicle after-sales service products. Recommendation information is generated based on the aforementioned commercial vehicle product portfolio. The recommendation information includes commercial vehicle order information and decision-making suggestions.

[0007] According to a second aspect of this application, a smart order reporting device for commercial vehicles is provided, the device comprising: The multi-source data integration module is used to acquire multi-source business data, including user input information, credit information, and policy information. The strategy generation module is used to generate a recommendation strategy based on the multi-source business data. The recommendation strategy includes credit risk constraints and policy constraints. The credit risk constraints are constraints obtained by using machine learning algorithms to perform credit risk assessment in conjunction with the multi-source business data. The policy constraints are constraints obtained by parsing the policy information. The product matching module is used to match a commercial vehicle product combination with the user input information according to the recommendation strategy. The commercial vehicle product combination includes at least one of commercial vehicle loan products, commercial vehicle insurance products, and commercial vehicle after-sales service products. The personalized recommendation algorithm module is used to generate recommendation information based on the commercial vehicle product portfolio. The recommendation information includes commercial vehicle order information and decision-making suggestions.

[0008] According to a third aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described intelligent order reporting method for commercial vehicles.

[0009] According to a fourth aspect of this application, a storage medium is provided that stores a computer program thereon, which, when executed by a processor, implements the above-described intelligent order reporting method for commercial vehicles.

[0010] By employing the above technical solutions, embodiments of this application provide a method, apparatus, device, and storage medium for intelligent commercial vehicle order placement. The method generates a recommendation strategy by acquiring multi-source business data, including user input information, credit information, and policy information. The recommendation strategy includes credit risk constraints obtained through credit risk assessment using machine learning algorithms combined with multi-source business data, and policy constraints obtained by parsing policy information. Then, according to the recommendation strategy, user input information is used to match commercial vehicle product combinations, and recommendation information is generated based on these combinations. By integrating multi-source business data and combining AI algorithms, this method can output personalized commercial vehicle product recommendation information and complete credit limit assessment, thereby improving the output efficiency and personalization of commercial vehicle order placement.

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic flowchart of the intelligent order reporting method for commercial vehicles provided in an embodiment of this application; Figure 2 This is a schematic diagram of the overall process for submitting a commercial vehicle order in an embodiment of this application; Figure 3 A schematic diagram of the credit risk assessment process provided in this application embodiment; Figure 4 This is a schematic diagram illustrating the dynamic adjustment of the matching process based on multi-turn dialogue, provided as an embodiment of this application. Figure 5 This is a schematic diagram illustrating the process of updating policy constraints based on real-time policies, provided as an embodiment of this application. Figure 6 A schematic diagram of the structure of the intelligent order reporting device for commercial vehicles provided in this application embodiment. Detailed Implementation

[0013] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0014] In this embodiment, a commercial vehicle is a car that collects freight or rental fees directly or indirectly and can be used for passenger and freight transportation or leasing. Commercial vehicles may include types such as freight trucks, buses, special-purpose vehicles, and rental vehicles.

[0015] For example, freight vehicles can include ordinary trucks, tractor-trailers, dump trucks, and box trucks, used for logistics transportation. Passenger vehicles are those with nine or more seats used for passenger transport, such as city buses, long-distance passenger buses, and tourist buses. Special-purpose vehicles are those directly involved in engineering construction or professional services and generate profit from them, such as concrete mixer trucks, water trucks, tanker trucks, and cranes. Rental vehicles, regardless of size or model, are also included in the category of commercial vehicles because they generate profit through rental fees.

[0016] For commercial vehicle businesses, different commercial vehicles can have different profit characteristics, corresponding to different commercial vehicle service invoices. Commercial vehicle service invoices refer to service invoices generated during the commercial vehicle business, such as insurance invoices, loan invoices, and after-sales service invoices. Commercial vehicle service invoices can be used to present relevant information about the commercial vehicle business at multiple business stages in which the user participates. Therefore, commercial vehicle service invoices directly affect the determination of vehicle insurance premiums, loan risk assessment, and the rights and obligations in insurance contracts.

[0017] With the rapid development of the commercial vehicle market, users face increasingly more choices and complex decision-making processes when selecting commercial vehicles and related financial products. In some embodiments, a service order recommendation system can be built to recommend suitable commercial vehicle-related products for user reference during the decision-making process. This system can rely on manual analysis or a simple rule engine. After receiving user input regarding their commercial vehicle service requirements, it analyzes user preferences based on the input information and then recommends relevant products accordingly.

[0018] However, order recommendation systems that rely on manual analysis and rule engines suffer from low output efficiency and struggle to meet personalized needs. Furthermore, because these systems rely on limited input information, they cannot fully utilize multi-dimensional data to provide accurate product recommendations, thus degrading the user experience.

[0019] To address the issues of low output efficiency and personalization in order recommendation systems, this application provides a smart order reporting method for commercial vehicles in some embodiments. This method integrates multi-source business data and combines it with AI algorithms to output personalized commercial vehicle product recommendation information and complete credit limit assessment, thereby improving the output efficiency and personalization of commercial vehicle order reporting.

[0020] The method can be applied to a commercial vehicle service platform, or to an electronic device that establishes a communication connection with the commercial vehicle service platform and has data processing capabilities. The electronic device includes, but is not limited to, computers, servers, mobile terminals, smart wearable devices, and industrial control computers. For ease of description, this application embodiment uses a commercial vehicle service platform as the executing entity of the method. It should be understood that the method can also be applied to other types of executing entities, which are not shown one by one in this application embodiment. Figure 1 As shown, the method includes: S101. Obtain multi-source business data.

[0021] In order to output commercial vehicle order forms, the commercial vehicle service platform can first obtain multi-source business data when executing commercial vehicle-related business. The multi-source business data includes user input information, credit information, and policy information.

[0022] User input information refers to relevant information obtained by the commercial vehicle service platform through user interaction interfaces, such as user interaction commands and user interaction data. Credit information refers to relevant information reflecting a user's current credit status, such as the user's credit record and repayment history. Policy information refers to macro and local regulations, financial incentives, and industry standards related to the entire lifecycle of commercial vehicles such as trucks, tractors, and special-purpose vehicles, including their purchase, operation, replacement, and scrapping. Specific policy information includes, but is not limited to, regional vehicle purchase subsidy policies, tax and fee reductions, and financial support policies, such as subsidies for new energy commercial vehicles, subsidies for scrapping and replacement or trade-in programs, and support policies such as vehicle and vessel tax and purchase tax reductions, toll discounts, and loan interest subsidies.

[0023] like Figure 2 As shown, different methods can be used to acquire different information from multi-source business data. For example, a commercial vehicle service platform can have a built-in multi-source data integration module, which includes input information processing, credit information acquisition, and policy information acquisition units. When acquiring multi-source business data, the commercial vehicle service platform can call the multi-source data integration module and obtain user input information through the input information processing unit, then parse user needs from the input information. Simultaneously, through the credit information acquisition unit, it integrates data from the credit system to obtain user credit records, repayment history, and other information. Furthermore, through the policy information acquisition unit, it obtains and parses the latest commercial vehicle-related policies and subsidy information in real time.

[0024] The user input information can include product requirements, which can trigger recommendations for commercial vehicle-related products. The commercial vehicle service platform supports multiple interaction methods, allowing users to input information through one or more combinations of these methods.

[0025] In some embodiments, if the commercial vehicle service platform supports text input, then when acquiring multi-source business data, the commercial vehicle service platform can first display an information input interface, which may include text input controls. During the display of the information input interface, the commercial vehicle service platform can also monitor the text information entered by the user through the text input controls in real time, and based on natural language processing technology, perform preprocessing, word segmentation, and intent recognition on the text information to obtain the user input information.

[0026] For example, if a user enters a text message such as "I want to rent a 9-seater MPV, please calculate the rental price" into the customer service window of the commercial vehicle business interface using an input device like a keyboard, the commercial vehicle service platform can obtain the inquiry text by listening to the text content in the text input box within the customer service window. Then, it can use a natural language processing model to analyze the intent of the inquiry text, thereby obtaining the user's input information. Accordingly, the user input information includes features such as the commercial vehicle business type "rental," the vehicle type "MPV," and the business requirement "price inquiry."

[0027] Because different interaction methods input different types of information, commercial vehicle service platforms need to employ different information acquisition and processing algorithms to obtain user input information. In some embodiments, the commercial vehicle service platform can have a built-in intelligent voice interaction module for voice interaction. When acquiring multi-source business data, it can first acquire the user's voice input and convert it into user-input text. The user-input text is a word segmentation set containing multiple keywords.

[0028] Next, user demand features are extracted from the user input text, wherein the user demand features are word vectors constructed from keywords associated with at least one commercial vehicle product. Then, user input information is generated by integrating the user demand features.

[0029] For example, a user inputs the message "My AA-plated truck needs its insurance renewed. It's a commercial vehicle, please provide a quote" through the voice assistant of a commercial vehicle service platform. The platform can then use speech recognition technology to convert the input speech into text, perform word segmentation, and obtain a keyword set: {"AA-plated", "truck", "insurance renewal", "commercial", "truck", "quote"}.

[0030] Next, word vectors are constructed by extracting user demand features. This involves selecting keywords related to commercial vehicle products from the word segmentation set and converting these words into machine-understandable word vectors. When constructing word vectors, the aforementioned discrete keywords can be transformed into vector values ​​in a multi-dimensional space. For example, "operation" might be mapped to a vector space location that simultaneously includes the features of "high risk" and "high coverage," so that the commercial vehicle service platform can understand the type of commercial vehicle—that is, a production tool requiring special rates.

[0031] After constructing word vectors, the commercial vehicle service platform can integrate all extracted features to generate a structured user demand information. This user demand information is a structured instruction that can be directly executed by the corresponding business system for commercial vehicles. For example, the user input might be: "User intent: Commercial vehicle insurance renewal quote; Key parameters: Vehicle brand: AA; Vehicle type: Truck; Usage: Commercial; Triggering event: Insurance expires; Output instruction: Trigger the quote engine, retrieve the rate table for 'AA brand commercial trucks,' and prepare to generate a quote."

[0032] Regarding the acquisition of policy information, a multi-level, automated data collection network can be established to ensure the timeliness and accuracy of policy information. For policy-related websites that provide standard API interfaces, structured data interfaces can be used for connection; for those that cannot be obtained through APIs, compliant web crawling technology can be used to periodically crawl news updates and announcements from policy-related websites. For example, policy-related websites can include official government websites, government data platforms, and industry association data.

[0033] S102. Generate recommendation strategies based on multi-source business data.

[0034] After acquiring multi-source business data, a recommendation strategy can be generated based on this data. This recommendation strategy includes credit risk constraints and policy constraints. Credit risk constraints are constraints obtained through credit risk assessment using machine learning algorithms combined with multi-source business data, and are used to recommend commercial vehicle products that match the current user's credit risk level.

[0035] To generate credit risk constraints, commercial vehicle service platforms can incorporate an AI-driven risk assessment module. This module uses machine learning algorithms, combined with multi-dimensional data such as user credit information, financial status, and policy information, to assess user credit risk. This allows for accurate prediction of a user's repayment ability, determination of an appropriate credit limit, and reduction of financial risk.

[0036] In some embodiments, when generating recommendation strategies based on multi-source business data, credit information, user basic data, and user behavior data can be extracted from the multi-source business data first. Multi-dimensional credit data is then generated by integrating the credit information, user basic data, and user behavior data. Credit features are then extracted from the multi-dimensional credit data, whereby the credit features are multi-dimensional feature vectors constructed from multiple credit information items; the credit information is used to characterize the current user's credit status, financial situation, and credit limit.

[0037] The credit features are then input into the credit assessment model to obtain the credit risk level output by the model. This credit assessment model is a machine learning model trained using multi-dimensional credit sample data. Credit risk constraints are then set based on the credit risk level.

[0038] For example, such as Figure 3 As shown, when conducting risk assessments in commercial vehicle financial insurance scenarios, commercial vehicle service platforms can call the risk control assessment module and obtain user credit-related data, including basic user data and user behavior data, through the risk control assessment module, in order to build user profiles by collecting all internal and external information.

[0039] Basic user data can include credit-related information such as identity, income, assets, and liabilities. For example, basic user data might include that User A is self-employed, 45 years old, has a monthly income of 20,000-30,000 RMB, owns a mortgaged property, and has no other business loans. User behavior data can include economically related data such as a user's consumption, repayment, and borrowing within a specific period. For example, User A's monthly fuel and toll payment records are stable, credit card payments have been made on time for 5 years, and there are no small loan records.

[0040] In addition to acquiring user credit-related data such as basic user data and user behavior data, credit information can also be obtained through third-party credit assessment agencies. Credit information can include overdue payment records, social relationships, vehicle records, etc. For example, user A has no overdue payment records, has a spouse as a guarantor, intends to purchase a popular model of brand AA, and has no recall records, etc.

[0041] After acquiring credit information, basic user data, and user behavior data, multi-dimensional credit data can be generated through data integration. This multi-dimensional credit data then undergoes data cleaning to remove invalid data. For example, data cleaning can remove duplicate or erroneous transaction records from user-uploaded bank statements and filter out consumption records not registered in the user's name, ensuring the quality of the data input into the model.

[0042] Next, feature extraction is performed on the multi-dimensional credit data to obtain credit characteristics. Commercial vehicle service platforms can extract key indicators from this multi-dimensional credit data to transform it into financially meaningful metrics. These key indicators can include repayment ability, repayment willingness, and stability. Repayment ability can be set using income and expenditure data within a preset period, such as average monthly disposable income = (average monthly cash flow - fuel and toll costs - mortgage payments) = 12,000 yuan. Repayment willingness can be determined by the number of historical delinquencies, such as 0 historical delinquencies. Stability can be determined using data such as years of employment, such as 8 years in the transportation industry.

[0043] After extracting credit features, feature engineering can be used to standardize and process these features, making them understandable to backend credit assessment models. For example, feature engineering can transform credit features into features more conducive to model classification, such as the income-to-debt ratio = monthly repayment amount / monthly disposable income = proposed monthly payment of 3000 yuan / 12000 yuan = 0.25. It can also transform "8 years in the industry" into a highly stable score value. This allows data from different dimensions (such as income and years of experience) to be calculated within the same vector space.

[0044] To ensure high-precision output from credit assessment models, training can be performed. This involves inputting processed data into a credit assessment model built on architectures such as decision trees and XGBoost, allowing the model to learn from multi-dimensional credit sample data. For example, by collecting business data from multiple users and labeling them based on their credit behavior (defaulting and repaying), multi-dimensional credit sample data can be created. This multi-dimensional sample data can then be used to train the credit assessment model, enabling it to leverage historical experience to identify patterns in credit risk.

[0045] After training, the credit assessment model needs to be validated, which means evaluating the model's predictive accuracy to determine its usability. For example, if the credit assessment model considers user B to be low-risk, and the corresponding validation set shows that 98% of people with similar characteristics in the past did indeed repay on time, then the credit assessment model has practical predictive ability.

[0046] The trained credit features are then input into the credit assessment model, which performs risk scoring to obtain the credit risk level output by the model. For example, if the model outputs a comprehensive risk score of 650 for user A, assuming a score range of 300-900, then user A's credit risk level is determined to be low risk, simplifying complex calculations into intuitive scores. After obtaining the comprehensive risk score, the result can be output, such as "Risk Score: 650 (Low Risk)," and credit risk constraints can be set based on this level.

[0047] Credit risk constraints can be used to determine which commercial vehicle products match a user's current credit risk level, and to provide decision-making recommendations. For example, for a low-risk credit risk score, the commercial vehicle service platform can provide a decision recommendation of "approving the application." Recommended down payment ratio: 30%, i.e., standard down payment. Recommended annualized interest rate: 6.5%, enjoying a premium customer rate, to guide sales personnel in completing the final contract signing.

[0048] In some embodiments, the AI-driven risk assessment module does not simply apply general machine learning models, but rather optimizes algorithms specifically for the commercial vehicle finance scenario to improve the accuracy and adaptability of risk assessment. The credit assessment model can employ a deep fusion network of multi-source heterogeneous features. By constructing a multimodal feature cross-fusion network, it can process financial data such as user credit reports and bank statements, and also access real-time user business intentions such as expanding fleets and changing routes, as well as subsidy or right-of-way rules extracted from policy information modules. By designing cross-feature layers, it automatically captures high-order interaction relationships between these heterogeneous features. For example, the model can identify the combined feature that "although the user has minor credit flaws, they have recently expressed a strong intention to expand cold chain transportation in voice inquiries, and the local area has just introduced a high subsidy policy for new energy refrigerated trucks," thereby more accurately assessing their future repayment ability.

[0049] To meet the interpretability requirements of financial regulators for risk control decisions, credit assessment models can also integrate interpretable hybrid expert systems. This involves combining a deep neural network with an interpretable rule engine to form expert rules based on policy analysis results. The risk score is derived by a weighted average of the deep model and the rule engine. Simultaneously, through attention mechanisms and Shapley values, the contribution of key features such as "the reduction of policy subsidies leading to increased cash flow pressure" to the final credit limit can be output, enabling credit approval personnel to clearly understand the AI's decision-making basis.

[0050] Furthermore, credit risk assessment does not simply output a single risk level; it can also generate a multi-dimensional set of quantitative risk results to support more refined credit decisions. In some embodiments, risk level labels can be constructed first, serving as the most intuitive result, classifying users into credit ratings such as AAA, AA, A, BBB, BB, B, C, and D. This rating is the final qualitative conclusion based on data from all dimensions, allowing approval personnel to quickly grasp the user's qualifications.

[0051] You can also set the Probability of Default (PD) as a core output of machine learning. By providing a specific percentage value, such as PD=2.3%, it represents the probability that a user will be more than 90 days overdue or commit other breaches as defined in the contract within a specific time window, thus forming the basis for calculating expected losses.

[0052] Credit risk assessment results can also include a risk score. To ensure compatibility with financial institutions' credit scoring systems, the Product Development (PD) can be mapped to a standardized credit score, with higher scores indicating lower risk. This score can be directly entered into the credit approval system.

[0053] Furthermore, when outputting credit risk assessment results, risk characteristic attribution can be performed. This involves combining interpretability algorithms such as the Shapley score to output the key factors leading to the risk outcome and their contribution. Examples include "Credit history record: -15 points (contribution 30%)", "Current debt ratio too high: -20 points (contribution 40%)", and "Policy subsidies about to expire: -8 points (contribution 16%)", providing a basis for subsequent risk mitigation measures.

[0054] AI-driven risk assessment can construct an intelligent credit decision-making loop that balances maximizing returns and minimizing risks, achieving precise measurement and pricing. By determining an appropriate credit limit, combining product digits (PD) and residual value predictions, a risk quantification model is used to calculate the user's expected loss (EL), and then, based on this, the maximum loan amount that should be granted to the user under an acceptable rate of return on capital is calculated. For example, for new energy vehicles with high PD but stable residual value, a higher loan-to-value ratio but also a higher interest rate might be offered. It can also achieve differentiated information recommendations, automatically matching different interest rates and repayment plans based on risk level. The lower the risk, the more favorable the interest rate; the higher the risk, the higher the interest rate or the lower the loan amount to cover potential losses, thus achieving refined information recommendations.

[0055] AI-driven risk assessment also helps reduce default losses for financial institutions, enabling pre-loan interception and post-loan early warning. Specifically, at the application stage, it accurately identifies high-risk users, directly rejecting their applications or requiring them to provide credit enhancement measures, such as increasing down payments or finding guarantors, thus curbing the generation of non-performing loans at the source. For users who have already received loans, continuous risk assessment can output dynamic early warning signals, helping financial institutions identify potential risky users in advance and take timely post-loan management measures, such as telephone collection, on-site visits, and asset preservation, mitigating risks at an early stage.

[0056] Policy constraints are constraints obtained by parsing policy information and used to recommend suitable commercial vehicle products in conjunction with current relevant commercial vehicle policies. To generate policy constraints and adapt to relevant commercial vehicle policies in real time, the commercial vehicle service platform can also incorporate a dynamic policy adaptation module. This module acquires and analyzes the latest policy information in real time, dynamically adjusts the recommendation strategy, and ensures that the recommendations comply with the latest policy requirements. Furthermore, the dynamic policy adaptation module can automatically update the recommended product combination based on policy changes, improving the timeliness and accuracy of the recommendations.

[0057] In some embodiments, when generating recommendation strategies based on multi-source business data, policy information can first be extracted from the multi-source business data. Real-time policy parameters can then be obtained by parsing the policy information, whereby the real-time policy parameters include the policy's scope of application and a policy calculation function. Next, the product matching scope is determined based on the policy's scope of application, and then policy constraints are set based on the product matching scope and the policy calculation function.

[0058] For example, when generating policy constraints, the original policy text can be obtained by crawling government websites, reading policy PDFs, or having business personnel input data. The original policy text might include: "From date A to date B, a one-time subsidy will be given to individuals or companies registered in XX region who scrap heavy-duty diesel trucks with emission standards of National III or below and purchase new energy trucks with National VI emission standards. The subsidy amount is calculated based on vehicle tonnage: 0.8 million yuan per light truck, 1.5 million yuan per medium truck, and 2.5 million yuan per heavy truck. The purchased new vehicle must be registered and licensed in XX region."

[0059] After obtaining the original policy documents, they can be preprocessed and formatted. Specifically, the parsing unit of the dynamic policy adaptation module can utilize Natural Language Processing (NLP) and large model technology to transform unstructured policy text into structured rule engine data. For example, for text data, content extraction can be performed to remove noisy information such as advertisements and navigation bars from web pages or PDFs, and to extract the main text, attachments, and metadata such as issuing unit, document number, and publication date.

[0060] Then, natural language processing technology is used to parse the policy information text. Specifically, a pre-trained NLP model (BERT) is used to perform entity recognition and relation extraction on the policy text to obtain real-time policy parameters. These real-time policy parameters include: the scope of application of the policy, i.e., vehicle type is light truck, medium truck, or heavy truck; emission standard is National III or below; usage nature is commercial; fuel type is diesel; registration location is XX region; the scrapping behavior requires scrapping the old vehicle; new vehicle requirements are National VI or new energy, and registration is required in XX region; and time range [date A, date B].

[0061] The real-time policy parameters also include the policy calculation function, which calculates the subsidy amount as a fixed amount matched according to the new vehicle model. The mapping table is as follows: if the new vehicle model is "light truck", then the subsidy is 0.8 million yuan; if the new vehicle model is "medium truck", then the subsidy is 1.5 million yuan; if the new vehicle model is "heavy truck", then the subsidy is 2.5 million yuan.

[0062] Next, the product matching scope is determined based on the policy's scope of application. This involves cross-referencing the policy's scope with the corresponding financial products offered by the commercial vehicle service platform to filter out eligible commercial vehicle products. For example, if the company's product database includes: Product 1, J6F light truck (applicable); Product 2, D6 light truck (applicable); Product 3, T7 heavy truck (applicable); Product 4, FM heavy truck (not applicable); Product 5, used National V heavy truck (not applicable), then the output result would be: Product Matching Scope = {Product 1, Product 2, Product 3}.

[0063] Then, through rule-based and quantification, the semantic parsing results are converted into computer-executable logical rules, enabling the setting of policy constraints based on product matching range and policy calculation functions. For example, policy constraints are DRL files or rule sets configured in the decision engine. The corresponding generated rule logic is: "WHEN Application Date BETWEEN 'Date A' AND 'Date B' AND Used Vehicle.Emission Standard = National III AND Used Vehicle.Usage Nature = Commercial AND Used Vehicle.Fuel Type = Diesel AND Used Vehicle.Registration Location = XX Region AND New Vehicle.Model IN [Light Truck, Medium Truck, Heavy Truck] AND New Vehicle.Registration Location = XX Region THEN Activate Subsidy Calculator; SWITCH (New Vehicle.Model) {CASE Light Truck: Subsidy Amount = 8000; CASE Medium Truck: Subsidy Amount = 15000; CASE Heavy Truck: Subsidy Amount = 25000}.

[0064] S103. According to the recommendation strategy, use user input information to match commercial vehicle product combinations.

[0065] After generating the recommendation strategy, the commercial vehicle service platform can match commercial vehicle products, that is, match a combination of commercial vehicle products using the user's input information according to the recommendation strategy. This combination of commercial vehicle products includes at least one of the following: commercial vehicle loan products, commercial vehicle insurance products, and commercial vehicle after-sales service products.

[0066] In order to match a suitable commercial vehicle product portfolio, in some embodiments, a user demand vector can be generated first based on user input information, and then the matching degree between the user input information and the commercial vehicle products can be calculated based on the user demand vector. The matching degree is calculated based on the semantic distance between the user demand vector and the product vector corresponding to the commercial vehicle product.

[0067] When calculating the matching degree between user input information and commercial vehicle products, the semantic distance between the user's demand vector and the corresponding product vector of the commercial vehicle product can be calculated first. Then, an initial match is calculated based on the semantic distance. Next, a user profile is constructed based on the user's basic data and user behavior data. Then, the user's personalized information, including usage habit information and product preference information, is determined based on the user profile. Finally, a matching weight is calculated according to the personalized information, and the initial matching degree is corrected using the matching weight to obtain the final matching degree between the user input information and the commercial vehicle product.

[0068] For example, a comprehensive user profile can be built based on user basic data, user behavior data, user credit information, and other integrated user information. A user profile is a three-dimensional portrait that includes static attributes, dynamic behaviors, creditworthiness, and business scenarios. Therefore, a user profile can include basic attribute information, dynamic intent and preference information, credit and financial information, etc.

[0069] The basic attribute information can include identity characteristics, user type, industry, and geographic location. Dynamic intent preference information can be obtained from the user's explicit needs and implicit preferences parsed from the voice input module. Explicit needs can be directly obtained from user input information. For example, if a user inputs "I want to change to a pure electric light truck with a long range for city delivery," the explicit need can be determined as a pure electric light truck. Implicit preferences can be determined by analyzing relevant data such as the user's historical driving behavior, APP browsing behavior, and historical maintenance records. For example, if a user repeatedly mentions keywords such as "comfort," "automatic transmission," and "residual value" in multiple rounds of dialogue, the implicit preference can be determined as vehicle performance characteristics such as comfort, automatic transmission, and resale value. Credit and financial information can include credit and risk control results, financial status, etc.

[0070] After determining personalized information such as usage habits and product preferences, matching weights can be calculated based on these personalized information. These matching weights can be used to selectively increase or decrease the matching degree of certain commercial vehicle products. Specifically, for commercial vehicle products that match the current user's applicable habits and product preferences, a higher matching weight can be set to increase the matching degree of the corresponding commercial vehicle product; while for commercial vehicle products that do not match the current user's applicable habits and product preferences, a lower matching weight can be set to decrease the matching degree of the corresponding commercial vehicle product.

[0071] After calculating the matching degree, a first set of candidate products can be generated based on the matching degree. This first set of candidate products includes commercial vehicle products with a matching degree greater than or equal to a matching degree threshold. Specifically, by traversing the matching degree of all products in the commercial vehicle product database and comparing it with a preset matching degree threshold, commercial vehicle products with a matching degree greater than or equal to the threshold are selected to form the first set of candidate products.

[0072] Then, according to the recommended strategy, a second set of alternative products is extracted from the first set of alternative products. That is, the commercial vehicle products in the first set of alternative products are filtered using the credit risk constraints and policy constraints in the recommended strategy to select commercial vehicle products that meet the credit risk constraints and policy constraints, thus forming the second set of alternative products.

[0073] In some embodiments, when extracting the second set of alternative products, products can be initially extracted based on credit risk constraints. For example, after determining the current user's credit risk level, products with credit risk requirements lower than the user's credit risk level from the first set of alternative products can be extracted to avoid recommending commercial vehicle products that cannot be purchased to the user.

[0074] After the initial product selection is completed, the policy constraints are used to further screen the initial product selection results. Specifically, commercial vehicle products that meet the policy constraints are extracted from the initial product selection results to recommend commercial vehicle products that meet the current policy incentives to users.

[0075] It should be noted that, in order to ensure the richness of the recommended results, the degree of policy constraints can be dynamically adjusted based on the actual commercial vehicle product screening results. For example, if there are few commercial vehicle products in the initial product selection results after the initial product selection, the policy constraints will not be used to screen the initial product selection results again, and the initial product selection results will be directly used as the second set of alternative products.

[0076] After extracting the second set of candidate products, the commercial vehicle products in the second set are ranked according to their matching degree, and a commercial vehicle product combination is generated based on the ranked commercial vehicle products. To this end, the commercial vehicle service platform can deploy a personalized recommendation algorithm module. This module can recommend the most suitable commercial vehicle product combination, including loan products, insurance products, and after-sales services, based on the user's comprehensive information and market product conditions. Market product conditions refer to the composition of the full lifecycle product library, which is a structured database updated in real time by the commercial vehicle service platform, covering all recommendable products throughout the entire lifecycle of commercial vehicles. Market product conditions can be a complete product matrix including the vehicle itself, financial solutions, insurance coverage, and after-sales services. Furthermore, the personalized recommendation algorithm module can dynamically adjust the recommendation strategy by analyzing user habits and preferences to improve user satisfaction.

[0077] S104. Generate recommendation information based on commercial vehicle product portfolio.

[0078] After matching the commercial vehicle product portfolio, the system can output the matched commercial vehicle product portfolio, that is, generate recommendation information based on the commercial vehicle product portfolio. The recommendation information can be output in the form of text, graphics, images, audio, video and multimedia messages to display the commercial vehicle product portfolio to users.

[0079] The recommended information includes commercial vehicle order placement and decision-making advice. Commercial vehicle order placement can include basic information about the commercial vehicle product and transaction-related details. Decision-making advice can include purchase recommendations and other decision-making-related content. It should be noted that commercial vehicle order placement and decision-making advice are only for reference and display purposes; they do not replace user decision-making, and the advice will not contain misleading content to ensure the fairness of the transaction.

[0080] In order to generate recommendation information, in some embodiments, when generating recommendation information based on commercial vehicle product portfolios, information to be displayed can be extracted from the commercial vehicle product portfolios first. The information to be displayed includes product name, product amount and product purchase link. Then, a commercial vehicle order form is generated based on the information to be displayed, which is used to display recommended commercial vehicle loan products, commercial vehicle insurance products and commercial vehicle after-sales service products, etc.

[0081] Then, the matching degree of commercial vehicle products in the commercial vehicle product portfolio is obtained, and first decision suggestion information is generated based on the matching degree, and second decision suggestion information is generated based on the credit risk level. The first decision suggestion information is used to provide product purchase reference for the current user; the second decision suggestion information is used to provide review reference for the product provider.

[0082] It is evident that when generating recommendation information, commercial vehicle service platforms can, on the one hand, integrate product name, product amount, and product purchase link to form and display commercial vehicle order forms. On the other hand, they can generate secondary decision-making suggestions based on credit risk levels, providing product purchase recommendations to users; simultaneously, they can also generate secondary decision-making suggestions based on credit risk levels, providing reviewers with a reference for their review.

[0083] By applying the technical solutions of the above embodiments, the intelligent order placement method for commercial vehicles provided in the above embodiments can realize intelligent order placement and product recommendation for commercial vehicle business based on multi-source data. The method can integrate multi-source data such as voice input, credit information, and policy information, and combine AI algorithms to provide users with personalized commercial vehicle product recommendations and complete credit limit assessment, thereby improving the output efficiency and personalization of recommendation information and solving the problem of low output efficiency and personalization of the order placement recommendation system.

[0084] In some embodiments, as a refinement and extension of the specific implementation of the above embodiments, and to fully illustrate the specific implementation process of this embodiment, some embodiments of this application also provide a method for intelligent order reporting for commercial vehicles. The difference between this method and the above embodiments is that it can dynamically adjust the recommendation strategy based on multiple rounds of user dialogue to provide personalized services. For example... Figure 4 As shown, the method includes: S201. Obtain multi-turn interactive dialogue data corresponding to the voice information; S202. Extract user demand features from multi-turn interactive dialogue data; S203. By comparing the characteristics of user needs in multi-round interactive dialogues, determine the information on changes in needs; S204. Use requirement change information to dynamically adjust target matching items.

[0085] For example, in order to achieve dynamic adjustment of recommendation strategies, commercial vehicle service platforms can build in an intelligent voice interaction module. The intelligent voice interaction module can acquire multi-turn interactive dialogue data corresponding to voice information and extract user demand features from the multi-turn interactive dialogue data.

[0086] Then, by comparing the user demand characteristics across multiple rounds of interactive dialogue, demand change information is determined. This demand change information includes at least one of product change information and strategy change information. The product change information is used to update user input information; the strategy change information is used to dynamically adjust the recommendation strategy.

[0087] Then, the target matching item is dynamically adjusted using the demand change information, wherein the target matching item includes at least one of user input information and recommendation strategy. That is, by comparing the user demand characteristics of multi-turn interactive dialogues, when it is determined that the demand change information includes product change information, the target matching item can be determined to be user input information, and the user input information can be updated according to the product change information. Similarly, when it is determined that the demand change information includes strategy change information, the recommendation strategy can be dynamically adjusted based on the strategy change information.

[0088] By applying the technical solutions of the above embodiments, the intelligent order reporting method for commercial vehicles described in the above embodiments can not only utilize natural language processing technology to achieve natural dialogue between users and the system, thus improving user experience, but also support multi-turn dialogue based on user voice input, dynamically adjust recommendation strategies, and provide personalized services.

[0089] In some embodiments, as a refinement and extension of the specific implementation of the above embodiments, and to fully illustrate the specific implementation process of this embodiment, some embodiments of this application also provide a method for intelligent order reporting for commercial vehicles. The difference between this method and the above embodiments is that, after generating a recommendation strategy based on multi-source business data, policy constraints can be updated, such as... Figure 5 As shown, the method includes: S301. Obtain the set of updated policies based on the product matching scope; S302. Parse the update policy parameters from the update policy set; S303. Determine policy update items by comparing real-time policy parameters and updated policy parameters; S304. Update policy constraints using the policy update item.

[0090] For example, the dynamic policy adaptation module built into the commercial vehicle service platform can generate a recommendation strategy based on multi-source business data and then obtain an updated policy set based on the product matching range. The updated policy set includes real-time policies obtained from multiple data sources, and the real-time policies are correlated with the product matching range.

[0091] The updated policy parameters, including policy calculation functions, are then parsed from the updated policy set. Policy update items are then determined by comparing real-time policy parameters with the updated policy parameters, and these update items are used to update policy constraints. Therefore, the dynamic policy adaptation module can dynamically adjust its recommendation strategy by acquiring and analyzing the latest policy information in real time, ensuring that the recommendation results meet the latest policy requirements. Furthermore, it automatically updates the recommended product portfolio based on policy changes, improving the timeliness and accuracy of the recommendations.

[0092] By applying the technical solutions of the above embodiments, the intelligent order placement method for commercial vehicles described in the above embodiments can improve the accuracy of recommendations, that is, significantly improve the accuracy and efficiency of product recommendations through multi-source data integration and AI algorithms. It can also reduce financial risks, that is, by using machine learning risk control models to accurately assess users' credit risk and reduce financial risks. The method can also optimize user experience, improving the user experience through intelligent voice interaction and personalized recommendations. Furthermore, it enables real-time data processing and dynamic policy adaptation, significantly improving the work efficiency of business personnel.

[0093] In some embodiments, as a specific implementation of the intelligent order reporting method for commercial vehicles described in the above embodiments, some embodiments of this application also provide an intelligent order reporting device for commercial vehicles, such as... Figure 6 As shown, the device includes: The multi-source data integration module is used to acquire multi-source business data, including user input information, credit information, and policy information. The strategy generation module is used to generate a recommendation strategy based on the multi-source business data. The recommendation strategy includes credit risk constraints and policy constraints. The credit risk constraints are constraints obtained by using machine learning algorithms to perform credit risk assessment in conjunction with the multi-source business data. The policy constraints are constraints obtained by parsing the policy information. The product matching module is used to match a commercial vehicle product combination with the user input information according to the recommendation strategy. The commercial vehicle product combination includes at least one of commercial vehicle loan products, commercial vehicle insurance products, and commercial vehicle after-sales service products. The personalized recommendation algorithm module is used to generate recommendation information based on the commercial vehicle product portfolio. The recommendation information includes commercial vehicle order information and decision-making suggestions.

[0094] By employing the above technical solutions, embodiments of this application provide a method, apparatus, device, and storage medium for intelligent commercial vehicle order placement. The method generates a recommendation strategy by acquiring multi-source business data, including user input information, credit information, and policy information. The recommendation strategy includes credit risk constraints obtained through credit risk assessment using machine learning algorithms combined with multi-source business data, and policy constraints obtained by parsing policy information. Then, according to the recommendation strategy, user input information is used to match commercial vehicle product combinations, and recommendation information is generated based on these combinations. By integrating multi-source business data and combining AI algorithms, this method can output personalized commercial vehicle product recommendation information and complete credit limit assessment, thereby improving the output efficiency and personalization of commercial vehicle order placement.

[0095] It should be noted that other corresponding descriptions of the functional units involved in the intelligent order reporting device for commercial vehicles provided in the embodiments of this application can be found in the corresponding descriptions in the intelligent order reporting method for commercial vehicles provided in the above embodiments, and will not be repeated here.

[0096] This application also provides a computer device, specifically a personal computer, server, network device, etc. The computer device includes a bus, processor, memory, and communication interface, and may also include input / output interfaces and a display device. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores location information. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the various method embodiments.

[0097] Those skilled in the art will understand that the structure of the computer device described above is only a partial structure related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. A specific computer device may include more or fewer components, or combine certain components, or have different component arrangements.

[0098] In one embodiment, a computer-readable storage medium is also provided, which may be non-volatile or volatile, and a computer program is stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0099] In one embodiment, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0100] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0101] Those skilled in the art will understand that all or part of the processes in 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 the computer program is executed, it can include the processes of the embodiments of the above methods.

[0102] Any references to memory, database, or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc.

[0103] Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can take many forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0104] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchain. The processors involved in the embodiments provided in this application may be, but are not limited to, general-purpose processors, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc.

[0105] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0106] The embodiments described above are merely examples of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application.

Claims

1. A method for intelligent order reporting for commercial vehicles, characterized in that, The method includes: Acquire multi-source business data, including user input information, credit information, and policy information; A recommendation strategy is generated based on the multi-source business data. The recommendation strategy includes credit risk constraints and policy constraints. The credit risk constraints are constraints obtained by using machine learning algorithms to perform credit risk assessment in conjunction with the multi-source business data. The policy constraints are constraints obtained by parsing the policy information. According to the recommendation strategy, the user input information is used to match a commercial vehicle product portfolio, which includes at least one of commercial vehicle loan products, commercial vehicle insurance products, and commercial vehicle after-sales service products. Recommendation information is generated based on the aforementioned commercial vehicle product portfolio. The recommendation information includes commercial vehicle order information and decision-making suggestions.

2. The method according to claim 1, characterized in that, Acquire multi-source business data, including: Obtain user-inputted voice information; The voice information is converted into user input text, which is a word segmentation set containing multiple keywords; Extract user demand features from the user input text, wherein the user demand features are word vectors constructed from keywords associated with at least one commercial vehicle product; The user input information is generated by integrating the user demand characteristics.

3. The method according to claim 2, characterized in that, The method further includes: Obtain the multi-turn interactive dialogue data corresponding to the voice information; Extract the user demand features from the multi-turn interactive dialogue data; By comparing the user demand characteristics in multi-round interactive dialogues, demand change information is determined. The demand change information includes at least one of product change information and strategy change information. The product change information is used to update the user input information. The strategy change information is used to dynamically adjust the recommendation strategy. The target matching item is dynamically adjusted using the demand change information, and the target matching item includes at least one of the user input information and the recommendation strategy.

4. The method according to claim 1, characterized in that, The recommendation strategy is generated based on the multi-source business data, including: The credit information is extracted from the multi-source business data, and basic user data and user behavior data are obtained. By integrating the credit information, the user's basic data, and the user's behavior data, multi-dimensional credit data is generated. Credit features are extracted from the multi-dimensional credit data. These credit features are multi-dimensional feature vectors constructed from multiple credit information. The credit information is used to characterize the current user's credit status, financial situation, and credit limit. The credit features are input into the credit assessment model to obtain the credit risk level output by the credit assessment model; the credit assessment model is a machine learning model trained using multi-dimensional credit sample data. The credit risk constraints are set according to the credit risk level.

5. The method according to claim 1, characterized in that, The recommendation strategy is generated based on the multi-source business data, including: Extract the policy information from the multi-source business data; By parsing the policy information, real-time policy parameters are obtained, including the policy scope and policy calculation function. The product matching scope shall be determined according to the scope of application of the aforementioned policy; The policy constraints are set based on the product matching range and the policy calculation function.

6. The method according to claim 5, characterized in that, After generating a recommendation strategy based on the multi-source business data, the method further includes: An updated policy set is obtained based on the product matching scope, and the updated policy set includes real-time policies obtained from multiple data sources; the real-time policies are correlated with the product matching scope. The updated policy parameters are parsed from the updated policy set, and the updated policy parameters include policy calculation functions; The policy update item is determined by comparing the real-time policy parameters and the updated policy parameters; Update the policy constraints using the policy update item.

7. The method according to claim 1, characterized in that, According to the recommendation strategy, the user input information is used to match commercial vehicle product combinations, including: Generate a user demand vector based on the user input information; Based on the user demand vector, the matching degree between the user input information and the commercial vehicle product is calculated. The matching degree is obtained based on the semantic distance between the user demand vector and the product vector corresponding to the commercial vehicle product. A first set of candidate products is generated based on the matching degree, and the first set of candidate products includes commercial vehicle products whose matching degree is greater than or equal to the matching degree threshold. According to the recommendation strategy, a second set of candidate products is extracted from the first set of candidate products; Based on the matching degree, the commercial vehicle products in the second candidate product set are sorted, and the commercial vehicle product combination is generated based on the sorted commercial vehicle products.

8. A smart order reporting device for commercial vehicles, characterized in that, The device includes: The multi-source data integration module is used to acquire multi-source business data, including user input information, credit information, and policy information. The strategy generation module is used to generate a recommendation strategy based on the multi-source business data. The recommendation strategy includes credit risk constraints and policy constraints. The credit risk constraints are constraints obtained by using machine learning algorithms to perform credit risk assessment in conjunction with the multi-source business data. The policy constraints are constraints obtained by parsing the policy information. The product matching module is used to match a commercial vehicle product combination with the user input information according to the recommendation strategy. The commercial vehicle product combination includes at least one of commercial vehicle loan products, commercial vehicle insurance products, and commercial vehicle after-sales service products. The personalized recommendation algorithm module is used to generate recommendation information based on the commercial vehicle product portfolio. The recommendation information includes commercial vehicle order information and decision-making suggestions.

9. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images