Vehicle type agile recommendation method and system based on digital twin model

Abstract

The invention discloses a vehicle type agile recommendation method and system based on a digital twin model, and belongs to the technical field of intelligent data processing.The vehicle type agile recommendation method based on the digital twin model comprises the steps that real-time passenger vehicle data and real-time user comment data are obtained, constructing a passenger vehicle type twin model library according to the real-time passenger vehicle data and the real-time user comment data; acquiring real-time demand data of the user for the vehicle type parameters and real-time importance score data of the user for the vehicle type attributes, and constructing a user demand twin model according to the real-time demand data of the user for the vehicle type parameters and the real-time importance score data of the user for the vehicle type attributes; and obtaining a vehicle model recommendation result based on a two-stage recommendation algorithm, the passenger vehicle model twin model and the user demand twin model. According to the method, the real-time performance, the accuracy and the individuation degree of vehicle model recommendation can be effectively improved, and the passenger vehicle models meeting the requirements of consumers can be more effectively and accurately recommended for the consumers.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent data processing, and particularly to a method and system for agile vehicle model recommendation based on a digital twin model. Background Art

[0002] With the continuous development of the new energy passenger vehicle market, consumers face an increasingly rich choice of vehicle models. However, they also face the challenge of how to select the most suitable vehicle model from numerous models. Traditional car recommendation algorithms mainly generate recommendation results through static similarity calculations. However, most of these algorithms rely on basic user information (such as vehicle model, price, power type, etc.), and fail to fully consider the individual demand differences of users in different fields (such as space, appearance, interior, etc.), as well as the multi-level changes of user needs over time. This leads to low real-time performance, accuracy, and user satisfaction of the recommendation results, and cannot timely reflect the changes in new vehicle information or user reviews in the market, resulting in a situation where the recommendation results are often not accepted by users. Summary of the Invention

[0003] The purpose of this application is to overcome the defects of the prior art and provide a method and system for agile vehicle model recommendation based on a digital twin model, which combines the multi-dimensional data of new energy passenger vehicles with the multi-level characteristics of user needs to achieve more personalized, accurate, and real-time updated recommendations.

[0004] In the first aspect, this application provides a method for agile vehicle model recommendation based on a digital twin model, including the following steps: Obtain real-time passenger vehicle data and real-time user review data, and construct a passenger vehicle model twin model library according to the real-time passenger vehicle data and the real-time user review data. The passenger vehicle model twin model library is dynamically updated according to new passenger vehicle data and new user review data; Obtain real-time demand data of users for vehicle model parameters and real-time importance scoring data of users for vehicle model attributes, and construct a user demand twin model according to the real-time demand data of users for vehicle model parameters and the real-time importance scoring data of users for vehicle model attributes; Obtain vehicle model recommendation results based on a two-stage recommendation algorithm, the passenger vehicle model twin model, and the user demand twin model.

[0005] Optionally, constructing a passenger vehicle model twin model library according to the real-time passenger vehicle data and the real-time user review data, and dynamically updating the passenger vehicle model twin model library according to new passenger vehicle data and new user review data includes: Perform structured processing on the real-time passenger vehicle data to obtain parameter feature data; Obtain the emotional average value of each vehicle model attribute based on the real-time user review data as word-of-mouth feature data; Vectorize the parameter feature data and the word-of-mouth feature data, and perform vector fusion on the vectorized parameter feature data and the word-of-mouth feature data to obtain the passenger vehicle model twin model; Store the passenger vehicle model twin models of all vehicle attributes in a database to obtain a passenger vehicle model twin model library, and the passenger vehicle model twin model library is dynamically updated according to new passenger vehicle data and new user review data.

[0006] Optionally, perform structured processing on the real-time passenger vehicle data to obtain parameter feature data, including: performing numerical standardization and categorical encoding on the real-time passenger vehicle data to obtain the parameter feature data.

[0007] Optionally, construct a user demand twin model according to the real-time demand data of the user for vehicle parameters and the real-time importance scoring data of the user for vehicle attributes, including: Obtain a vehicle parameter demand feature vector based on the real-time demand data of the user for vehicle parameters; Normalize the real-time importance scoring data of the user for vehicle attributes to obtain a vehicle attribute demand weight vector; Fuse the vehicle parameter demand feature vector and the vehicle attribute demand weight vector to obtain the user demand twin model.

[0008] Optionally, the formula for normalizing the importance scoring data of the user for vehicle attributes is: where, is the demand weight of the i-th vehicle attribute, n is the total number of vehicle attributes, is the importance scoring data of the i-th vehicle attribute.

[0009] Optionally, obtaining the vehicle model recommendation result based on the two-stage recommendation algorithm, the passenger vehicle model twin model library, and the user demand twin model includes: Screen the passenger vehicle model twin model library in a preset order to obtain candidate vehicle models; Obtain the vehicle attribute demand weight vector of the candidate vehicle model based on the user demand twin model; Obtain the word-of-mouth feature vector of the candidate vehicle model based on the passenger vehicle model twin model; Calculate the weighted cosine similarity between the vehicle attribute demand weight vector and the word-of-mouth feature vector to obtain their weighted cosine similarity; sort the candidate vehicle models in descending order according to the weighted cosine similarity, and select the top several candidate vehicle models as the vehicle model recommendation result.

[0010] Optionally, the preset order sequentially includes: purchase budget, cruising range, vehicle type level, power type, and charging time.

[0011] Optionally, in the process of sequentially screening the passenger vehicle model digital twin model in the preset order to obtain candidate models, if the number of candidate models is lower than the preset threshold, the restrictive conditions are cancelled in reverse order or the numerical condition thresholds are relaxed according to a preset ratio until the number of candidate models reaches the preset threshold.

[0012] Optionally, the weighted cosine similarity between the vehicle attribute demand weight vector and the word-of-mouth feature vector is calculated based on the following formula to obtain the weighted cosine similarity between the two: where Similarity is the weighted cosine similarity between the vehicle attribute demand weight vector and the word-of-mouth feature vector; is the demand weight of the i-th vehicle attribute, is the emotional average value of the i-th vehicle attribute, and n is the total number of vehicle attributes.

[0013] In a second aspect, the present application also provides a vehicle model agile recommendation system based on a digital twin model, including: A data acquisition module, configured to obtain real-time passenger vehicle data, real-time user comment data, real-time user demand data for vehicle model parameters, and real-time importance scoring data of users for vehicle model attributes; A passenger vehicle model digital twin model construction module, configured to construct a passenger vehicle model digital twin model according to the real-time passenger vehicle data and the real-time user comment data, and store the passenger vehicle model digital twin model in a database; A user demand digital twin model construction module, configured to construct a user demand digital twin model according to the user demand data for vehicle model parameters and the importance scoring data of users for vehicle model attributes; A recommendation result generation module, configured to obtain a vehicle model recommendation result based on a two-stage recommendation algorithm, the passenger vehicle model digital twin model, and the user demand digital twin model.

[0014] The present application provides a vehicle model agile recommendation method and system based on a digital twin model. By collecting real-time passenger vehicle data, real-time user comment data, and real-time user demand data for vehicle model parameters, constructing a passenger vehicle model digital twin model and a real-time user demand digital twin model, and obtaining a vehicle model recommendation result based on a two-stage recommendation algorithm, a passenger vehicle model digital twin model, and a user demand digital twin model, it can reflect the vehicle model characteristics and user needs in real time and accurately, improve the accuracy and timeliness of recommendations; the screening and adjustment mechanism can ensure the diversity of recommendation results, meet the personalized needs of different users, and enhance the user's car purchase experience and decision-making efficiency.

[0015] In order to make the above features and advantages of the invention more obvious and understandable, specific embodiments will be given below and described in detail in conjunction with the accompanying drawings as follows. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 and Figure 2 is a flowchart of a method for agile vehicle model recommendation based on a digital twin model provided in an embodiment of the present application.

[0018] Figure 3 is a flowchart of step S10 in the method for agile vehicle model recommendation based on a digital twin model provided in an embodiment of the present application.

[0019] Figure 4 is a flowchart of step S20 in the method for agile vehicle model recommendation based on a digital twin model provided in an embodiment of the present application.

[0020] Figure 5 is a flowchart of step S30 in the method for agile vehicle model recommendation based on a digital twin model provided in an embodiment of the present application.

[0021] Figure 6 is a schematic structural diagram of a system for agile vehicle model recommendation based on a digital twin model provided in another embodiment of the present application. Detailed Embodiments

[0022] To make the objectives and technical solutions of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions of the embodiments of the present application in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the described embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application.

[0023] In one embodiment, please refer to Figure 1 and Figure 2 , the present application provides a method for agile vehicle model recommendation based on a digital twin model. The method for agile vehicle model recommendation based on a digital twin model may include the following steps: S10 to S30.

[0024] S10: Acquire real-time passenger car data and real-time user review data, and construct a passenger car model twin model library based on the real-time passenger car data and the real-time user review data, wherein the passenger car model twin model library is dynamically updated based on new passenger car data and new user review data.

[0025] S20: Obtain user demand data on vehicle model parameters and user real-time importance rating data on vehicle model attributes, and build a user demand twin model based on the user demand data on vehicle model parameters and user real-time importance rating data on vehicle model attributes.

[0026] S30: Obtain vehicle model recommendation results based on the two-stage recommendation algorithm, the passenger vehicle model twin model library and the user demand twin model.

[0027] In the vehicle model agile recommendation method based on the digital twin model of this application, by obtaining real-time passenger car data and real-time user comment data, it is possible to accurately present the actual conditions of the vehicle model, such as various parameters and configurations, so that the basic information based on the recommendation has timeliness; by constructing a passenger vehicle model twin model and a user demand twin model, it is possible to accurately fit the user's personalized needs and avoid the blindness of general recommendations; through a two-stage recommendation algorithm, it is possible to highlight key factors and make the calculated similarity more scientific and reasonable. This application can capture the actual situation of the vehicle model and the personalized needs of the user in real time and accurately, improve the fit between the recommendation results and the user's expectations, provide users with more targeted and timely vehicle model recommendations, and effectively help users quickly screen out their favorite models and improve the efficiency of car purchase decisions.

[0028] As an example, the vehicle model agile recommendation method based on the digital twin model of the present application can be used to recommend vehicle models such as gasoline vehicles, electric vehicles, and new energy vehicles. Taking new energy vehicles as an example, steps S10 to S30 are introduced in detail below.

[0029] In step S10, refer to Figure 1 In step S10, real-time passenger car data and real-time user comment data are obtained, and a passenger car model twin model library is constructed based on the real-time passenger car data and the real-time user comment data. The passenger car model twin model library is dynamically updated based on new passenger car data and new user comment data.

[0030] As an example, see Figure 2 In step S10, web crawler technology can be used to obtain feature data (i.e., real-time passenger car data and real-time user comment data) uploaded in real time by the physical entity of the passenger car (i.e., the passenger car); a passenger car model twin model is constructed based on the real-time passenger car data and the real-time user comment data to form a passenger car model twin model library, and the twin model is iterated in real time.

[0031] As an example, in step S10, constructing a passenger car model twin based on the real-time passenger car data and the real-time user comment data may include: performing structured processing on the real-time passenger car data to obtain parameter feature data; obtaining the emotional average value of each vehicle model attribute based on the real-time user comment data as word-of-mouth feature data; vectorizing the parameter feature data and the word-of-mouth feature data, and performing vector fusion on the vectorized parameter feature data and word-of-mouth feature data to obtain the passenger car model twin; storing the passenger car model twins of all vehicle model attributes in a database to obtain a passenger car model twin library, and the passenger car model twin library is dynamically updated according to new passenger car data and new user comment data.

[0032] Specifically, please combine Figures 1 to 2 refer to Figure 3 , step S10 may include the following steps: S101~S105.

[0033] S101: Use web crawler technology to obtain real-time passenger car data and real-time user comment data.

[0034] S102: Perform structured processing on the real-time passenger car data to obtain parameter feature data.

[0035] S103: Use natural language algorithms to analyze each piece of the real-time user comment data according to vehicle model attributes to obtain the emotional values of each vehicle model attribute.

[0036] S104: Calculate the average value of the emotional values of each vehicle model attribute respectively to obtain the emotional average value of each vehicle model attribute as word-of-mouth feature data.

[0037] S105: Vectorize the parameter feature data and the word-of-mouth feature data, and perform vector fusion on the vectorized parameter feature data and word-of-mouth feature data to obtain the passenger car model twin; in this step, perform multi-dimensional vector fusion on the parameter feature data and the word-of-mouth feature data to obtain the passenger car model twin.

[0038] As an example, in step S101, the parameter data of the passenger car and the real-time user comment data can be obtained in real time through the API interface and web crawler technology. Specifically, by searching data sources such as automotive vertical media, automaker official websites, and social media platforms, determine their open API interfaces, and use web crawler technology to crawl real-time new energy passenger car data from automotive vertical media, automaker official websites, and social media platforms, and real-time user comment data can be crawled from social media platforms.

[0039] As an example, the user comment data may include but is not limited to: vehicle model level, power type, cruising range, charging time, and guiding price, etc.

[0040] As an example, in step S102, since the captured real-time passenger vehicle data may contain a large amount of noise and non-standard information, it is necessary to structurally process the real-time passenger vehicle data to eliminate the dimension difference. The real-time passenger vehicle data can be parsed and converted into a structured format.

[0041] As an example, the structural processing of the real-time passenger vehicle data to obtain parameter feature data includes: numerically normalizing and class encoding the real-time passenger vehicle data to obtain the parameter feature data; that is, the structural processing may include: numerical normalization, class encoding. Specifically, numerical data can be normalized, for example, the charging time is unified in hours; class data can be encoded, for example, the vehicle model level is mapped to discrete labels.

[0042] As an example, in step S103, a pre-trained natural language algorithm is used to perform domain classification and sentiment analysis on each real-time user comment data. All real-time user comment data of a single vehicle model are classified according to vehicle model attributes, and sentiment analysis is performed on each vehicle model attribute respectively to obtain multiple sentiment values in each domain. The sentiment value is set in the interval (0, 1) to reflect the user's satisfaction with a certain domain.

[0043] As an example, the vehicle model attributes may include but are not limited to: space (i.e., interior space), battery life, appearance, interior, cost performance, driving texture, in-vehicle intelligent system.

[0044] As an example, in step S104, the mean value of the sentiment values of each vehicle model attribute is calculated respectively to obtain the sentiment average score of each vehicle model attribute as the comprehensive score of the corresponding domain, that is, the word-of-mouth feature data.

[0045] As an example, in step S105, the parameter feature data and the word-of-mouth feature data are written in vector form, and then vector splicing and fusion are performed to obtain the passenger vehicle model twin model.

[0046] As an example, the passenger vehicle model twin models of all vehicle model attributes are stored in a database to obtain a passenger vehicle model twin model library.

[0047] As an example, the passenger vehicle model twin model library can adopt a distributed database. The passenger vehicle model twin model library can be dynamically updated. The passenger vehicle model twin model library supports high-concurrency reading and writing and real-time update, and can be dynamically updated according to new passenger vehicle data and new user comment data.

[0048] As an example, it can be set to a timed scanning mode to scan the data source at regular intervals, detect changes in the data source, and initialize the passenger car model twin model of the vehicle model when a new vehicle model is added; when a vehicle model is taken off the shelf, remove the passenger car model twin model of the corresponding vehicle model; when the vehicle model parameters or review data are updated, trigger a partial recalculation to ensure that the passenger car model twin model library is synchronized with the market status.

[0049] As an example, regularly scan data sources such as automotive vertical media, automaker official websites, and social media platforms to detect parameter changes or new reviews. If an update is detected, re-determine the passenger car model twin model of the affected vehicle model and replace the old version of the passenger car model twin model.

[0050] As an example, for a new vehicle model, if the initial review data is insufficient, the missing fields can be filled with the historical review average of vehicle models of the same brand or the same price range.

[0051] In step S20, please refer to Figure 1 step S20 in, obtain the real-time demand data of the user for vehicle model parameters and the real-time importance scoring data of the user for vehicle model attributes, and construct a user demand twin model according to the real-time demand data of the user for vehicle model parameters and the real-time importance scoring data of the user for vehicle model attributes.

[0052] As an example, please refer to Figure 2 , in step S20, a questionnaire can be formed based on the real-time demand feedback of the user, and a user demand twin model can be constructed according to the questionnaire.

[0053] As an example, in step S20, the constructing of the user demand twin model according to the real-time demand data of the user for vehicle model parameters and the importance scoring data of the user for vehicle model attributes may include: obtaining a vehicle model parameter demand feature vector based on the real-time demand data of the user for vehicle model parameters; normalizing the real-time importance scoring data of the user for vehicle model attributes to obtain a vehicle model attribute demand weight vector; and fusing the vehicle model parameter demand feature vector and the vehicle model attribute demand weight vector to obtain the user demand twin model.

[0054] As an example, please combine Figure 1 and Figure 2 refer to Figure 4 , step S20 may include the following steps: S201~S204.

[0055] S201: Obtain the real-time demand data of the user for vehicle model parameters and the real-time importance scoring data of the user for vehicle model attributes through a real-time questionnaire.

[0056] S202: Perform data processing (such as vectorization) on the real-time demand data of the user for vehicle model parameters to obtain a vehicle model parameter demand feature vector.

[0057] S203: Process the real-time importance scoring data of the vehicle model attributes by the user (for example, normalization processing) to obtain the vehicle model attribute requirement weight vector.

[0058] S204: Fuse the vehicle model parameter requirement feature vector and the vehicle model attribute requirement weight vector (for example, multi-dimensional vector fusion) to obtain the user requirement twin model.

[0059] As an example, in step S201, the user fills out a questionnaire through the interaction interface. The user can fill in the requirement data for vehicle model parameters and the importance scoring data for vehicle model attributes in the questionnaire. The requirement data for the vehicle model parameters may include but are not limited to: vehicle model level preference, power type preference, minimum cruising range, maximum charging time, and purchase budget range; the importance scoring data for vehicle model attributes includes but is not limited to the importance levels of space, battery life, appearance, interior, cost performance, driving texture, and in-vehicle intelligent system, with a score of 0 - 10.

[0060] As an example, in step S202, the data in the questionnaire is sorted out and parsed. The multi-selection parameters can be converted into one-hot encoding, and the numerical parameters are normalized. The requirement data of the user for the vehicle model parameters after sorting out and parsing is written in vector form to obtain the vehicle model parameter requirement feature vector.

[0061] As an example, in step S203, the importance scoring data of the vehicle model attributes of the user after sorting out and parsing is normalized, the weights of each field are calculated, and the weights are written in vector form to obtain the vehicle model attribute requirement weight vector.

[0062] As an example, the formula for normalizing the importance scoring data of the vehicle model attributes of the user is: where is the requirement weight of the i-th vehicle model attribute, n is the total number of vehicle model attributes, is the importance scoring data of the i-th vehicle model attribute; in this embodiment, n = 7. Through normalization processing, it can be ensured that the sum of the weights of each field is 1.

[0063] As an example, in step S204, the vehicle model parameter requirement feature vector and the vehicle model attribute requirement weight vector are concatenated to obtain the user requirement twin model. The user requirement twin model integrates the requirements of the user for specific vehicle model parameters and the importance weight information of different vehicle model attributes, enabling the user requirement twin model to more comprehensively reflect the user's requirement characteristics.

[0064] In step S30, please refer toFigure 1 In step S30, based on the two-stage recommendation algorithm, the passenger car model twin model library, and the user demand twin model, a vehicle model recommendation result is obtained.

[0065] As an example, please refer to Figure 5 , step S30 may include the following steps: S301 to S304.

[0066] S301: Screen the passenger car model twin model library in a preset order to obtain candidate vehicle models.

[0067] S302: Based on the user demand twin model, obtain the vehicle attribute demand weight vector of the candidate vehicle models.

[0068] S303: Based on the passenger car model twin model, obtain the word-of-mouth feature vector of the candidate vehicle models.

[0069] S304: Calculate the weighted cosine similarity between the vehicle attribute demand weight vector and the word-of-mouth feature vector to obtain their weighted cosine similarity; sort the candidate vehicle models in descending order according to the weighted cosine similarity, and select the top several candidate vehicle models as the vehicle model recommendation result.

[0070] As an example, in step S301, when screening the passenger car model twin model library in a preset order, the preset order may but is not limited to being, in sequence, the purchase budget, the cruising range, the vehicle model level, the power type, and the charging time, and the priority of the preset order decreases in sequence.

[0071] As an example, in the process of screening the passenger car model twin model library in a preset order to obtain candidate vehicle models, if the number of candidate vehicle models is lower than a preset threshold, the restrictive conditions are cancelled in reverse order or the numerical condition thresholds are relaxed by a preset ratio until the number of candidate vehicle models reaches the preset threshold.

[0072] Specifically, after each step of screening is completed, the number of candidate vehicle models is checked in real time. If the number of candidate vehicle models is lower than the threshold x , then a dynamic relaxation mechanism is triggered until the number of obtained candidate vehicle models is greater than or equal to the threshold x .

[0073] Specifically, the dynamic relaxation mechanism may include: relaxing the screening conditions in reverse priority order, or relaxing the screening conditions of numerical parameters by a preset ratio.

[0074] As an example, the vehicle model level may include: subcompact cars, compact cars, midsize cars, large cars, etc.

[0075] As an example, the power type may include: fuel vehicles, pure electric vehicles, plug-in hybrid vehicles, etc.

[0076] As an example, for numerical parameters, the screening conditions can be expanded according to a preset ratio. For example, for the car purchase budget, the upper limit of the budget can be increased by 10%. For categorical parameters, similar categories can be merged. For example, "SUV" and "Crossover SUV" can be merged into the same category until the candidate vehicle model threshold x requirements are met.

[0077] In one example, the quantity threshold of the candidate vehicle models x can be set to 10 vehicles. First, according to the car purchase budget range set by the user, vehicle models with prices within this range are screened out from the passenger vehicle model twin model library. Among the vehicle model set after the car purchase budget screening; then, screening is performed according to the user's expected cruising range requirement; next, according to the vehicle model level preferred by the user, vehicle models that meet the level are selected from the vehicle models obtained from the previous screening; then, according to the power type expected by the user, the vehicle models are screened again, and the vehicle models that meet the power type requirement are retained; finally, for vehicle models involving charging, according to the user's acceptable charging time range, vehicle models with charging times that meet the conditions are screened out. After each screening step is completed, check whether the number of candidate vehicle models is less than 10. If the number of candidate vehicle models is less than 10, the user needs to relax the current screening conditions. For non-numerical screening conditions, the restrictive conditions of the current screening step are cancelled in turn. Specifically, after screening by the car purchase budget, 60 candidate vehicle models are obtained. Then, after screening by the cruising range, 30 candidate vehicle models remain. After screening by the vehicle model level, 12 candidate vehicle models remain. Then, after screening by the power type, 8 candidate vehicle models remain. At this time, the number of candidate vehicle models is less than 10. Since the power type is the current screening step and is a non-numerical condition, the screening restriction of the power type is relaxed. The original screening condition was only pure electric vehicles. After relaxation, it can include fuel vehicles and hybrid vehicles. After re-screening, it is found that 15 candidate vehicle models remain, meeting the threshold requirement, and the screening process ends.

[0078] As an example, in step S302, the vehicle model attribute requirement weight vector of the candidate vehicle models is extracted from the user demand twin model , where correspond to the normalized weight values of space, battery life, appearance, interior, cost performance, driving texture, and in-vehicle intelligent system respectively, that is correspond to the demand weights of seven vehicle model attributes of space, battery life, appearance, interior, cost performance, driving texture, and in-vehicle intelligent system respectively.

[0079] As an example, in step S303, the word-of-mouth feature vector of the candidate vehicle models is extracted from the passenger vehicle model twin model , where The score values corresponding to space, battery life, appearance, interior, cost performance, driving experience, and in-vehicle intelligent system respectively, that is, the word-of-mouth score value, which is also The emotional average values corresponding to seven vehicle model attributes: space, battery life, appearance, interior, cost performance, driving experience, and in-vehicle intelligent system respectively.

[0080] As an example, in step S304, the weighted cosine similarity between the vehicle model attribute demand weight vector and the word-of-mouth feature vector is calculated based on the following formula to obtain the weighted cosine similarity between the two, that is, calculate the vehicle model attribute demand weight vector of the candidate vehicle model and the word-of-mouth feature vector The weighted cosine similarity Similarity, and the expression is: where Similarity is the weighted cosine similarity between the vehicle model attribute demand weight vector and the word-of-mouth feature vector; is the demand weight of the i-th vehicle model attribute, is the emotional average value (i.e., the word-of-mouth score value) of the i-th vehicle model attribute, and n is the total number of vehicle model attributes. In this embodiment, n = 7.

[0081] Furthermore, the candidate vehicle models are sorted in descending order of similarity, and the top several candidate vehicle models are selected as the vehicle model recommendation results. Specifically, the top 10 candidate vehicle models can be selected as the vehicle model recommendation results and a recommendation list is output.

[0082] As an example, if the number of candidate vehicle models is less than 10, all vehicle models are output and the similarity rankings are marked.

[0083] In the vehicle model agile recommendation method based on the digital twin model of the present application, real-time passenger vehicle data and user comment data are obtained through web crawlers and API interfaces, and a passenger vehicle model twin model library is constructed, which can comprehensively present the actual situation and word-of-mouth of vehicle models, and provide a real and dynamic vehicle model portrait; the demand data of users for vehicle model parameters and the importance scores of vehicle model attributes are collected through real-time questionnaires, and a user demand twin model is constructed to accurately meet the personalized needs of users; vehicle models are screened in a preset order, and the similarity is calculated in combination with the multi-level weighted cosine similarity matching algorithm (that is, the formula for calculating the weighted cosine similarity in the above embodiment), which can synthesize multi-dimensional information to ensure the scientific nature of the recommendation results; the dynamic relaxation mechanism can ensure that a sufficient number of candidate vehicle models are screened out. The method of the present application can provide highly accurate, personalized and timely vehicle model recommendations for users, and greatly improve the efficiency of users' car purchase decision-making.

[0084] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the indication of the arrows, these steps are not necessarily executed in the sequence indicated by the arrows. Unless there is a clear indication in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other sequences. Moreover, at least a part of the steps in the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0085] In another embodiment, please refer to Figure 6 , the present application also provides a vehicle model agile recommendation system based on a digital twin model, which may include: a data acquisition module 1, a passenger vehicle model twin model construction module 2, a user demand twin model construction module 3, and a recommendation result generation module 4. Among them, the data acquisition module 1 is used to obtain real-time passenger vehicle data, real-time user comment data, real-time user demand data for vehicle model parameters, and real-time importance score data of vehicle model attributes for users; the passenger vehicle model twin model construction module 2 is used to construct a passenger vehicle model twin model according to the real-time passenger vehicle data and the real-time user comment data, and store the passenger vehicle model twin model in a database; the user demand twin model construction module 3 is used to construct a user demand twin model according to the demand data of the vehicle model parameters for the user and the importance score data of the vehicle model attributes for the user; the recommendation result generation module 4 is used to obtain a vehicle model recommendation result based on a two-stage recommendation algorithm, the passenger vehicle model twin model library, and the user demand twin model.

[0086] As an example, the vehicle model agile recommendation system based on a digital twin model in this embodiment can be used to execute the vehicle model agile recommendation method based on a digital twin model in the above embodiment.

[0087] As an example, the system of the present application adopts a microservices architecture, and collects data and outputs recommendation results through API interfaces.

[0088] As an example, the system can be set for timed update, and data collection, similarity calculation, and model maintenance are performed in the background at regular intervals to ensure that the recommendation system always runs based on the latest data.

[0089] In the above-mentioned vehicle model agile recommendation system based on the digital twin model, the real-time passenger vehicle data, user comment data, user demand data, and importance scoring data are obtained through the data acquisition module 1, which can timely capture market dynamics and changes in user demands, ensuring the timeliness and comprehensiveness of the recommendation basis; the passenger vehicle model twin model construction module 2 can comprehensively depict information such as the actual parameters, configurations, and word-of-mouth of vehicle models, providing accurate vehicle model portraits and making the recommendation results more in line with the actual situation of vehicle models; the user demand twin model construction module 3 can deeply explore users' personalized demands, ensuring that the recommendation direction meets users' expectations and avoiding the blindness of recommendations; the recommendation result generation module 4 can comprehensively consider multi-dimensional information, scientifically measure the matching degree between vehicle models and user demands, and make the recommendation results more reasonable and targeted. The system of this application can comprehensively, real-time, and accurately reflect the characteristics of vehicle models and users' personalized demands, improve the accuracy and timeliness of recommendations, provide users with more high-quality vehicle model recommendation services that meet their needs, and enhance users' car purchase experience and decision-making efficiency.

[0090] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0091] Although this application has been disclosed as above with embodiments, it is not intended to limit this application. Any person with ordinary knowledge in the technical field to which this application pertains may make some modifications and refinements without departing from the spirit and scope of this application. Therefore, the protection scope of this application shall be subject to that defined by the appended patent application scope.

Claims

1. A vehicle model agile recommendation method based on a digital twin model, characterized in that: The following steps are involved: Acquire real-time passenger car data and real-time user review data, and build a passenger car model twin model library according to the real-time passenger car data and the real-time user review data, wherein the passenger car model twin model library is dynamically updated according to new passenger car data and new user review data; Acquire real-time demand data of users for vehicle model parameters and real-time importance scoring data of users for vehicle model attributes, and build a user demand twin model according to the real-time demand data of users for vehicle model parameters and real-time importance scoring data of users for vehicle model attributes; The vehicle model recommendation result is obtained based on the two-stage recommendation algorithm, the passenger vehicle model twin model library and the user demand twin model.

2. The vehicle model agile recommendation method based on digital twin model according to claim 1 is characterized in that: A passenger vehicle model twin model library is constructed according to the real-time passenger vehicle data and the real-time user comment data, and the passenger vehicle model twin model library is dynamically updated according to new passenger vehicle data and new user comment data, including: Performing structured processing on the real-time passenger car data to obtain parameter characteristic data; Based on the real-time user review data, the average sentiment value of each vehicle type attribute is obtained as word-of-mouth feature data; Vectorizing the parameter feature data and the word-of-mouth feature data, and performing vector fusion on the vectorized parameter feature data and the word-of-mouth feature data to obtain the passenger vehicle model twin model; The passenger vehicle model twin models of all vehicle model attributes are stored in a database to obtain a passenger vehicle model twin model library, and the passenger vehicle model twin model library is dynamically updated according to new passenger vehicle data and new user comment data.

3. The vehicle model agile recommendation method based on digital twin model according to claim 2 is characterized in that: The real-time passenger car data is subjected to structured processing to obtain parameter characteristic data, including: numerical standardization and category coding of the real-time passenger car data to obtain the parameter characteristic data.

4. The vehicle model agile recommendation method based on digital twin model according to claim 2 is characterized in that: A user demand twin model is constructed according to the user's real-time demand data for vehicle model parameters and the user's real-time importance rating data for vehicle model attributes, including: Obtaining a vehicle model parameter demand feature vector based on the user's real-time demand data for vehicle model parameters; Normalizing the user's real-time importance score data for vehicle model attributes to obtain a vehicle model attribute demand weight vector; The vehicle model parameter requirement feature vector and the vehicle model attribute requirement weight vector are merged to obtain the user requirement twin model.

5. The vehicle model agile recommendation method based on digital twin model according to claim 4 is characterized in that: The formula for normalizing the importance score data of the user on the vehicle model attributes is: in, is the demand weight of the i-th vehicle model attribute, n is the total number of vehicle model attributes, Importance scoring data for the i-th vehicle model attribute.

6. The vehicle model agile recommendation method based on digital twin model according to claim 4 is characterized in that: The vehicle model recommendation result is obtained based on the two-stage recommendation algorithm, the passenger vehicle model twin model library and the user demand twin model, including: Screening the passenger vehicle model twin model library in sequence according to a preset order to obtain candidate vehicle models; Acquire a vehicle attribute demand weight vector of the candidate vehicle model based on the user demand twin model; Acquire a reputation feature vector of a candidate vehicle model based on the passenger vehicle model twin model; A weighted cosine similarity calculation is performed on the vehicle attribute requirement weight vector and the word-of-mouth feature vector to obtain a weighted cosine similarity between the two; the candidate vehicle models are arranged in descending order according to the weighted cosine similarity, and the top several candidate vehicle models are selected as the vehicle model recommendation results.

7. The vehicle model agile recommendation method based on digital twin model according to claim 6 is characterized in that: The preset order includes: car purchase budget, cruising range, vehicle model level, power type and charging time.

8. The vehicle model agile recommendation method based on digital twin model according to claim 6 is characterized in that: In the process of screening the passenger vehicle model twin models in sequence in a preset order to obtain candidate models, if the number of candidate models is lower than a preset threshold, the restriction conditions are canceled in reverse order or the numerical condition threshold is relaxed according to a preset ratio until the number of candidate models reaches the preset threshold.

9. The vehicle model agile recommendation method based on digital twin model according to claim 6 is characterized in that: The weighted cosine similarity of the vehicle attribute demand weight vector and the word-of-mouth feature vector is calculated based on the following formula to obtain the weighted cosine similarity between the two: Wherein, Similarity is the weighted cosine similarity between the vehicle attribute demand weight vector and the word-of-mouth feature vector; is the demand weight of the i-th vehicle model attribute, is the average sentiment value of the i-th vehicle model attribute, and n is the total number of vehicle model attributes.

10. A vehicle model agile recommendation system based on a digital twin model, characterized in that: include: The data collection module is used to obtain real-time passenger car data, real-time user comment data, real-time user demand data for vehicle model parameters, and real-time importance rating data of users on vehicle model attributes; A passenger vehicle twin model building module, used to build a passenger vehicle twin model according to the real-time passenger vehicle data and the real-time user comment data, and store the passenger vehicle twin model in a database; A user demand twin model construction module is used to construct a user demand twin model based on the user's demand data for vehicle model parameters and the user's importance score data for vehicle model attributes; A recommendation result generation module is used to obtain vehicle model recommendation results based on a two-stage recommendation algorithm, the passenger vehicle model twin model library and the user demand twin model.

Images