Dynamic car rental recommendation method based on multi-dimensional user portraits

Abstract

The invention provides a dynamic car rental recommendation method based on a multi-dimensional user portrait, and belongs to the technical field of car rental recommendation. The method integrates multi-source heterogeneous data of a ferry ticketing system, a car rental platform and a third-party data source, constructs a dynamic multi-dimensional user portrait model, and combines a user portrait tag library and real-time scene parameters to obtain a dynamic car rental recommendation result. A personalized car rental recommendation list is generated by adopting a service matching engine fusing a collaborative filtering algorithm and a weighted scoring strategy, a recommendation result is output through at least one of a short message, an applet message push system or an intelligent voice outbound system according to the historical touch preference of a user, and user feedback data is collected to optimize model parameters. According to the method, the problems of data islanding, single portrait dimension and dynamic mismatch between the service and the demand in the prior art are solved, and the matching efficiency of the car rental service and the user satisfaction are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of smart port technology, and in particular to a dynamic car rental recommendation method based on multi-dimensional user portraits. Background Art

[0002] Currently, the Qiongzhou Strait, as an important channel connecting the Hainan Free Trade Port with the inland, not only faces the key challenge of intelligent transformation of transportation services, but also has the following problems:

[0003] 1. The existing system faced bottlenecks in cross-platform and cross-scenario data integration. The API connection between the ferry ticketing system and the service platform was inefficient, resulting in a discrete distribution of user behavior data. The utilization rate of passenger travel characteristics was low, making it impossible to establish a complete travel demand forecasting model.

[0004] 2. The defects of single and static portrait dimensions. Traditional user portraits only rely on basic attributes (such as age, gender, etc.) and shallow historical records. There are problems such as the lack of dynamic spatiotemporal dimensions, insufficient accuracy of preference calculation, and difficulty in multimodal data fusion.

[0005] 3. There is a dynamic mismatch between services and demand, and digital services are insufficiently deep. The service recommendations for passengers are single and highly overlapping, failing to reflect differentiated service characteristics. This leads to low service efficiency and a waste of service resources.

[0006] Therefore, a dynamic car rental recommendation method based on multi-dimensional user portraits is proposed to solve the above problems. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a dynamic car rental recommendation method based on multi-dimensional user portraits to at least solve the above problems.

[0008] The technical solution adopted in the present invention is as follows:

[0009] A dynamic car rental recommendation method based on multi-dimensional user portraits, the method comprising the following steps:

[0010] S1. Data integration: Real-time access to multi-source heterogeneous data from ferry ticketing systems, car rental platforms, and third-party data sources through standardized API interfaces. The data includes at least user identity information, historical order records, voyage trajectories, and real-time traffic status data.

[0011] S2. Profile Modeling: Based on cleaned multi-source data, a dynamic user profile model is constructed, which includes a three-layer labeling system consisting of basic attributes, behavioral characteristics, and scenario preferences. The scenario preference layer uses a clustering algorithm to identify user travel scenario characteristics.

[0012] S3, Recommendation Generation: Combining the user portrait tag library with real-time scenario parameters, a service matching engine that integrates collaborative filtering algorithms and weighted scoring strategies is used to generate a personalized car rental recommendation list;

[0013] S4. Service reach: Based on the user's historical reach preferences, the recommendation results are output through at least one of SMS, mini-program message push, or intelligent voice outbound call system, and user feedback data is collected to optimize model parameters.

[0014] Furthermore, in the dynamic user portrait model:

[0015] The basic attribute layer includes user age segments and traveler statistics extracted from real-name data, as well as vehicle model preference quantitative indicators, price sensitivity classification, and cross-sea time preference identification results derived from historical order analysis;

[0016] The behavioral feature layer includes peak travel frequency statistics and cross-platform service usage analysis based on time series analysis;

[0017] The scenario preference layer uses a K-means clustering algorithm to identify typical features of at least family travel scenarios and business travel scenarios, where the basis for determining a family travel scenario is that the companions include children ≤ 12 years old or elderly people ≥ 65 years old and the historical car rental orders contain safety seat rental records.

[0018] Furthermore, the price sensitivity rating is calculated using the following quantitative formula:

[0019] Price sensitivity = 0.4 × (discount amount / total amount) + 0.4 × economy car rental ratio + 0.3 × (1-additional service selection rate)

[0020] The calculation results are divided into three levels according to the percentile of the user group: high sensitivity, medium sensitivity and low sensitivity, among which the top 20% are high-sensitive users, the middle 60% are medium-sensitive users, and the bottom 20% are low-sensitive users.

[0021] Furthermore, in the recommendation generation step, the car rental service recommendation priority Score is determined by the following linear weighted formula:

[0022] Score=α·S user +β·S service +γ·S context

[0023] Among them, the user portrait similarity uses cosine similarity to calculate the matching degree of the feature vectors of the query user and the service frequent customers. The cosine similarity calculation formula is:

[0024]

[0025] Among them, Uq and U s are the feature vectors of query users and service frequent customers respectively;

[0026] The service resource saturation S_service is defined as 1-(current load / service capacity), with a value range of [0,1];

[0027] Real-time scene parameters are dynamically adjusted according to the environment status:

[0028]

[0029] The basic value S base The algorithm is calculated based on dynamic factors such as the real-time supply-demand ratio and geographic proximity, and the weight coefficient satisfies α+β+γ=1, and is continuously optimized through offline reinforcement learning strategies.

[0030] Furthermore, it also includes the steps of car rental demand forecasting:

[0031] The LSTM-based time series prediction model is used. The input feature matrix includes the historical order volume, port throughput, and holiday factors within the T time window. The calculation formula of the LSTM time series prediction model is:

[0032]

[0033] Among them, the input feature matrix It contains the time series observations of historical order volume, port throughput and holiday factors, time window length T, feature dimension d, future period Δ, LSTM hidden layer dimension is set to 64, parameter set Θ is optimized by time series back propagation, and the output layer is optimized by weight matrix and the bias term b o Mapping the hidden state to the future car rental demand forecast The training goal is to minimize the mean square error between the predicted value and the true value:

[0034]

[0035] in, is the mean square error, which is used to measure the average square deviation between the predicted value and the true value; N is the total number of training samples, which is used to average the errors of all samples when calculating the loss; y i is the actual car rental demand of the i-th sample, that is, the actual observation value; is the model's predicted car rental demand for the i-th sample, which is calculated from the hidden state through the weight matrix and bias term;

[0036] The model outputs the predicted value of car rental demand in the future period Δ, and optimizes the network parameters by minimizing the mean square error loss function between the predicted value and the true value.

[0037] Furthermore, in the service access step:

[0038] For family travel scenarios, users will be given priority in recommending models equipped with safety seats, and the recommendation weight of four-wheel drive SUV models will be automatically increased in rainy weather conditions;

[0039] For users in business travel scenarios, we match them with high-end car models with in-car WiFi and business reception insurance, and provide them with quick car pick-up and return services at airports / high-speed rail stations.

[0040] Furthermore, the portrait modeling step also includes a dynamic update mechanism:

[0041] A time-decay weight model is used, with a weight of 0.8 assigned to data from the past three months and a weight of 0.5 assigned to data from 3-6 months;

[0042] Real-time updates are triggered by event-driven methods. When a user places a new ferry ticket order, portrait recalculation is immediately initiated, and scene labels are added or adjusted.

[0043] Furthermore, the identification of the family travel scenario further includes:

[0044] Indicate the children / elderly passengers' travel status by their age in the ferry booking;

[0045] Users whose safety seat rental records account for ≥20% of their historical orders are marked as users with strong family travel characteristics.

[0046] Furthermore, the identification of the business travel scenario further includes:

[0047] For users who have taken ferry trips ≥ 2 times per month within six months and rent high-end cars;

[0048] Link their workday attendance records and business hospitality insurance purchase records for joint verification.

[0049] Furthermore, it is characterized in that the vehicle type preference quantitative index is calculated by the following formula:

[0050] Vehicle model preference = Σ(number of historical leases of the vehicle model × time decay coefficient) / total number of leases, where the time decay coefficient is weighted as 0.9 for 1 year, 0.6 for 1-2 years, and 0.3 for more than 2 years.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. Breakthrough in the efficiency of global data fusion

[0053] Real-time data connection across business systems is achieved, and user feature dimensions are expanded.

[0054] 2. Dynamic portrait accuracy improved

[0055] Demand forecast errors are reduced, the timeliness of user preferences is enhanced, and the recognition rate of typical scenarios reaches a new high.

[0056] 3. Revolutionary improvement in resource matching efficiency

[0057] The response time for emergency dispatch has been shortened, the inter-trip service guarantee rate has made breakthrough progress, and the vehicle-ship service chain is seamlessly connected.

[0058] 4. Deeply tap the potential of scenario-based services

[0059] The differences in recommendations for home / business scenarios have improved user satisfaction, and the service coverage for extended trips has been extended. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0061] Figure 1 This is a schematic diagram of the overall structure of a dynamic car rental recommendation method based on multi-dimensional user portraits proposed in an embodiment of the present invention.

[0062] Figure 2 This is a data flow diagram of a dynamic car rental recommendation method based on multi-dimensional user portraits proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of the present invention more apparent, exemplary embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described in the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present invention.

[0064] In the following description, numerous specific details are provided to provide a more thorough understanding of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced without one or more of these details. In other instances, certain technical features well known in the art are not described to avoid confusion with the present invention.

[0065] It should be understood that the present invention can be implemented in different forms and should not be interpreted as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to make disclosure thorough and complete and to fully convey the scope of the present invention to those skilled in the art.

[0066] The purpose of the terms used herein is only to describe specific embodiments and is not intended to limit the present invention. When used herein, the singular forms "a", "an", and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "comprising" and / or "comprising", when used in this specification, determine the presence of the features, integers, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts and / or groups. When used herein, the term "and / or" includes any and all combinations of the relevant listed items.

[0067] In order to fully understand the present invention, a detailed structure will be provided in the following description to illustrate the technical solution proposed by the present invention. Optional embodiments of the present invention are described in detail below. However, in addition to these detailed descriptions, the present invention may also have other implementations.

[0068] Reference Figure 1-2 The present invention provides a dynamic car rental recommendation method based on multi-dimensional user portraits, characterized in that the method comprises the following steps:

[0069] S1. Data integration: Real-time access to multi-source heterogeneous data from ferry ticketing systems, car rental platforms, and third-party data sources through standardized API interfaces. The data includes at least user identity information, historical order records, voyage trajectories, and real-time traffic status data.

[0070] S2. Profile Modeling: Based on cleaned multi-source data, a dynamic user profile model is constructed, which includes a three-layer labeling system consisting of basic attributes, behavioral characteristics, and scenario preferences. The scenario preference layer uses a clustering algorithm to identify user travel scenario characteristics.

[0071] S3, Recommendation Generation: Combining the user portrait tag library with real-time scenario parameters, a service matching engine that integrates collaborative filtering algorithms and weighted scoring strategies is used to generate a personalized car rental recommendation list;

[0072] S4. Service reach: Based on the user's historical reach preferences, the recommendation results are output through at least one of SMS, mini-program message push, or intelligent voice outbound call system, and user feedback data is collected to optimize model parameters.

[0073] For example, a ferry ticketing system can call its order query interface to obtain the departure point (e.g., Xuwen Port), destination (e.g., Haikou New Seaport), flight time, and vehicle type (gasoline / new energy vehicle) entered by the user when purchasing the ticket. A car rental platform can obtain a user's historical car rental records, including the type of vehicle (e.g., 5-seater sedan / 7-seater SUV), rental duration, and additional services (child seats, insurance type). Third-party data can be connected to the AutoNavi Map API to obtain real-time traffic status. Data integration can integrate the original data of the "Ferry Butler" ticketing system and car rental platform through the API interface, covering user identity information, historical orders, voyage trajectories, etc. Portrait modeling can use the Spark distributed computing framework to clean and normalize multi-source heterogeneous data, and generate a dynamic tag library through spatiotemporal feature extraction algorithms and classification models. Service matching can be based on the car rental vehicle feature library (model, price, service type) and real-time scene parameters (weather, port traffic), and use collaborative filtering algorithms to perform weighted matching of service resources. Marketing reach can integrate SMS gateways, mini-program message push, and intelligent voice outbound call systems to push services according to user reach preferences and collect feedback data to form a closed-loop optimization. User feedback data includes user praise rate, recommendation acceptance rate and vehicle turnover rate.

[0074] In the dynamic user portrait model:

[0075] The basic attribute layer includes user age segments and traveler statistics extracted from real-name data, as well as vehicle model preference quantitative indicators, price sensitivity classification, and cross-sea time preference identification results derived from historical order analysis;

[0076] The behavioral feature layer includes peak travel frequency statistics and cross-platform service usage analysis based on time series analysis;

[0077] The scenario preference layer uses a K-means clustering algorithm to identify typical features of at least family travel scenarios and business travel scenarios, where the basis for determining a family travel scenario is that the companions include children ≤ 12 years old or elderly people ≥ 65 years old and the historical car rental orders contain safety seat rental records.

[0078] For example, the basic attribute layer can also obtain the user's native place, travel purpose and historical car rental records. For the user's native place and travel purpose, the departure and destination information filled in when purchasing the ferry ticket can be combined with historical orders to analyze the user's geographical distribution (for example, most tourists outside the island come from Guangdong, Beijing, etc.), and infer the purpose of tourism, visiting relatives or business trips. For historical car rental records, the user's past orders on the "Ferry Butler" platform or the car rental platform can be integrated to extract the model preference (calculated based on the historical rental record of the model and time weighting), rental period, price sensitivity (based on the use of coupons, purchase of additional services, and The characteristics such as the weighted proportion of economic vehicle models) and the tendency to use new energy vehicles are obtained: based on whether the user has chosen new energy special ship transportation (such as the "Green Source") or whether the user has purchased a new energy car ticket, new energy rental vehicles and charging services on the island are recommended. The identification of the time preference for crossing the sea can be combined with the ferry flight reservation time (such as the Spring Festival return peak is concentrated from the 5th to the 8th day of the first lunar month), to predict the time period of car rental demand after the user arrives at the port, extract high-frequency rental period segments (such as weekend travel of less than 3 days, long holiday travel of more than 7 days), establish a rental period template library for holidays such as the Spring Festival / National Day, and identify whether the user has holiday car rental behavior.

[0079] The price sensitivity rating is calculated using the following quantitative formula:

[0080] Price sensitivity = 0.4 × (discount amount / total amount) + 0.4 × economy car rental ratio + 0.3 × (1-additional service selection rate)

[0081] The calculation results are divided into three levels according to the percentile of the user group: high sensitivity, medium sensitivity and low sensitivity, among which the top 20% are high-sensitive users, the middle 60% are medium-sensitive users, and the bottom 20% are low-sensitive users.

[0082] In the recommendation generation step, the car rental service recommendation priority Score is determined by the following linear weighted formula:

[0083] Score=α·S user +β·S service +γ·S context

[0084] Among them, the user portrait similarity uses cosine similarity to calculate the matching degree of the feature vectors of the query user and the service frequent customers. The cosine similarity calculation formula is:

[0085]

[0086] Among them, U q and U s are the feature vectors of query users and service frequent customers respectively;

[0087] The service resource saturation S_service is defined as 1-(current load / service capacity), with a value range of [0,1];

[0088] Real-time scene parameters are dynamically adjusted according to the environment status:

[0089]

[0090] The basic value S base The algorithm is calculated based on dynamic factors such as the real-time supply-demand ratio and geographic proximity, and the weight coefficient satisfies α+β+γ=1, and is continuously optimized through offline reinforcement learning strategies.

[0091] It also includes the steps of car rental demand forecasting:

[0092] The LSTM-based time series prediction model is used. The input feature matrix includes the historical order volume, port throughput, and holiday factors within the T time window. The calculation formula of the LSTM time series prediction model is:

[0093]

[0094] Among them, the input feature matrix It contains the time series observations of historical order volume, port throughput and holiday factors, time window length T, feature dimension d, future period Δ, LSTM hidden layer dimension is set to 64, parameter set Θ is optimized by time series back propagation, and the output layer is optimized by weight matrix and the bias term b o Mapping the hidden state to the future car rental demand forecast The training goal is to minimize the mean square error between the predicted value and the true value:

[0095]

[0096] in, is the mean square error, which is used to measure the average square deviation between the predicted value and the true value; N is the total number of training samples, which is used to average the errors of all samples when calculating the loss; y i is the actual car rental demand of the i-th sample, that is, the actual observation value; is the model's predicted car rental demand for the i-th sample, which is calculated from the hidden state through the weight matrix and bias term;

[0097] The model outputs the predicted value of car rental demand in the future period Δ, and optimizes the network parameters by minimizing the mean square error loss function between the predicted value and the true value.

[0098] In the service access steps:

[0099] For family travel scenarios, users will be given priority in recommending models equipped with safety seats, and the recommendation weight of four-wheel drive SUV models will be automatically increased in rainy weather conditions;

[0100] For users in business travel scenarios, we match them with high-end car models with in-car WiFi and business reception insurance, and provide them with quick car pick-up and return services at airports / high-speed rail stations.

[0101] The portrait modeling step also includes a dynamic update mechanism:

[0102] A time-decay weight model is used, with a weight of 0.8 assigned to data from the past three months and a weight of 0.5 assigned to data from 3-6 months;

[0103] Real-time updates are triggered by event-driven methods. When a user places a new ferry ticket order, portrait recalculation is immediately initiated, and scene labels are added or adjusted.

[0104] The identification of the family travel scenario further includes:

[0105] Indicate the children / elderly passengers' travel status by their age in the ferry booking;

[0106] Users whose safety seat rental records account for ≥20% of their historical orders are marked as users with strong family travel characteristics.

[0107] The identification of the business travel scenario further includes:

[0108] For users who have taken ferry trips ≥ 2 times per month within six months and rent high-end cars;

[0109] Link their workday attendance records and business hospitality insurance purchase records for joint verification.

[0110] The vehicle model preference quantitative index is calculated using the following formula:

[0111] Vehicle model preference = Σ(number of historical leases of the vehicle model × time decay coefficient) / total number of leases, where the time decay coefficient is weighted as 0.9 for 1 year, 0.6 for 1-2 years, and 0.3 for more than 2 years.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A dynamic car rental recommendation method based on multi-dimensional user portraits, characterized in that: The method comprises the following steps: S1. Data integration: Real-time access to multi-source heterogeneous data from ferry ticketing systems, car rental platforms, and third-party data sources through standardized API interfaces. The data includes at least user identity information, historical order records, voyage trajectories, and real-time traffic status data. S2. Profile Modeling: Based on cleaned multi-source data, a dynamic user profile model is constructed, which includes a three-layer labeling system consisting of basic attributes, behavioral characteristics, and scenario preferences. The scenario preference layer uses a clustering algorithm to identify user travel scenario characteristics. S3, Recommendation Generation: Combining the user portrait tag library with real-time scenario parameters, a service matching engine that integrates collaborative filtering algorithms and weighted scoring strategies is used to generate a personalized car rental recommendation list; S4. Service reach: Based on the user's historical reach preferences, the recommendation results are output through at least one of SMS, mini-program message push, or intelligent voice outbound call system, and user feedback data is collected to optimize model parameters.

2. The method according to claim 1, characterized in that In the dynamic user portrait model: The basic attribute layer includes user age segments and traveler statistics extracted from real-name data, as well as vehicle model preference quantitative indicators, price sensitivity classification, and cross-sea time preference identification results derived from historical order analysis; The behavioral feature layer includes peak travel frequency statistics and cross-platform service usage analysis based on time series analysis; The scenario preference layer uses a K-means clustering algorithm to identify typical features of at least family travel scenarios and business travel scenarios, where the basis for determining a family travel scenario is that the companions include children ≤ 12 years old or elderly people ≥ 65 years old and the historical car rental orders contain safety seat rental records.

3. The method according to claim 2, characterized in that The price sensitivity rating is calculated using the following quantitative formula: Price sensitivity = 0.4 × (discount amount / total amount) + 0.4 × economy car rental ratio + 0.3 × (1-additional service selection rate) The calculation results are divided into three levels according to the percentile of the user group: high sensitivity, medium sensitivity and low sensitivity, among which the top 20% are high-sensitive users, the middle 60% are medium-sensitive users, and the bottom 20% are low-sensitive users.

4. The method according to claim 1, wherein In the recommendation generation step, the car rental service recommendation priority Score is determined by the following linear weighted formula: Score=α·S user +β·S service +γ·S context Among them, the user portrait similarity uses cosine similarity to calculate the matching degree of the feature vectors of the query user and the service frequent customers. The cosine similarity calculation formula is: Among them, U q and U s are the feature vectors of query users and service frequent customers respectively; The service resource saturation S_service is defined as 1-(current load / service capacity), with a value range of [0,1]; Real-time scene parameters are dynamically adjusted according to the environment status: The basic value S base The algorithm is calculated based on dynamic factors such as the real-time supply-demand ratio and geographic proximity, and the weight coefficient satisfies α+β+γ=1, and is continuously optimized through offline reinforcement learning strategies.

5. The method according to claim 1, characterized in that It also includes the steps of car rental demand forecasting: The LSTM-based time series prediction model is used. The input feature matrix includes the historical order volume, port throughput, and holiday factors within the T time window. The calculation formula of the LSTM time series prediction model is: Among them, the input feature matrix It contains the time series observations of historical order volume, port throughput and holiday factors, time window length T, feature dimension d, future period Δ, LSTM hidden layer dimension is set to 64, parameter set Θ is optimized by time series back propagation, and the output layer is optimized by weight matrix and the bias term b o Mapping the hidden state to the future car rental demand forecast The training goal is to minimize the mean square error between the predicted value and the true value: in, is the mean square error, which is used to measure the average square deviation between the predicted value and the true value; N is the total number of training samples, which is used to average the errors of all samples when calculating the loss; y i is the actual car rental demand of the i-th sample, that is, the actual observation value; is the model's predicted car rental demand for the i-th sample, which is calculated from the hidden state through the weight matrix and bias term; The model outputs the predicted value of car rental demand in the future period Δ, and optimizes the network parameters by minimizing the mean square error loss function between the predicted value and the true value.

6. The method according to claim 1, characterized in that In the service access steps: For family travel scenarios, users will be given priority in recommending models equipped with safety seats, and the recommendation weight of four-wheel drive SUV models will be automatically increased in rainy weather conditions; For users in business travel scenarios, we match them with high-end car models with in-car WiFi and business reception insurance, and provide them with quick car pick-up and return services at airports / high-speed rail stations.

7. The method according to claim 1, characterized in that The portrait modeling step also includes a dynamic update mechanism: A time-decay weight model is used, with a weight of 0.8 assigned to data from the past three months and a weight of 0.5 assigned to data from 3-6 months; Real-time updates are triggered by event-driven methods. When a user places a new ferry ticket order, portrait recalculation is immediately initiated, and scene labels are added or adjusted.

8. The method according to claim 2, characterized in that The identification of the family travel scenario further includes: Indicate the children / elderly passengers' travel status by their age in the ferry booking; Users whose safety seat rental records account for ≥20% of their historical orders are marked as users with strong family travel characteristics.

9. The method according to claim 2, characterized in that The identification of the business travel scenario further includes: For users who have taken ferry trips ≥ 2 times per month within six months and rent high-end cars; Link their workday attendance records and business hospitality insurance purchase records for joint verification.

10. The method according to claim 2, characterized in that The vehicle model preference quantitative index is calculated using the following formula: Vehicle model preference = Σ(number of historical leases of the vehicle model × time decay coefficient) / total number of leases, where the time decay coefficient is weighted as 0.9 for 1 year, 0.6 for 1-2 years, and 0.3 for more than 2 years.

Images