Training method and device of second-hand car comprehensive profitability evaluation model

Abstract

The invention discloses a second-hand vehicle comprehensive profitability evaluation model training method and device, and the method comprises the steps: determining a second-hand vehicle comprehensive profitability evaluation model, and carrying out the training; the training process specifically comprises the following steps: acquiring historical transaction data of a target second-hand vehicle, wherein the historical transaction data comprises multi-dimensional feature data and corresponding service result label data; constructing a training sample according to the multi-dimensional feature data and the corresponding service result label data; and iteratively optimizing the weight coefficient of each dimension and the initial parameter of the scoring reference offset item in the second-hand vehicle comprehensive profitability evaluation model by adopting a machine learning algorithm of supervised learning according to the training sample. According to the method, modeling is performed on the second-hand vehicle profitability evaluation through five dimensions of price, supply and demand, vehicle condition, new vehicle and business experience, training is performed, and the trained second-hand vehicle comprehensive profitability evaluation model can perform quantitative evaluation on the second-hand vehicle profitability more comprehensively and reliably.

Description

Technical Field

[0001] This application relates to the field of used car profitability assessment technology, specifically to a training method and apparatus for a comprehensive used car profitability assessment model. Background Technology

[0002] With the continuous growth in car ownership and the shortening of consumers' car replacement cycle, the used car market has shown rapid expansion in recent years. Against this backdrop, how to scientifically and systematically assess the profitability of used vehicles has become a key aspect for car dealers, financial leasing companies, and trading platforms to optimize inventory management, formulate pricing strategies, and control operational risks.

[0003] Currently, used car dealers rely mainly on personal experience and subjective judgment to assess the profitability of vehicles during the procurement and sales process. The factors they usually consider include: the market price of similar models, an intuitive assessment of the vehicle's condition, and a subjective judgment of market demand (supply and demand).

[0004] While some e-commerce platforms offer price reference functions, these are limited to single-dimensional valuations, such as estimates based on vehicle condition or market listing price. Therefore, existing methods for assessing the profitability of used vehicles have the following shortcomings:

[0005] Highly subjective: It relies mainly on the experience of sales or purchasing personnel, resulting in unstable outcomes.

[0006] One-dimensional approach: focusing only on price or vehicle condition, ignoring supply and demand changes and market dynamics.

[0007] Lack of data integration and quantitative scoring: A unified Z-score standardized evaluation system has not been formed among the various indicators.

[0008] Difficult to update dynamically: Existing methods cannot automatically adjust profit forecasts according to market conditions.

[0009] Therefore, there is an urgent need for a systematic, multi-dimensional, and quantifiable comprehensive profitability assessment model to provide used car dealers with a scientific basis for decision-making, thereby supporting the refined operation and high-quality development of the used car business. Summary of the Invention

[0010] To address this, this application provides a training method and apparatus for a comprehensive profitability assessment model for used cars, thereby solving the problems of strong subjectivity, single dimension, and lack of data fusion and quantitative scoring in existing used car profitability assessment methods.

[0011] To achieve the above objectives, this application provides the following technical solution:

[0012] Firstly, a training method for a comprehensive profitability assessment model for used cars includes:

[0013] The comprehensive profitability assessment model for used cars is determined as follows:

[0014]

[0015] in, This indicates the overall profitability score for used cars. Indicates the weight of the price dimension. This indicates the score in the price dimension. Indicates the weight of supply and demand dimensions. This indicates the score based on the supply and demand dimension. Indicates the weight of the vehicle condition dimension. This indicates the score for the vehicle condition dimension. Indicates the weight of the new car dimension. This indicates the score for the new car dimension. Indicates the empirically adjusted dimension weights. This represents the score of the experience-corrected dimension. Indicates the scoring benchmark bias;

[0016] The training process for the comprehensive profitability assessment model for used cars is as follows:

[0017] Obtain historical transaction data of the target used car. The historical transaction data includes multi-dimensional feature data and corresponding business result tag data. The multi-dimensional feature data includes price dimension data, supply and demand dimension data, vehicle condition dimension data, new car dimension data, and experience correction dimension data.

[0018] Training samples are constructed based on the multi-dimensional feature data and the corresponding business result label data;

[0019] Based on the training samples, a supervised learning machine learning algorithm is used to iteratively optimize the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model for used cars.

[0020] Preferably, the formula for calculating the price dimension score is as follows:

[0021]

[0022] in, Z represents the sensitivity adjustment coefficient for the price dimension. standardization, Indicates the target price for the vehicle. This indicates the market price of vehicles. This represents the benchmark bias for the price dimension.

[0023] Preferably, the formula for calculating the supply and demand dimension score is as follows:

[0024]

[0025] in, Indicates the demand score. Z represents the sensitivity adjustment coefficient for the demand dimension. standardization, This represents the average number of views of the target vehicle listing over a certain period of time. This represents the average number of page views of other car listings of the same model year as the target car within a certain period. This indicates the scoring benchmark bias item for the demand dimension; Indicates the supply score. Z represents the sensitivity adjustment coefficient in the supply dimension. standardization, This indicates the number of vehicles of the same model as the target vehicle currently on sale. This indicates the number of other models in the same series as the target vehicle currently on sale. This represents the scoring benchmark bias term for the supply dimension.

[0026] Preferably, the formula for calculating the vehicle condition dimension score is as follows:

[0027]

[0028] in, The score indicates the vehicle's age. Z represents the sensitivity adjustment coefficient for the vehicle age dimension. standardization, Indicates the age of the target vehicle. This indicates the age of other vehicles of the same model available in the market. This indicates the baseline bias for the vehicle age dimension. Indicates mileage score, Z represents the sensitivity adjustment coefficient for the mileage dimension. standardization, Indicates the target vehicle mileage. This indicates the mileage of other vehicles of the same model available in the market. This represents the benchmark bias term for the mileage dimension.

[0029] Preferably, the formula for calculating the new vehicle dimension score is as follows:

[0030]

[0031] in, Z represents the sensitivity adjustment coefficient for the new car dimension. standardization, This indicates the new car price corresponding to the target vehicle. Indicates the price of the target car. This indicates the baseline bias for the new vehicle dimension.

[0032] Preferably, the formula for calculating the experience-corrected dimension score is:

[0033]

[0034] in, This represents the sensitivity adjustment coefficient for the experience-corrected dimension. This represents the business experience index provided by business experts.

[0035] Preferably, when iteratively optimizing the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the used car comprehensive profitability assessment model using a supervised learning machine learning algorithm based on the training samples, the process specifically includes:

[0036] The output value of the comprehensive profitability assessment model for used cars is used as the predicted value, and the business result label data is used as the actual label value.

[0037] A loss function is constructed based on the predicted value and the true label value;

[0038] The machine learning algorithm iteratively optimizes the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model of used cars based on the loss function.

[0039] Preferably, the machine learning algorithm employs one or more of a multinomial regression model, a decision tree model, or a gradient boosting tree.

[0040] Preferably, the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model for used cars are set based on business experience.

[0041] Secondly, a training device for a comprehensive profitability assessment model for used cars includes:

[0042] The model determination module is used to determine the comprehensive profitability assessment model for used cars. The comprehensive profitability assessment model for used cars is as follows:

[0043]

[0044] in, This indicates the overall profitability score for used cars. Indicates the weight of the price dimension. This indicates the score in the price dimension. Indicates the weight of supply and demand dimensions. This indicates the score based on the supply and demand dimension. Indicates the weight of the vehicle condition dimension. This indicates the score for the vehicle condition dimension. Indicates the weight of the new car dimension. This indicates the score for the new car dimension. Indicates the empirically adjusted dimension weights. This represents the score of the experience-corrected dimension. Indicates the scoring benchmark bias;

[0045] The model training module is used to train the comprehensive profitability assessment model for used cars; the training process of the comprehensive profitability assessment model for used cars is as follows:

[0046] Obtain historical transaction data of the target used car. The historical transaction data includes multi-dimensional feature data and corresponding business result tag data. The multi-dimensional feature data includes price dimension data, supply and demand dimension data, vehicle condition dimension data, new car dimension data, and experience correction dimension data.

[0047] Training samples are constructed based on the multi-dimensional feature data and the corresponding business result label data;

[0048] Based on the training samples, a supervised learning machine learning algorithm is used to iteratively optimize the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model for used cars.

[0049] Compared with the prior art, this application has at least the following beneficial effects:

[0050] This application provides a training method for a comprehensive profitability assessment model for used cars, including: determining the comprehensive profitability assessment model for used cars and training it; the training process specifically involves: acquiring historical transaction data of the target used car, including multi-dimensional feature data and corresponding business result label data; constructing training samples based on the multi-dimensional feature data and corresponding business result label data; and using a supervised learning machine learning algorithm based on the training samples to iteratively optimize the initial parameters of the weight coefficients of each dimension and the bias term of the scoring benchmark in the comprehensive profitability assessment model for used cars. This application models and trains a comprehensive profitability assessment model for used cars based on five dimensions: price, supply and demand, vehicle condition, new car, and business experience. The trained comprehensive profitability assessment model for used cars can more comprehensively and reliably quantify the profitability of used cars, overcoming the problems of traditional models that rely on experience-based judgment, have single dimensions, and lack data fusion and quantitative scoring. Attached Figure Description

[0051] To more intuitively illustrate the prior art and this application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary drawings, those skilled in the art are able to easily make conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0052] Figure 1 This is a flowchart illustrating a training method for a comprehensive profitability assessment model for used cars, as provided in Embodiment 1 of this application. Detailed Implementation

[0053] The present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0054] In the description of this application: unless otherwise stated, "a plurality of" means two or more. The terms "first," "second," "third," etc., in this application are intended to distinguish the objects referred to and do not have any special meaning in terms of technical connotation (e.g., they should not be construed as an emphasis on importance or order). Expressions such as "including," "comprising," and "having" also mean "not limited to" (certain units, components, materials, steps, etc.).

[0055] The terms used in this application, such as "upper," "lower," "left," "right," and "middle," are generally used to indicate the general relative positional relationship for the purpose of intuitive understanding by referring to the accompanying drawings, and are not absolute limitations on the positional relationship in the actual product.

[0056] Example 1

[0057] This embodiment provides a training method for a comprehensive profitability assessment model for used cars. This method models the profitability of used cars across five dimensions: price, supply and demand, vehicle condition, comparison with new cars, and business experience. This model guides procurement decisions, sales strategies, and inventory management, thereby improving overall operational efficiency and profit margins. Specifically, used car profitability refers to the expected unit profit achievable from the time a used car is purchased until it is sold, taking into account vehicle price, vehicle condition, supply and demand, comparison with new cars, and experience factors.

[0058] This embodiment provides a training method for a comprehensive profitability assessment model for used cars, including:

[0059] S1: The comprehensive profitability assessment model for used cars is determined as follows:

[0060]

[0061] in, This indicates the overall profitability score for used cars. Indicates the weight of the price dimension. This indicates the score in the price dimension. Indicates the weight of supply and demand dimensions. This indicates the score based on the supply and demand dimension. Indicates the weight of the vehicle condition dimension. This indicates the score for the vehicle condition dimension. Indicates the weight of the new car dimension. This indicates the score for the new car dimension. Indicates the empirically adjusted dimension weights. This represents the score of the experience-corrected dimension. This indicates the scoring benchmark bias.

[0062] Specifically, in terms of price:

[0063] This embodiment compares and analyzes vehicle prices from both internal and market perspectives. Price characteristic data includes: the highest, lowest, and average prices of the same model of vehicle operated within the enterprise, both within the province and nationwide; the highest, lowest, and average prices of the same model currently on the market; a price comparison analysis of "similar vehicles" currently on the market [referring to comparable vehicles within the range of key parameters (e.g., age difference ≤ 1.5 years, mileage difference ≤ 15,000 km)] (age ± 1.5 years, mileage ± 15,000 km), and calculation of a cost-effectiveness correction factor; and a comparison of transaction data over the past three months (highest, lowest, and average prices of the same model). The price data is then processed through... The standardized score (i.e., the price dimension score) is obtained as follows:

[0064]

[0065] in, Z represents the sensitivity adjustment coefficient for the price dimension. Standardization, that is Indicates the use of Standardization principles will According to the whole The distribution of values ​​(mean and standard deviation) is linearly transformed and mapped to a standard score within the interval of 0 to 100. Indicates the target price for the vehicle. This indicates the market price of vehicles. This represents the benchmark bias for the price dimension.

[0066] Supply and demand dimensions:

[0067] Supply-side indicators are based on the number of vehicles in the company's inventory (same model, year, and series) and the number of vehicles on the market (same model, year, and series). Demand-side indicators include: a comparison of the maximum and average number of users acquired by the target vehicle and the number of users acquired by the same model over the past 3 weeks; ranking of average daily user retention over the past 3 weeks; and ranking of page views. Therefore, the formula for calculating the supply and demand dimension scores is as follows:

[0068]

[0069] in, Indicates the demand score. Z represents the sensitivity adjustment coefficient for the demand dimension. standardization, This represents the average number of views of the target vehicle listing over a certain period of time. This represents the average number of page views (21 days) of other car listings of the same model year as the target car. This indicates the scoring benchmark bias item for the demand dimension; Indicates the supply score. Z represents the sensitivity adjustment coefficient in the supply dimension. standardization, This indicates the number of vehicles of the same model (brand, model, engine displacement, and main configuration) available for sale as the target vehicle. This indicates the number of other models in the same series as the target vehicle currently on sale. This represents the scoring benchmark bias term for the supply dimension.

[0070] Vehicle condition:

[0071] Vehicle condition data includes: ranking of its own vehicle inspection report level; mileage and vehicle age ranking; and mileage and age distribution ranking among similar models in the market. Therefore, the formula for calculating the vehicle condition dimension score is:

[0072]

[0073] in, The score indicates the vehicle's age. Z represents the sensitivity adjustment coefficient for the vehicle age dimension. standardization, Indicates the age of the target vehicle. This indicates the age of other vehicles of the same model available in the market. This indicates the baseline bias for the vehicle age dimension. Indicates mileage score, Z represents the sensitivity adjustment coefficient for the mileage dimension. standardization, Indicates the target vehicle mileage. This indicates the mileage of other vehicles of the same model available in the market. This represents the benchmark bias term for the mileage dimension.

[0074] New car perspective:

[0075] New car feature data includes: price difference trend analysis with the latest model year and lowest-spec new cars in the same series; and changes in the depreciation rate of used cars relative to new cars. Therefore, the formula for calculating the score for the new car dimension is:

[0076]

[0077] in, Z represents the sensitivity adjustment coefficient for the new car dimension. standardization, This indicates the new car price corresponding to the target vehicle. Indicates the price of the target car. This indicates the baseline bias for the new vehicle dimension.

[0078] Experience-based correction dimension:

[0079] Experience-based adjustments involve adding or deducting points for specific car models based on the sales staff's experience with the target vehicle. Therefore, the formula for calculating the score in the experience-based adjustment dimension is:

[0080]

[0081] in, This represents the score of the experience-corrected dimension. This represents the sensitivity adjustment coefficient for the experience-corrected dimension. This refers to the business experience index provided by business experts (which is a vehicle series preference correction coefficient based on historical business experience or expert experience).

[0082] In this embodiment, the empirical correction coefficient can be gradually reduced as the number of application samples accumulates, and a more objective and stable correction coefficient can be obtained by fitting more sample data, so as to improve the accuracy and stability of profitability scoring.

[0083] S2: Training a comprehensive profitability assessment model for used cars.

[0084] Specifically, the training process for the comprehensive profitability assessment model for used cars is as follows:

[0085] S201: Obtain historical transaction data for the target used car. The historical transaction data includes multi-dimensional feature data and corresponding business result label data. The multi-dimensional feature data includes price dimension data, supply and demand dimension data, vehicle condition dimension data, new car dimension data, and experience correction dimension data.

[0086] More specifically, in this embodiment, the model training uses historical real-world operating data as supervised samples, and the training samples include at least:

[0087] Multi-dimensional characteristic data of the target used car (price dimension indicators, supply and demand dimension indicators, vehicle condition dimension indicators, new car price difference indicators and experience correction indicators) and corresponding business result label data, including but not limited to: actual transaction profit or profit margin, actual sales cycle (number of days on sale).

[0088] S202: Construct training samples based on multi-dimensional feature data and corresponding business result label data;

[0089] More specifically, each historical vehicle constitutes a complete training sample, used to establish a mapping relationship between features and actual profit results.

[0090] S203: Based on the training samples, supervised learning machine learning algorithms are used to iteratively optimize the initial parameters of the weight coefficients of each dimension and the bias term of the scoring benchmark in the comprehensive profitability assessment model for used cars.

[0091] More specifically, this embodiment uses historical real business results as supervision signals and employs machine learning methods to train the parameters, including: using the output value of the used car comprehensive profitability assessment model as the model prediction value; and using actual transaction profit, sales cycle, or a combination of both as the true label value. A loss function is constructed, for example:

[0092]

[0093] in, Show the actual label value. This represents the model's predicted value.

[0094] The following parameters are iteratively optimized by minimizing the loss function: weight coefficients for each dimension. Bias items in each dimension and the combined weight of each dimension The training models that can be used include, but are not limited to: multinomial regression models, decision tree models, gradient boosting trees, or combinations thereof.

[0095] In this embodiment, the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model for used cars are set based on business experience during the cold start phase, including the weight coefficients of each dimension. Bias items in each dimension and the combined weight of each dimension .

[0096] In this embodiment, after training is complete, the optimized parameters are used as input for the next round of scoring calculations, which are then used to generate subsequent vehicle profitability scores. Model parameters can be updated periodically according to time windows (e.g., weekly, monthly) or sample size thresholds. Historical parameter versions are also retained for model performance comparison and backtesting analysis.

[0097] This embodiment utilizes machine learning methods to dynamically fit and optimize the comprehensive profitability assessment model for used cars. Specifically, it optimizes the weights of each component. In the cold start phase, empirical values ​​can be assigned. As sample data accumulates, machine learning methods (such as multinomial regression and decision tree models) can be used for dynamic fitting and optimization. The purpose of dynamic fitting and optimization is not to replace the profitability scoring formula itself, but to continuously refine the parameters in the scoring functions of each dimension (including...). , and dimensional weights This allows the generated comprehensive profitability score to more accurately reflect real business results, thereby improving the predictive ability and stability of the scoring formula in actual business scenarios.

[0098] The training method for a comprehensive profitability assessment model for used cars provided in this embodiment has the following advantages:

[0099] 1. A multi-dimensional integrated profitability evaluation system

[0100] This embodiment expands the assessment of used car profitability from a single price judgment to a comprehensive system encompassing five dimensions: price, supply and demand, vehicle condition, price difference with new cars, and experience-based adjustments. Through multi-source data collection and Z-score standardization, it achieves comparability and unified quantification of information across different dimensions. This system can dynamically reflect market changes, making the assessment results more stable and scientific, and providing a data-driven basis for used car transaction decisions.

[0101] 2. Application of Z-score-based standardized scoring algorithm in the used car industry

[0102] This embodiment employs the Z-score standardization algorithm to transform data with different dimensions and distributions into a unified standard score. This method effectively solves the problem of scale inconsistency among multi-dimensional features of used cars, ensuring reasonable weighting and strong interpretability of various indicators in the comprehensive score. By introducing weighting coefficients and bias terms, the system can automatically optimize model parameters based on data accumulation, improving the accuracy and robustness of the profitability score.

[0103] 3. Dynamic weighting model revised based on business experience

[0104] This embodiment introduces a "business experience index" as a correction term at the algorithm level to make targeted adjustments to the model's output.

[0105] During the cold start phase, initial weights can be set based on expert experience, and as data accumulates, machine learning models (such as multinomial regression or decision trees) can be used for dynamic fitting.

[0106] This "experience + data" fusion mechanism enhances the applicability and interpretability of the model in real business scenarios, ensuring that the scoring results are based on objective statistics and conform to actual industry practices.

[0107] In summary, the training method for the comprehensive profitability assessment model of used cars provided in this embodiment achieves a more comprehensive and reliable quantitative assessment of the profitability of used cars by constructing a comprehensive evaluation model covering five dimensions: price, supply and demand, vehicle condition, new car price difference, and business experience. This overcomes the shortcomings of traditional methods that rely on experience-based judgment, have limited dimensions, and lack dynamic adjustment. This method can automatically update the assessment results based on real-time market data, providing a scientific basis for used car procurement, pricing, and inventory management, effectively improving vehicle turnover efficiency and overall profitability, and has good practicality and promotional value.

[0108] Example 2

[0109] This embodiment provides a training device for a comprehensive profitability assessment model for used cars, comprising:

[0110] The model determination module is used to determine the comprehensive profitability assessment model for used cars. The comprehensive profitability assessment model for used cars is as follows:

[0111]

[0112] in, This indicates the overall profitability score for used cars. Indicates the weight of the price dimension. This indicates the score in the price dimension. Indicates the weight of supply and demand dimensions. This indicates the score based on the supply and demand dimension. Indicates the weight of the vehicle condition dimension. This indicates the score for the vehicle condition dimension. Indicates the weight of the new car dimension. This indicates the score for the new car dimension. Indicates the empirically adjusted dimension weights. This represents the score of the experience-corrected dimension. Indicates the scoring benchmark bias;

[0113] The model training module is used to train the comprehensive profitability assessment model for used cars; the training process of the comprehensive profitability assessment model for used cars is as follows:

[0114] Obtain historical transaction data of the target used car. The historical transaction data includes multi-dimensional feature data and corresponding business result tag data. The multi-dimensional feature data includes price dimension data, supply and demand dimension data, vehicle condition dimension data, new car dimension data, and experience correction dimension data.

[0115] Training samples are constructed based on the multi-dimensional feature data and the corresponding business result label data;

[0116] Based on the training samples, a supervised learning machine learning algorithm is used to iteratively optimize the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model for used cars.

[0117] For details on the specific implementation of each module in the training device for a comprehensive profitability assessment model for used cars, please refer to the above description of the limitations of the training method for a comprehensive profitability assessment model for used cars, which will not be repeated here.

[0118] The technical features of the above embodiments can be combined in any way (as long as there is no contradiction in the combination of these technical features). For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; these embodiments not explicitly written should also be considered to be within the scope of this specification.

Claims

1. A training method for a comprehensive profitability assessment model for used cars, characterized in that, include: The comprehensive profitability assessment model for used cars is determined as follows: in, This indicates the overall profitability score for used cars. Indicates the weight of the price dimension. This indicates the score in the price dimension. Indicates the weight of supply and demand dimensions. This indicates the score based on the supply and demand dimension. Indicates the weight of the vehicle condition dimension. This indicates the score for the vehicle condition dimension. Indicates the weight of the new car dimension. This indicates the score for the new car dimension. Indicates the empirically adjusted dimension weights. This represents the score of the experience-corrected dimension. Indicates the scoring benchmark bias; The training process for the comprehensive profitability assessment model for used cars is as follows: Obtain historical transaction data of the target used car. The historical transaction data includes multi-dimensional feature data and corresponding business result tag data. The multi-dimensional feature data includes price dimension data, supply and demand dimension data, vehicle condition dimension data, new car dimension data, and experience correction dimension data. Training samples are constructed based on the multi-dimensional feature data and the corresponding business result label data; Based on the training samples, a supervised learning machine learning algorithm is used to iteratively optimize the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model for used cars.

2. The training method for the comprehensive profitability assessment model for used cars according to claim 1, characterized in that, The formula for calculating the price dimension score is as follows: in, Z represents the sensitivity adjustment coefficient for the price dimension. standardization, Indicates the target price for the vehicle. This indicates the market price of vehicles. This represents the benchmark bias for the price dimension.

3. The training method for the comprehensive profitability assessment model for used cars according to claim 1, characterized in that, The formula for calculating the supply and demand dimension score is as follows: in, Indicates the demand score. Z represents the sensitivity adjustment coefficient for the demand dimension. standardization, This represents the average number of views of the target vehicle listing over a certain period of time. This represents the average number of page views of other car listings of the same model year as the target car within a certain period. This indicates the scoring benchmark bias item for the demand dimension; Indicates the supply score. Z represents the sensitivity adjustment coefficient in the supply dimension. standardization, This indicates the number of vehicles of the same model as the target vehicle currently on sale. This indicates the number of other models in the same series as the target vehicle currently on sale. This represents the scoring benchmark bias term for the supply dimension.

4. The training method for the comprehensive profitability assessment model for used cars according to claim 1, characterized in that, The formula for calculating the vehicle condition dimension score is as follows: in, The score indicates the vehicle's age. Z represents the sensitivity adjustment coefficient for the vehicle age dimension. standardization, Indicates the age of the target vehicle. This indicates the age of other vehicles of the same model available in the market. This indicates the baseline bias for the vehicle age dimension. Indicates mileage score, Z represents the sensitivity adjustment coefficient for the mileage dimension. standardization, Indicates the target vehicle mileage. This indicates the mileage of other vehicles of the same model available in the market. This represents the benchmark bias term for the mileage dimension.

5. The training method for the comprehensive profitability assessment model for used cars according to claim 1, characterized in that, The formula for calculating the new vehicle dimension score is as follows: in, Z represents the sensitivity adjustment coefficient for the new car dimension. standardization, This indicates the new car price corresponding to the target vehicle. Indicates the price of the target car. This indicates the baseline bias for the new vehicle dimension.

6. The training method for the comprehensive profitability assessment model for used cars according to claim 1, characterized in that, The formula for calculating the experience-corrected dimension score is as follows: in, This represents the sensitivity adjustment coefficient for the experience-corrected dimension. This represents the business experience index provided by business experts.

7. The training method for the comprehensive profitability assessment model for used cars according to claim 1, characterized in that, When iteratively optimizing the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the used car comprehensive profitability assessment model using a supervised learning machine learning algorithm based on the training samples, the specific steps include: The output value of the comprehensive profitability assessment model for used cars is used as the predicted value, and the business result label data is used as the actual label value. A loss function is constructed based on the predicted value and the true label value; The machine learning algorithm iteratively optimizes the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model of used cars based on the loss function.

8. The training method for the comprehensive profitability assessment model for used cars according to claim 1, characterized in that, The machine learning algorithm employs one or more of the following: multinomial regression model, decision tree model, or gradient boosting tree.

9. The training method for the comprehensive profitability assessment model for used cars according to claim 1, characterized in that, The initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model for used cars are set based on business experience.

10. A training device for a comprehensive profitability assessment model for used cars, characterized in that, include: The model determination module is used to determine the comprehensive profitability assessment model for used cars. The comprehensive profitability assessment model for used cars is as follows: in, This indicates the overall profitability score for used cars. Indicates the weight of the price dimension. This indicates the score in the price dimension. Indicates the weight of supply and demand dimensions. This indicates the score based on the supply and demand dimension. Indicates the weight of the vehicle condition dimension. This indicates the score for the vehicle condition dimension. Indicates the weight of the new car dimension. This indicates the score for the new car dimension. Indicates the empirically adjusted dimension weights. This represents the score of the experience-corrected dimension. Indicates the scoring benchmark bias; The model training module is used to train the comprehensive profitability assessment model for used cars; the training process of the comprehensive profitability assessment model for used cars is as follows: Obtain historical transaction data of the target used car. The historical transaction data includes multi-dimensional feature data and corresponding business result tag data. The multi-dimensional feature data includes price dimension data, supply and demand dimension data, vehicle condition dimension data, new car dimension data, and experience correction dimension data. Training samples are constructed based on the multi-dimensional feature data and the corresponding business result label data; Based on the training samples, a supervised learning machine learning algorithm is used to iteratively optimize the initial parameters of the weight coefficients of each dimension and the scoring benchmark bias term in the comprehensive profitability assessment model for used cars.

Images