New energy second-hand vehicle value evaluation method and equipment

Abstract

The invention relates to the technical field of value assessment, in particular to a new energy second-hand vehicle value assessment method and equipment, which can obtain vehicle basic data, use historical data, website data and power battery data of a to-be-assessed vehicle to form a multi-source data set; performing feature extraction on the data in the multi-source data set to obtain conventional factor features, behavior factor features, heat factor features and battery factor features; constructing a new energy second-hand vehicle value evaluation model according to an integrated learning method; and inputting the features into a new energy second-hand vehicle value evaluation model to obtain a value prediction result of the to-be-evaluated vehicle. According to the technical scheme, multi-dimensional information, especially the influence of the battery health state on the vehicle residual value, can be obtained, and more comprehensive and accurate second-hand vehicle value evaluation can be provided; through the ensemble learning method, the accuracy of value evaluation can be improved, and the stability of the model is enhanced through ensemble learning, so that the prediction result is more reliable.

Description

Technical Field

[0001] The present invention relates to the technical field of value assessment, and particularly relates to a method and device for evaluating the value of used new energy vehicles. Background Art

[0002] With the continuous progress of technology and the gradual reduction of costs, the market retention of new energy vehicles has been increasing year by year, and the trading demand for used new energy vehicles has also been growing vigorously. However, the existing used vehicle value assessment system has exposed many defects when dealing with used new energy vehicles.

[0003] Most traditional used vehicle valuation methods are designed based on the characteristics of fuel vehicles and do not fully consider the unique attributes of new energy vehicles. Among them, the battery, as the core component of new energy vehicles, factors such as its health status, charging cycle times, and remaining capacity have a great impact on the vehicle's cruising range, performance, and actual value. However, in traditional valuation models, these key factors are often ignored.

[0004] The currently common valuation methods mainly include: manual experience valuation, rule-driven models, and simple regression models based on historical data.

[0005] Manual experience valuation mainly relies on the personal experience of appraisers and subjective judgments of the market. Differences in the professional levels and experiences of different appraisers will lead to large fluctuations in the evaluation results, with strong subjectivity. Moreover, manual evaluation has low efficiency, is difficult to meet the market demand for large-scale and rapid transactions, and is prone to human errors during the evaluation process.

[0006] Rule-driven models set valuation rules based on fixed factors such as vehicle brand, model, and service life, and cannot flexibly adapt to the dynamic changes of the market. New energy vehicle technology iterates rapidly, and the market supply and demand relationship and consumer preferences change frequently. Rule-driven models are difficult to adjust in real time, resulting in the disconnection between the valuation results and the actual market value.

[0007] The simple regression model based on historical data only performs weighted calculations on some vehicle attributes through simple algorithms, ignoring the complex interaction between multi-source information such as vehicle condition data, market condition data, and historical transaction data. For example, it fails to comprehensively consider the internal relationship between battery status, driving mileage, and service life, making the valuation results unable to accurately reflect the true value of used new energy vehicles.

[0008] Therefore, the existing valuation methods generally have the problem of insufficient data utilization, cannot effectively integrate multi-source information, and are insensitive to market price fluctuations, making it difficult to provide timely and accurate valuation services and unable to meet the needs of the rapid development and flexible changes of the used new energy vehicle market. Summary of the Invention

[0009] In view of this, the purpose of the present invention is to provide a method and device for evaluating the value of new energy used vehicles, so as to solve the problem that the existing valuation methods cannot meet the needs of the rapid development and flexible changes of the new energy used vehicle market.

[0010] According to the first aspect of the embodiments of the present invention, a method for evaluating the value of new energy used vehicles is provided, including:

[0011] Obtain the vehicle basic data, usage history data, website data and power battery data of the vehicle to be evaluated, and form a multi-source data set;

[0012] Extract features from the data in the multi-source data set to obtain conventional factor features, behavior factor features, popularity factor features and battery factor features;

[0013] Construct a value evaluation model for new energy used vehicles according to the integrated learning method;

[0014] Input the conventional factor features, behavior factor features, popularity factor features and battery factor features into the value evaluation model for new energy used vehicles to obtain the value prediction result of the vehicle to be evaluated.

[0015] Preferably, the popularity factor features include: vehicle model scoring features, vehicle model comment features and popularity features.

[0016] Preferably, extracting the vehicle model comment features from the website data includes:

[0017] Obtain the comment data of users on the vehicle model to be evaluated from the website data, and clean the comment data;

[0018] Statistically obtain the total number of comments on the vehicle model from the cleaned comment data;

[0019] Use the pre-trained sentiment analysis model to judge the sentiment direction of each comment in the comment data, and statistically obtain the number of positive comments and the number of negative comments;

[0020] Perform extremely large normalization processing and extremely small normalization processing on the total number of comments, the number of positive comments and the number of negative comments respectively;

[0021] Use the entropy weight method to calculate the weight coefficients of the total number of comments, the number of positive comments and the number of negative comments respectively;

[0022] According to the weight coefficients, perform weighted normalization processing on the total number of comments, the number of positive comments and the number of negative comments;

[0023] According to the data after weighted normalization processing, obtain the positive ideal solution and the negative ideal solution of the total number of comments, the number of positive comments and the number of negative comments respectively;

[0024] Calculate the Euclidean distance between each vehicle model and the positive and negative ideal solutions, calculate the relative closeness of each vehicle model according to the Euclidean distance, obtain the comprehensive score according to the relative closeness, and use the comprehensive score as the vehicle model scoring feature.

[0025] Preferably, construct a new energy used vehicle value evaluation model according to the ensemble learning method, including:

[0026] Obtain used vehicle transaction data, and obtain the key features affecting the price from the used vehicle transaction data;

[0027] Input the key features into the ensemble learning model for training to obtain a new energy used vehicle value evaluation model.

[0028] Preferably, the method further includes:

[0029] The ensemble learning model includes multiple basic models, and weights are assigned to each basic model according to the training situation;

[0030] When obtaining the value prediction result of the vehicle to be evaluated:

[0031] Use all the basic models to make predictions respectively to obtain the preliminary prediction results corresponding to each basic model;

[0032] Calculate the weighted average of the preliminary prediction results corresponding to all the basic models as the final prediction result.

[0033] Preferably, the method further includes:

[0034] The ensemble learning model includes a first-layer model group and a second-layer model. The first-layer model group includes multiple basic models of different types, and the second-layer model is a single model;

[0035] When obtaining the value prediction result of the vehicle to be evaluated:

[0036] Use the basic models in the first layer to make predictions respectively to obtain the preliminary prediction results corresponding to each basic model;

[0037] Use the second-layer model to make a re-prediction according to all the preliminary prediction results to obtain the final prediction result.

[0038] Preferably, perform feature extraction on the data in the multi-source dataset, including:

[0039] Extract the conventional factor features from the vehicle basic data; the conventional factor features include: basic features, vehicle condition features, vehicle model level features, and market capacity features;

[0040] Extract the behavior factor features from the usage history data;

[0041] Extract the popularity factor features from website data;

[0042] Extract the battery factor features from the power battery data; the battery factor features include: charging features and battery rating features.

[0043] According to the second aspect of the embodiments of the present invention, there is provided a new energy used car value evaluation device, including:

[0044] A main controller, and a memory connected to the main controller;

[0045] The memory stores program instructions therein;

[0046] The main controller is configured to execute the program instructions stored in the memory to perform the method described in any one of the above.

[0047] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0048] It can be understood that the technical solution shown in the present invention can obtain the vehicle basic data, usage history data, website data and power battery data of the vehicle to be evaluated, and constitute a multi-source data set; extract features from the data in the multi-source data set to obtain conventional factor features, behavior factor features, popularity factor features and battery factor features; construct a new energy used car value evaluation model according to the ensemble learning method; input the features into the new energy used car value evaluation model to obtain the value prediction result of the vehicle to be evaluated. The technical solution shown in the present invention can obtain multi-dimensional information, especially the impact of the battery health state on the vehicle residual value, and can provide a more comprehensive and accurate used car value evaluation; through the ensemble learning method, the accuracy of the value evaluation can be improved, and the ensemble learning enhances the stability of the model, making the prediction result more reliable.

[0049] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention.

[0051] Figure 1 It is a schematic diagram of the steps of a new energy used car value evaluation method shown according to an exemplary embodiment;

[0052] Figure 2 It is a schematic diagram of a data processing flow shown according to an exemplary embodiment;

[0053] Figure 3 It is a flowchart of the processing of vehicle model comment features shown according to an exemplary embodiment. Specific Embodiments

[0054] Here, exemplary embodiments will be described in detail, and examples thereof are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0055] In one embodiment, Figure 1 is a schematic diagram of the steps of a new energy used car value evaluation method shown according to an exemplary embodiment. Refer to Figure 1 to provide a new energy used car value evaluation method, including:

[0056] Step S11: Obtain the vehicle basic data, usage history data, website data, and power battery data of the vehicle to be evaluated, and form a multi-source data set;

[0057] Step S12: Extract features from the data in the multi-source data set to obtain conventional factor features, behavior factor features, popularity factor features, and battery factor features;

[0058] Step S13: Construct a new energy used car value evaluation model according to the ensemble learning method;

[0059] Step S14: Input the conventional factor features, behavior factor features, popularity factor features, and battery factor features into the new energy used car value evaluation model to obtain the value prediction result of the vehicle to be evaluated.

[0060] It can be understood that the technical solution shown in this embodiment can obtain the vehicle basic data, usage history data, website data, and power battery data of the vehicle to be evaluated, and form a multi-source data set; extract features from the data in the multi-source data set to obtain conventional factor features, behavior factor features, popularity factor features, and battery factor features; construct a new energy used car value evaluation model according to the ensemble learning method; input the features into the new energy used car value evaluation model to obtain the value prediction result of the vehicle to be evaluated. The technical solution shown in this embodiment can obtain multi-dimensional information, especially the impact of the battery health status on the vehicle residual value, and can provide a more comprehensive and accurate used car value evaluation; through the ensemble learning method, the accuracy of the value evaluation can be improved, and the ensemble learning enhances the stability of the model, making the prediction result more reliable.

[0061] In specific practice, refer to Figure 2First, in step S11, the data sources mainly include vehicle basic data, usage history data, website data, and power battery data. Vehicle basic data: includes the basic information of the vehicle, vehicle condition data, and market data. Usage history data: includes the usage data of the vehicle. Website data: includes vehicle model ratings, vehicle model reviews, and search indices. Power battery data: includes charging records and battery status.

[0062] In step S12, feature extraction is performed on the data in the multi-source dataset to obtain conventional factor features, behavior factor features, popularity factor features, and battery factor features.

[0063] It should be noted that this step includes:

[0064] Extract the conventional factor features from the vehicle basic data; the conventional factor features include: basic features, vehicle condition features, vehicle model level features, and market capacity features.

[0065] Extract the behavior factor features from the usage history data; the behavior factor features are the behavior characteristics of the vehicle owner when using the vehicle.

[0066] Extract the popularity factor features from the website data; the popularity factor features include: vehicle model rating features, vehicle model review features, and popularity features.

[0067] Extract the battery factor features from the power battery data; the battery factor features include: charging features and battery rating features.

[0068] In a preferred embodiment, due to the subjectivity of the vehicle model review features, the following method can be used to obtain the vehicle model review features. See Figure 3 It should be noted that extracting the vehicle model review features from the website data includes:

[0069] Step S21: Obtain the review data of the user on the vehicle model to be evaluated from the website data, and perform data cleaning on the review data.

[0070] First, obtain the review data. The review data of users on a specific vehicle model can be collected from platforms such as automotive forums, social media, and review websites. After obtaining the data, perform data cleaning operations. The main purpose is to remove duplicate reviews, irrelevant reviews, and noise data to ensure the validity of the review content.

[0071] Step S22: Statistically obtain the total number of reviews of the vehicle model from the cleaned review data.

[0072] Statistically count the total number of all review entries for each vehicle model as an indicator to measure the degree of attention of the vehicle model. When dealing with a specific vehicle model, select all the reviews of the vehicle model.

[0073] Step S23: Use the pre-trained sentiment analysis model to determine the sentiment direction of each comment in the comment data, and count the number of positive comments and negative comments.

[0074] Perform sentiment analysis on each comment. To accurately capture the true feelings of users towards a specific vehicle model, in this embodiment, a pre-trained sentiment analysis model is used to perform positive and negative sentiment judgments on each comment.

[0075] The model selected is BERT-Sentiment, which is a specific application or fine-tuned version of deep learning models based on the Transformer architecture (such as BERT) in sentiment analysis tasks. It is pre-trained on a large-scale corpus and further optimized on a specific dataset, capable of capturing complex semantic and context information, and is suitable for scenarios that require high-precision sentiment classification.

[0076] The sentiment analysis model used needs to have multilingual support. Considering that comments may come from different countries and regions, the sentiment analysis model has the ability to process multiple languages to ensure that it can accurately parse the content of comments in various languages.

[0077] The sentiment analysis model has a custom adjustment function. For the professional terms and industry characteristics in the automotive field, the pre-trained model is fine-tuned to improve its accuracy in a specific field. For example, special attention is paid to keywords related to vehicle performance, driving experience, battery life, etc.

[0078] The sentiment analysis model has the function of context understanding. The model not only identifies the sentiment tendency of individual words, but also can understand the overall sentiment color of sentences and even paragraphs. By considering context information, misjudgments caused by isolated words and sentences are avoided.

[0079] The sentiment analysis model has the function of sentiment intensity evaluation. In addition to simple positive / negative binary classification, a sentiment intensity evaluation mechanism is introduced, that is, different degrees of weights are assigned according to the degree of sentiment expression in the comment. For example, although "very good" and "good" are both positive evaluations, the former has a higher sentiment intensity.

[0080] Uncertainty processing of the sentiment analysis model: For ambiguous or equivocal comments, the model will mark them as neutral or uncertain categories and make a comprehensive judgment in combination with other features to reduce errors.

[0081] Step S24: Perform extremely large normalization processing and extremely small normalization processing on the total number of comments, the number of positive comments, and the number of negative comments respectively.

[0082] After counting the number of positive comments and negative comments, calculate the weights by the entropy weight method.

[0083] First, perform data standardization: For the total number of comments, the number of positive comments, and the number of negative comments, perform maximization and minimization standardization respectively.

[0084] Maximization:

[0085] Minimization:

[0086] Among them, X is the value after standardization, x is the value in the original data, min is the minimum value in this group of data, and max is the maximum value in this group of data. Through the maximization formula, the original data is mapped to the interval [0,1]. The larger the original data, the closer the standardized value is to 1, which can highlight the differences in larger values in the data. Through the minimization formula, the original data is mapped to the interval [0,1]. The smaller the original data, the closer the standardized value is to 1, which can highlight the differences in larger values in the data.

[0087] Step S25: Use the entropy weight method to calculate the weight coefficients of the total number of comments, the number of positive comments, and the number of negative comments respectively.

[0088] By calculating the probability matrix, and then calculating the entropy weight.

[0089] Preferably, after calculating the weights, the weights can also be adjusted based on data volatility to ensure that the weights match the data volatility.

[0090] Evaluate data volatility: Calculate the data volatility metrics (such as standard deviation, variance) of each feature. Data volatility measures the degree of dispersion of the data. The commonly used metrics are standard deviation and variance. Taking the standard deviation as an example, assume there is a set of data on the total number of comments of a certain vehicle model, and then calculate the standard deviation. The variance is the square of the standard deviation. By calculating these metrics, we can clearly understand the fluctuation of each feature (such as the total number of comments, the number of positive comments, the number of negative comments). The greater the data fluctuation, the greater the difference in the values of this feature, and its importance in the evaluation may also be different.

[0091] Compare the entropy weight results with volatility: Identify the mismatched situations. After calculating the entropy weights previously, compare the entropy weight results with the data volatility obtained through the above calculations. If the data volatility of a certain feature is large, but the weight calculated by the entropy weight is low, or vice versa, it means that there is a mismatched situation. For example, if the data of the number of positive comments of a certain vehicle model fluctuates significantly, but the entropy weight shows that its weight is low, this may mean that the importance of this feature was not fully reflected in the previous calculations.

[0092] Application adjustment rule: Perform unconventional standardization according to the data characteristics of the fields to ensure that the weight after processing matches the data fluctuation. When a mismatch is found, perform unconventional standardization according to the data characteristics of each field (i.e., each feature). This standardization is not a simple linear transformation, but a targeted processing based on the data distribution, feature properties, etc. For example, for some data with special distributions, specific function transformations may be required to readjust their values. Through such processing, the finally determined weight size can match the data fluctuation, making the weight more reasonably reflect the importance of the feature in evaluating the attention degree of the vehicle model, user sentiment tendency, etc., and then more accurately calculate the comprehensive score of the vehicle comment feature, providing a more reliable basis for the value evaluation of new energy used cars.

[0093] After that, use the TOPSIS method to calculate the comprehensive score. The TOPSIS method (Technique for Order Preference by Similarity to an Ideal Solution), that is, the Technique for Order Preference by Similarity to an Ideal Solution, is a commonly used multi-index comprehensive evaluation method.

[0094] Step S26: Perform weighted normalization processing on the total number of comments, the number of positive comments, and the number of negative comments according to the weight coefficient.

[0095] Based on the weights calculated and adjusted by the entropy weight method before, perform weighted normalization on the total number of comments, the number of positive comments, and the number of negative comments after the standardization process. Through this step, the weights of each feature are considered, enabling different features to participate in the operation reasonably according to their importance in subsequent calculations.

[0096] Step S27: According to the data after the weighted normalization process, obtain the positive ideal solution and the negative ideal solution of the total number of comments, the number of positive comments, and the number of negative comments respectively.

[0097] The positive ideal solution is a vector composed of the optimal values in each feature, and the negative ideal solution is a vector composed of the worst values in each feature. For extremely large indicators such as the total number of comments and the number of positive comments (the larger the value, the better), the positive ideal solution is the maximum value of this feature among all vehicle models, and the negative ideal solution is the minimum value; for extremely small indicators such as the number of negative comments (the smaller the value, the better), the positive ideal solution is the minimum value of this feature among all vehicle models, and the negative ideal solution is the maximum value. For example, assuming there are data on the number of positive comments of 10 vehicle models, then the maximum value in this group of data is the positive ideal solution of the feature of the number of positive comments, and the minimum value is the negative ideal solution. After determining the positive and negative ideal solutions, it provides a reference standard for calculating the gap between each vehicle model and the ideal state in the subsequent calculation.

[0098] Step S28: Calculate the Euclidean distances of each vehicle model from the positive and negative ideal solutions, calculate the relative closeness degree of each vehicle model based on the Euclidean distances, obtain the comprehensive score according to the relative closeness degree, and use the comprehensive score as the vehicle model scoring feature.

[0099] Calculate the Euclidean distances of each vehicle model from the positive and negative ideal solutions to measure the gap between each vehicle model and the ideal state. For the i-th vehicle model, its Euclidean distance from the positive ideal solution The calculation formula is:

[0100]

[0101] x ij is the j-th weighted normalized eigenvalue of the i-th vehicle model, is the positive ideal solution of the j-th feature.

[0102] For the i-th vehicle model, its Euclidean distance from the negative ideal solution The calculation formula is:

[0103]

[0104] is the negative ideal solution of the j-th feature.

[0105] By calculating the Euclidean distances, it can be intuitively seen how close each vehicle model is to the positive and negative ideal solutions. The smaller the distance, the closer the vehicle model is to the ideal state in these features.

[0106] Calculate the relative closeness degree: Based on the Euclidean distances calculated above, calculate the relative closeness degree S i :

[0107]

[0108] The value range of the relative closeness degree is between [0, 1]. S i The closer it is to 1, the closer the vehicle model is to the positive ideal solution and the farther it is from the negative ideal solution, that is, the better the vehicle model performs in the comprehensive evaluation; S i The closer it is to 0, the closer the vehicle model is to the negative ideal solution and the worse the comprehensive performance. The finally obtained relative closeness degree is the comprehensive score of each vehicle model, which is used to evaluate the attention degree of the vehicle model, the user's emotional tendency, etc., and provides an important basis for the value evaluation of new energy used cars.

[0109] In another embodiment, regarding the construction and training of the new energy used car value evaluation model, in order to improve the accuracy and stability of pricing prediction, the present invention adopts the Ensemble Learning method, and inputs the comprehensive scoring result into the ensemble learning model for the final price prediction. Ensemble Learning can significantly improve the overall performance of the model, reduce the risk of overfitting, and enhance the robustness to data noise by combining the prediction results of multiple basic models (weak learners).

[0110] Use XGBoost as the core ensemble learning model for used car pricing prediction. XGBoost is an efficient implementation based on the gradient boosting framework and has the following remarkable advantages:

[0111] Regularization mechanism: Built-in L1 and L2 regularization terms, effectively preventing overfitting and ensuring the generalization ability of the model on new data. Efficient computational performance: Through parallel computing and memory optimization, XGBoost can quickly process large-scale data sets, meeting the requirements of real-time and efficiency. High-precision prediction: XGBoost emphasizes capturing subtle features, especially performing well in the case of non-linear relationships and complex feature combinations, and is very suitable for tasks with diverse features such as used car pricing. Flexibility: Supports custom loss functions and evaluation metrics, allowing adjustment of the model optimization direction according to actual business needs.

[0112] It should be noted that constructing a new energy used car value evaluation model according to the ensemble learning method includes: obtaining used car transaction data, and deriving key features affecting the price from the used car transaction data; inputting the key features into the ensemble learning model for training to obtain a new energy used car value evaluation model.

[0113] When performing data processing, feature selection is carried out first. By analyzing a large amount of used car transaction data, key features with the greatest impact on the price are identified, such as vehicle model, year, mileage, battery health status, etc. These features are used as input variables into the ensemble learning model to ensure that the model can focus on the most important information.

[0114] At the same time, some features need to be transformed. Some original features may need to be appropriately transformed or encoded, such as converting categorical variables into numerical representations (one-hot encoding), or normalizing continuous variables to meet the requirements of different models and improve model performance.

[0115] During model training and optimization, to evaluate the generalization ability of the model, the K-fold cross-validation technique is adopted during the training process. This helps ensure that the model not only performs well on the training set but also has reliable predictive performance on unseen data. At the same time, methods such as Grid Search, Random Search, or Bayesian optimization are used to systematically explore different combinations of hyperparameters, find the optimal configuration, and further improve the performance of the model.

[0116] It should be noted that the ensemble learning model includes multiple base models, and weights are assigned to each base model according to the training situation. When the value prediction result of the vehicle to be estimated is obtained by the model constructed in this way:

[0117] All the base models are used to make predictions separately to obtain the preliminary prediction results corresponding to each base model; the weighted average of the preliminary prediction results corresponding to all the base models is calculated as the final prediction result.

[0118] The weighted average method is adopted. Different weights are assigned to each base model according to their performance, and then the weighted average is calculated as the final prediction result. This method can make full use of the advantages of each model and reduce the bias that may be brought by a single model.

[0119] It should be noted that the ensemble learning model includes a first-layer model group and a second-layer model. The first-layer model group includes multiple base models of different types, and the second-layer model is a single model. When the value prediction result of the vehicle to be estimated is obtained by the model constructed in this way:

[0120] The base models in the first layer are used to make predictions separately to obtain the preliminary prediction results corresponding to each base model; the second-layer model is used to make a re-prediction based on all the preliminary prediction results to obtain the final prediction result.

[0121] The stacking method is used to construct a multi-layer ensemble model, that is, multiple different types of base models are used in the first layer to generate preliminary predictions, and then another model (such as linear regression or logistic regression) is used in the second layer to re-train these preliminary predictions to obtain the final output. This way often achieves better results.

[0122] When the new energy used car value evaluation method shown in the present invention is implemented, through the integrated learning method, the accuracy of used car pricing prediction has been significantly improved. Especially when dealing with complex and diverse data, the model can more accurately reflect the market price dynamics. The integrated learning enhances the stability of the model, reduces the influence of outliers and noise, and makes the prediction results more reliable. The data source of the present invention is extensive (adding website data, market data, battery data, etc.) and has stronger multi-source data integration ability; it can achieve complex feature extraction and integrated feature calculation; by introducing the integrated learning technology, the accuracy and applicability of the valuation are improved.

[0123] In another embodiment, there is provided a new energy used car value evaluation device, including:

[0124] A main controller, and a memory connected to the main controller;

[0125] The memory, in which program instructions are stored;

[0126] The main controller is used to execute the program instructions stored in the memory to execute the method described in any one of the above.

[0127] It can be understood that the same or similar parts in the above embodiments can be referred to each other, and the content not detailed in some embodiments can be seen in the same or similar content of other embodiments.

[0128] It should be noted that in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0129] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of the code including one or more executable instructions for implementing a specific logical function or process. The scope of the preferred embodiments of the present invention includes additional implementations, where the functions can be executed in a manner that is not shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0130] It should be understood that each part of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0131] Those of ordinary skill in the art can understand that all or part of the steps carried by the method of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0132] In addition, in each embodiment of the present invention, each functional unit can be integrated into a processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. When the above integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0133] The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disc, etc.

[0134] In the description of this specification, the descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0135] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for evaluating the value of used new energy vehicles, characterized in that, Including: Obtain the vehicle basic data, usage history data, website data, and power battery data of the vehicle to be evaluated, and form a multi-source dataset; Extract features from the data in the multi-source dataset to obtain conventional factor features, behavioral factor features, popularity factor features, and battery factor features; Construct a new energy used car value evaluation model according to the ensemble learning method; Input the conventional factor features, behavioral factor features, popularity factor features, and battery factor features into the new energy used car value evaluation model to obtain the value prediction result of the vehicle to be evaluated.

2. The method according to claim 1, characterized in that, The popularity factor features include: model score features, model review features, and popularity features.

3. The method according to claim 2, wherein Extract model review features from website data, including: Obtain the review data of the vehicle model to be evaluated by users from website data, and clean the review data; Statistically obtain the total number of reviews of the model from the cleaned review data; Use a pre-trained sentiment analysis model to judge the sentiment direction of each review in the review data, and count the number of positive reviews and negative reviews; Perform extremely large normalization processing and extremely small normalization processing on the total number of reviews, the number of positive reviews, and the number of negative reviews respectively; Use the entropy weight method to calculate the weight coefficients of the total number of reviews, the number of positive reviews, and the number of negative reviews respectively; Perform weighted normalization processing on the total number of reviews, the number of positive reviews, and the number of negative reviews according to the weight coefficients; According to the data after weighted normalization processing, obtain the positive ideal solution and negative ideal solution of the total number of reviews, the number of positive reviews, and the number of negative reviews respectively; Calculate the Euclidean distance between each model and the positive and negative ideal solutions, calculate the relative closeness of each model according to the Euclidean distance, obtain the comprehensive score according to the relative closeness, and use the comprehensive score as the model score feature.

4. The method according to claim 1, characterized in that, Construct a new energy used car value evaluation model according to the ensemble learning method, including: Obtain used car transaction data, and obtain the key features affecting the price from the used car transaction data; Input the key features into the ensemble learning model for training to obtain a new energy used car value evaluation model.

5. The method according to claim 4, wherein Also including: There are multiple basic models in the ensemble learning model, and weights are assigned to each basic model according to the training situation; When obtaining the value prediction result of the vehicle to be evaluated: Use all the basic models to make predictions respectively to obtain the preliminary prediction results corresponding to each basic model; Calculate the weighted average of the preliminary prediction results corresponding to all the basic models as the final prediction result.

6. The method according to claim 4, wherein Also including: The ensemble learning model includes a first-layer model group and a second-layer model. The first-layer model group includes multiple basic models of different types, and the second-layer model is a single model; When obtaining the value prediction result of the vehicle to be evaluated: Use the basic models of the first layer to make predictions respectively to obtain the preliminary prediction results corresponding to each basic model; Use the second-layer model to make a re-prediction according to all the preliminary prediction results to obtain the final prediction result.

7. The method according to claim 1, characterized in that Extract features from the data in the multi-source dataset, including: Extract the conventional factor features from the vehicle basic data; the conventional factor features include: basic features, vehicle condition features, vehicle type level features, and market capacity features; Extract the behavior factor features from the usage history data; Extract the popularity factor features from the website data; Extract the battery factor features from the power battery data; the battery factor features include: charging features and battery rating features.

8. A new energy used car value evaluation device, characterized in that, Include: A main controller and a memory connected to the main controller; The memory stores program instructions; The main controller is configured to execute the program instructions stored in the memory and execute the method according to any one of claims 1 to 7.

Images