A second-hand car transaction information management system and method

Abstract

The present invention provides a second-hand car transaction information management system and method, which relates to the field of vehicle transaction technology. The method includes: receiving information of a target second-hand car input by a user; dividing the information of the target second-hand car into low-sensitivity information and high-sensitivity information according to the degree of influence of the information of the target second-hand car on the transaction price; quantifying the high-sensitivity information and the low-sensitivity information to generate high-sensitivity data and low-sensitivity data; inputting the low-sensitivity data and the high-sensitivity data into a pre-trained prediction model to obtain an intermediate transaction price; constructing a transaction price interval according to the distribution of the training sample error value of the prediction model; and feeding back the transaction price interval to the user. The present invention makes predictions based on the information of the target second-hand car, and adopts an interval estimation method for the prediction value, so as to accurately predict the transaction residual value of the second-hand car.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle transaction, and in particular to a second-hand car transaction information management system and method. Background Art

[0002] The used car transaction information management system is a digital platform for managing the used car buying and selling process, including vehicle information management, used car residual value estimation and transaction management. In this system, the residual value estimation of used cars is the most important part, that is, providing the current price of used cars, or retrieving matching vehicles based on the price provided by the user.

[0003] Currently, traditional residual value assessment methods, such as the replacement cost method and the current market price method, are still used by many used car appraisers in offline assessments due to their low cost and simple operation. These methods and models rely heavily on the personal experience of relevant analysts, and can neither guarantee the accuracy of the assessment nor be promoted on a large scale.

[0004] In view of this, the present invention is proposed. Summary of the invention

[0005] In view of the above problems, the present invention proposes a used car transaction information management system and method to solve the problems existing in the prior art that the residual value assessment of used cars relies on subjective judgment and has high time and labor costs, and the online assessment standards are not unified and the results have large errors.

[0006] In a first aspect, the present invention provides a used car transaction information management system, comprising:

[0007] An input module, used to receive information of a target used car input by a user; and to classify the information of the target used car into low-sensitivity information and high-sensitivity information according to the degree of influence of the information of the target used car on the transaction price;

[0008] A pre-processing module, used for quantifying the high-sensitivity information and the low-sensitivity information to generate high-sensitivity data and low-sensitivity data;

[0009] A prediction module, used for inputting the low-sensitivity data and the high-sensitivity data into a pre-trained prediction model to obtain an intermediate transaction price; wherein the prediction model includes a lightweight gradient boosting machine, a fully connected layer and an output layer; the lightweight gradient boosting machine is used for inputting the low-sensitivity data and outputting a first feature vector, the fully connected layer is used for performing weighted summation on the high-sensitivity data to obtain a second feature vector; the output layer is used for fusing the first feature vector and the second feature vector to obtain an intermediate transaction price;

[0010] An interval estimation module, used to construct an interval of transaction prices according to the distribution of error values ​​of training samples of the prediction model;

[0011] The output module is used to feed back the transaction price range to the user.

[0012] In a second aspect, the present invention provides a second-hand car transaction information management method, comprising:

[0013] receiving information of a target used car input by a user; and dividing the information of the target used car into low-sensitivity information and high-sensitivity information according to the degree of influence of the information of the target used car on the transaction price;

[0014] quantifying the high-sensitivity information and the low-sensitivity information to generate high-sensitivity data and low-sensitivity data;

[0015] Inputting the low-sensitivity data and the high-sensitivity data into a pre-trained prediction model to obtain an intermediate transaction price; wherein the prediction model includes a lightweight gradient boosting machine, a fully connected layer and an output layer; the lightweight gradient boosting machine is used to input the low-sensitivity data and output a first feature vector, the fully connected layer is used to perform weighted summation on the high-sensitivity data to obtain a second feature vector; the output layer is used to fuse the first feature vector and the second feature vector to obtain an intermediate transaction price;

[0016] Constructing a transaction price interval according to the distribution of the error values ​​of the training samples of the prediction model;

[0017] The transaction price range is fed back to the user.

[0018] Compared with the prior art, the present invention has the following technical effects:

[0019] 1. Use the prediction model to make predictions based on the information of the target used car, and use the interval estimation method for the predicted value to accurately predict the transaction residual value of the used car.

[0020] 2. This embodiment provides a new prediction model, including a lightweight gradient boosting machine, a fully connected layer and an output layer. The prediction model can learn the first eigenvector that affects the transaction price from the low-sensitivity data, directly perform weighted summation using the high-sensitivity data, and fuse it with the first eigenvector to obtain the intermediate transaction price, so as to highlight the impact of the high-sensitivity data on the transaction price.

[0021] 3. The present invention adopts a lightweight gradient boosting machine for feature extraction, which is beneficial to improving computational efficiency and solving the overfitting problem in the training process.

[0022] The description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above description and other purposes, features and advantages of the present invention more obvious and easy to understand, the preferred embodiments are specifically cited and described in detail as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0024] Figure 1 A schematic diagram of the structure of a used car transaction information management system provided by an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of the structure of a prediction model provided by an embodiment of the present invention;

[0026] Figure 3 A schematic diagram of a sliding window provided by an embodiment of the present invention;

[0027] Figure 4 A flowchart of a method for managing used car transaction information provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0029] In the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be connected, detachably connected, or integrated; it can be mechanically connected or electrically connected; it can be directly connected or indirectly connected through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0030] Example 1

[0031] The present invention proposes a second-hand car transaction information management system, see Figure 1,The system includes : an input module, a pre-processing module, a prediction module, an interval estimation module and an output module.

[0032] The input module is used to receive the information of the target used car input by the user.

[0033] The system provided in this embodiment has a user-oriented interface, and an input box for used car information is displayed on the interface. The user can enter the information of the used car in the input box, which is called the target used car information for the convenience of description and distinction, such as the information of the used car that the user wants to buy, or the information of the used car that the user wants to sell.

[0034] The target used car information is an indicator that can affect the transaction price. In order to guide the user to input and avoid entering unnecessary information, the target used car information name is displayed on the interface to guide the user to input. Optionally, the target used car information includes: vehicle attributes, performance characteristics, usage characteristics, historical transaction volume, vehicle model ownership, accident information and repair and maintenance information. The specific content of each information is introduced in detail below.

[0035] Vehicle attributes include model, brand and manufacturer's suggested retail price of new cars; performance characteristics include transmission type, displacement, drive type, horsepower, torque, maximum speed and fuel consumption per 100 kilometers; usage characteristics include vehicle age, usage nature, number of license plate changes and mileage; accident information includes the type (such as collision, scratches) and number of accidents involving the used car; repair and maintenance information includes the type (such as film wrapping, oil change) and number of repairs and maintenance performed on the used car.

[0036] Collect the target used car information entered by the user on the interface. It can be seen that some of this information is numerical and some is text.

[0037] For different models, the main information that affects the residual value of the in-use vehicle may be different. For example, for luxury brand vehicles, vehicle attributes are the main information that affects the residual value of the in-use vehicle, and for ordinary vehicles, performance characteristics and usage characteristics are the main information that affects the residual value of the in-use vehicle. Accident information affects the residual value of almost all models of in-use vehicles. Based on this, this embodiment first divides the information of the target used car into low-sensitivity information and high-sensitivity information according to the degree of influence of the information of the target used car on the transaction price (including personal transaction price, car dealer purchase price or car dealer selling price). The degree of influence of low-sensitivity information on the transaction price (i.e., the residual value of the in-use vehicle) is lower than the degree of influence of high-sensitivity information on the transaction price.

[0038] Optionally, the high sensitivity information and low sensitivity information of a certain type of vehicle can be determined based on expert experience or questionnaires, or can be calculated based on statistical methods. Specifically, the input module includes: a receiving unit and a dividing unit. The receiving unit is used to receive the information of the target used car input by the user; the dividing unit is used to use a sensitivity analysis method to divide the information of the target used car into low sensitivity information and high sensitivity information according to the degree of influence of the information of the target used car on the transaction price.

[0039] In a specific implementation, a single factor analysis method is used to determine the sensitivity of information. The transaction price of each vehicle of a certain model is collected in advance, such as an individual transaction price, and the information of each vehicle. The transaction prices with different information but the same other information are summarized, and the degree of influence of the different information on the transaction price when it changes (such as changes in the number of repairs and maintenance, or changes in accident information) can be determined, which can be measured in percentage. Information whose degree of influence exceeds the set value is determined as high-sensitivity information, and information whose degree of influence does not exceed the set value is determined as low-sensitivity information. In actual operation, it may not be possible to collect multiple transaction prices for single information changes, and a multi-factor analysis method can be used to determine the sensitivity of information. Summarize the transaction prices of multiple different information but the same remaining information, and determine the degree of influence of the multiple information changes on the transaction price. The input module sends the high-sensitivity information and the low-sensitivity information to the pre-processing module.

[0040] The pre-processing module is used to quantify the high-sensitivity information and the low-sensitivity information to generate high-sensitivity data and low-sensitivity data.

[0041] Specifically, the text in the high-sensitivity information and the low-sensitivity information is converted into a numerical value using one-hot encoding. The original numerical values ​​in the high-sensitivity information and the low-sensitivity information are retained. Then, the numerical values ​​formed above are normalized to obtain high-sensitivity data and low-sensitivity data.

[0042] The prediction module is used to input the low sensitivity data and the high sensitivity data into the pre-trained prediction model to obtain the intermediate transaction price. This embodiment adopts a prediction model based on deep learning to obtain the intermediate transaction price according to the low sensitivity data and the high sensitivity data. Among them, the intermediate transaction price includes the car dealer's purchase price, the car dealer's selling price and the personal transaction price. That is to say, the prediction model will output a 3D vector, and the 3 values ​​in the vector represent the car dealer's purchase price, the car dealer's selling price and the personal transaction price respectively.

[0043] The structure, training process and functions of the prediction model are described in detail below.

[0044] See also Figure 2The prediction model includes a lightweight gradient boosting machine LightGBM, a fully connected layer, and an output layer. In order to enable the prediction model to have the ability to predict prices, the prediction model needs to be trained first.

[0045] Specifically, training samples are collected, and the training samples include high-sensitivity data and low-sensitivity data of used vehicles for training. Specifically, information of a large number of vehicles that have completed transactions in the historical period (referred to as used vehicles for training) is collected, and the information of used vehicles for training is divided into low-sensitivity information and high-sensitivity information according to the degree of influence of the information of used vehicles for training on the transaction price; then the high-sensitivity information and low-sensitivity information are quantified to generate high-sensitivity data and low-sensitivity data. The labels are the intermediate transaction prices of used vehicles for training: the car dealer's purchase price, the car dealer's selling price, and the personal transaction price. The parameters in the prediction model to be trained are initialized. The low-sensitivity data of used vehicles for training are input into the lightweight gradient boosting machine of the prediction model to be trained, and the high-sensitivity data of used vehicles for training are input into the fully connected layer of the prediction model to be trained, and the intermediate transaction price is output. According to the difference between the intermediate transaction price output by the prediction model (which can be called the training intermediate transaction price) and the actual intermediate transaction price, a loss function is constructed; and the loss function is minimized by iterating the parameters in the prediction model.

[0046] During the use process after the prediction model training is completed, the lightweight gradient boosting machine is used to input low-sensitivity data and output the first eigenvector. The fully connected layer is used to perform weighted summation of high-sensitivity data to obtain the second eigenvector; the output layer is used to fuse the first eigenvector and the second eigenvector to obtain the intermediate transaction price. Among them, the lightweight gradient boosting machine is an improved algorithm for the gradient boosting decision tree. It mainly optimizes the optimal split point search strategy and leaf growth strategy of the decision tree. It has the characteristics of fast training speed, low memory usage, and high accuracy. It is suitable for classification, regression, and sorting tasks. LightGBM is an integrated model composed of several sub-models according to a certain strategy. If the model input value is x, the LightGBM integrated model is , the i-th sub-model is , there are n sub-models in total. Then the relationship between the LightGBM integrated model and the sub-model is as shown in formula (1):

[0047] ; (1)

[0048] in, is the coefficient determined by the sub-model combination strategy. Except for the fitting target of which is the true target, the fitting target of other sub-models is the error of the previous sub-model, which is equivalent to the correction of the previous sub-model. The outputs of these sub-models are weighted and calculated according to a certain combination strategy. In this way, the sub-models that originally had weak predictive ability form an integrated model with strong predictive ability and generalization, which can extract information from complex feature vectors and accurately and effectively evaluate prices. In this embodiment, LightGBM learns the first eigenvector M (a 1×3 eigenvector) that affects the transaction price from low-sensitivity data, and inputs the first eigenvector M into the output layer; at the same time, the eigenvector G composed of multiple high-sensitivity data is input into the fully connected layer. In the fully connected layer, multiple high-sensitivity data are multiplied by the corresponding weights and then summed to obtain the second eigenvector Q, which is actually a value. The output layer is specifically used to multiply the first eigenvector M and the second eigenvector Q correspondingly to obtain the intermediate transaction price R (a 1×3 eigenvector).

[0049] Specifically, assume that for the vehicle input by the user, the high-sensitivity data are vehicle age, mileage, and the number of collision accidents, and the rest are low-sensitivity data. Input the low-sensitivity data into LightGBM to obtain the first feature vector M=(y1, y2, y3), where y1, y2, y3 are the elements in the first feature vector, and the vehicle age g1, mileage g2, and the number of collision accidents g3 constitute the feature vector G and input it into the fully connected layer, where G=(g1, g2, g3). After the fully connected layer, the second feature vector Q=w1×g1+w2×g2+w3×g3 is obtained. It should be noted that w1, w2, w3 in the fully connected layer, and the parameters in LightGBM need to be obtained during the training of the prediction model. After the output layer, R=(y1×Q, y2×Q, y3×Q) is obtained. y1×Q is the price that the dealer buys the car, y2×Q is the price that the dealer sells the car, and y3×Q is the personal transaction price.

[0050] The interval estimation module is used to construct the interval of transaction prices according to the distribution of the error values ​​of the training samples of the prediction model.

[0051] The error value of the training sample is the difference between the intermediate transaction price (which can be called the intermediate transaction price for training) output by the prediction model and the actual intermediate transaction price after the high-sensitivity data and low-sensitivity data of the used cars for training are input into the prediction model. When there are a large number of training samples, the error values ​​of these training samples will show a certain distribution, reflecting the fluctuation of the predicted intermediate transaction price. Taking into account the actual business scenarios and the fluctuation of the evaluation price due to the defects of the evaluation model itself, the present invention adopts a dynamic algorithm of interval estimation based on a sliding window, starting from the error interval of the model, and drawing on the idea of ​​the 3σ principle of the normal distribution. Specifically, the interval estimation module includes: an error statistics unit, a fitting unit and an interval construction unit.

[0052] Among them, the error statistics unit is used to sort the intermediate transaction prices for training. It can be sorted from large to small or from small to large, and if they are the same, they can be sorted in parallel. Since there are three types of intermediate transaction prices for training in this embodiment, each type of training intermediate transaction price (the error value of the dealer's purchase price, the error value of the dealer's selling price, or the error value of the personal transaction price) is sorted separately to obtain three sorting results, so as to obtain the dynamic range of each intermediate transaction price later. After sorting, for each sorting, the training intermediate transaction prices that are close by mistake are placed in a close position. The sliding window method is used to perform window selection on the training intermediate transaction prices (dealer's purchase price, dealer's selling price, or personal transaction price) output by all training samples, and in each window, the training intermediate transaction price located in the middle of the window is determined. and training intervals . See Figure 3 , the size of the sliding window is r+1. The closest training intermediate transaction price to the middle of the window is the r training intermediate transaction prices selected by the window (in Figure 3 The training interval is determined by the mean m and standard deviation d of the error value (i.e., the difference from the actual transaction price) of the training intermediate transaction price in the window; see Figure 3 When the sliding step of the sliding window is 1, there is a sliding window centered on each training intermediate transaction price, and the mean m and standard deviation d of the error value of the training intermediate transaction price in the sliding window are calculated. The multiple k of the standard deviation d is used as the radius of the model error interval to determine the training interval The fitting unit is used to convert the middle transaction price in the middle of the window As the independent variable, the values ​​of the two endpoints of the training interval are taken as the dependent variables, and the first function J1 corresponding to each dependent variable is fitted ( ) and the second function J2 ( ). Optionally, the first function and the second function are fitted using a regression model to obtain the upper and lower bound functions of a dynamic range of a certain intermediate transaction price. The interval construction unit is used to convert the intermediate transaction price obtained by the prediction module into Input them into the first function and the second function respectively to get the transaction price range [J1 ( ), J2 ( ))], the transaction price range includes the range of the dealer's purchase price, the dealer's selling price and the personal transaction price. This embodiment uses the fitting function to directly input the intermediate transaction price obtained by the prediction module into the upper and lower bound functions to obtain the transaction price range, thereby improving the prediction efficiency of the dynamic range in actual business, saving computing power, and realizing the rapid prediction of the dynamic range.

[0053] Optionally, the system further includes: a screening module, which is used to screen out the correct interval of the dealer's purchase price, the dealer's selling price and the personal transaction price according to the set size relationship after constructing the transaction price interval through the interval estimation module. Among them, the set size relationship is determined according to the market situation. Due to the dealer's reservation of profit margins, the dealer's purchase price is the lowest of the three prices, the dealer's selling price is the highest, and the individual car owner sells directly without a middleman to earn the difference, and the transaction price is generally in the middle. Based on this, the set size relationship includes: the upper limit of the dealer's purchase price interval is less than the lower limit of the personal transaction price, and the upper limit of the personal transaction price is less than the lower limit of the dealer's selling price. If the set size relationship is met, the transaction price interval is considered correct, otherwise, the transaction price interval is considered wrong and the prediction result is discarded.

[0054] The output module is used to feed back the transaction price range to the user, and the transaction price range can be specifically displayed on the interface. Since there are three types of transaction price ranges, in order to improve the pertinence and user experience, the transaction price range required by the user can be displayed according to the user's identity. Specifically, the system also includes a login module, which is used to verify the user's identity according to the user's login information, and the identity includes car dealers and individuals. For car dealers, the range of car dealers' purchase prices and car dealers' selling prices can be displayed, and for individuals, the range of personal transaction prices can be displayed.

[0055] The above embodiment describes a scheme in which a user inputs vehicle information for price evaluation. The system can also receive prices to recommend vehicles. Specifically, the input module is also used to receive the price and transaction type (including buy and sell) input by the user; determine the transaction price interval that matches the input price according to the user's identity and transaction type. Assuming that the user is a car dealer, the transaction type is buy, and the input price is 10w, then search for an interval including 10w in all intervals of the car dealer's purchase price pre-stored in the system, such as [9w, 12w], [5w, 10.5w], and then determine the information of the used car according to the matching transaction price interval; and feed back the information of the used car to the user. For example, [9w, 12w] is the interval of the car dealer's purchase price of used car S, then the vehicle attributes, performance characteristics, usage characteristics, historical transaction volume, vehicle model ownership, accident information and repair and maintenance information of used car S are displayed on the system interface.

[0056] The above-mentioned embodiments of the present invention solve the problems existing in the prior art that the residual value assessment of in-use vehicles relies on subjective judgment and has high time and labor costs, and the online assessment standards are not unified, the result errors are large, and the interval range is large. Compared with the prior art, the present invention covers a variety of information that affects the residual value of the vehicle, obtains training samples based on historical transaction data of used cars, uses a prediction model based on LightGBM for price evaluation, and constructs a model-based dynamic interval rapid prediction method to achieve price and dynamic interval prediction covering dealer purchase prices, dealer selling prices, and personal transaction prices, improve the accuracy and efficiency of the prediction, and provide support for residual value assessment business scenarios such as OEMs, used car dealers, and financial institutions.

[0057] Example 2

[0058] See also Figure 4 This embodiment provides a second-hand car transaction information management method, which is applicable to a second-hand car transaction information management system. The method includes:

[0059] S110, receiving information of a target used car input by a user; and dividing the information of the target used car into low-sensitivity information and high-sensitivity information according to the degree of influence of the information of the target used car on the transaction price.

[0060] S120: quantify the high-sensitivity information and the low-sensitivity information to generate high-sensitivity data and low-sensitivity data.

[0061] S130: Input the low-sensitivity data and the high-sensitivity data into a pre-trained prediction model to obtain an intermediate transaction price.

[0062] S140: Constructing a transaction price interval according to the distribution of the training sample error values ​​of the prediction model.

[0063] S150: Feedback the transaction price range to the user.

[0064] This embodiment has the technical features and corresponding technical effects of the above-mentioned system, which will not be described in detail here.

[0065] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A used car transaction information management system, characterized in that: include: An input module, used for receiving information of a target used car input by a user; According to the influence degree of the information of the target used car on the transaction price, the information of the target used car is divided into low-sensitivity information and high-sensitivity information; The information of the target used car includes: vehicle attributes, performance characteristics, usage characteristics, historical transaction volume, vehicle model ownership, accident information and repair and maintenance information; the vehicle attributes include vehicle model, brand and manufacturer's suggested price of new cars; performance characteristics include transmission type, displacement, drive type, horsepower, torque, maximum speed and fuel consumption per 100 kilometers; usage characteristics include vehicle age, usage nature, number of license plate changes and mileage; accident information includes the type and number of accidents that occurred in the used car; and repair and maintenance information includes the type and number of repairs and maintenance performed on the used car; A pre-processing module, used for quantifying the high-sensitivity information and the low-sensitivity information to generate high-sensitivity data and low-sensitivity data; A prediction module, used for inputting the low sensitivity data and the high sensitivity data into a pre-trained prediction model to obtain an intermediate transaction price; wherein the prediction model includes a lightweight gradient boosting machine, a fully connected layer and an output layer; the lightweight gradient boosting machine is used for inputting the low sensitivity data and outputting a first feature vector, the fully connected layer is used for performing weighted summation on the high sensitivity data to obtain a second feature vector; the output layer is used for correspondingly multiplying the first feature vector and the second feature vector to obtain an intermediate transaction price; An interval estimation module, used to construct an interval of transaction prices according to the distribution of the error values ​​of the training samples of the prediction model; the interval of transaction prices includes the interval of the car dealer's purchase price, the interval of the car dealer's selling price and the interval of the personal transaction price; An output module, used for feeding back the transaction price range to the user; Wherein, the interval estimation module comprises: The error statistics unit is used to sort the intermediate transaction prices for training; use the sliding window method to perform window selection on the intermediate transaction prices for training output by all training samples; determine the intermediate transaction price for training and the training interval located in the middle of each window; wherein the training interval is determined according to the mean and standard deviation of the error values ​​of the intermediate transaction prices for training in the window; the training samples include high-sensitivity data and low-sensitivity data of used cars for training; A fitting unit, used for taking the middle transaction price in the middle of the window as an independent variable and the values ​​of the two end points of the training interval as dependent variables, and fitting the first function and the second function corresponding to each dependent variable; The interval construction unit is used to input the intermediate transaction price obtained by the prediction module into the first function and the second function respectively to obtain the interval of the transaction price.

2. The system according to claim 1, characterized in that The input module comprises: A receiving unit, used for receiving information of a target used car input by a user; The classification unit is used to classify the information of the target used car into low-sensitivity information and high-sensitivity information according to the influence of the information of the target used car on the transaction price by using a sensitivity analysis method.

3. The system according to claim 1, characterized in that Also includes: The screening module is used to screen out the correct dealer purchase price interval, dealer selling price interval and personal transaction price interval according to the set size relationship after the transaction price interval is constructed by the interval estimation module.

4. The system according to claim 1, characterized in that The system also includes a login module; The login module is used to verify the identity of the user according to the user's login information, and the identity includes the car dealer and the individual; The input module is also used to: receive the price and transaction type input by the user; The output module is also used to determine a transaction price range that matches the input price according to the identity of the user and the transaction type; determine the information of the used car according to the matching transaction price range; and feed back the used car information to the user.

5. A method for managing second-hand car transaction information, characterized in that: include: Receive information of a target used car input by a user; According to the influence degree of the information of the target used car on the transaction price, the information of the target used car is divided into low-sensitivity information and high-sensitivity information; The information of the target used car includes: vehicle attributes, performance characteristics, usage characteristics, historical transaction volume, vehicle model ownership, accident information and repair and maintenance information; the vehicle attributes include vehicle model, brand and manufacturer's suggested price of new cars; performance characteristics include transmission type, displacement, drive type, horsepower, torque, maximum speed and fuel consumption per 100 kilometers; usage characteristics include vehicle age, usage nature, number of license plate changes and mileage; accident information includes the type and number of accidents that occurred in the used car; and repair and maintenance information includes the type and number of repairs and maintenance performed on the used car; quantifying the high-sensitivity information and the low-sensitivity information to generate high-sensitivity data and low-sensitivity data; Inputting the low-sensitivity data and the high-sensitivity data into a pre-trained prediction model to obtain an intermediate transaction price; wherein the prediction model includes a lightweight gradient boosting machine, a fully connected layer and an output layer; the lightweight gradient boosting machine is used to input the low-sensitivity data and output a first feature vector, the fully connected layer is used to perform weighted summation on the high-sensitivity data to obtain a second feature vector; the output layer is used to multiply the first feature vector and the second feature vector correspondingly to obtain an intermediate transaction price; According to the distribution of the error values ​​of the training samples of the prediction model, a transaction price interval is constructed; the transaction price interval includes a dealer's purchase price interval, a dealer's selling price interval and a personal transaction price interval; Feedback the transaction price range to the user; Among them, the interval estimation module includes: The error statistics unit is used to sort the intermediate transaction prices for training; use the sliding window method to perform window selection on the intermediate transaction prices for training output by all training samples; determine the intermediate transaction price for training and the training interval located in the middle of each window; wherein the training interval is determined according to the mean and standard deviation of the error values ​​of the intermediate transaction prices for training in the window; the training samples include high-sensitivity data and low-sensitivity data of used cars for training; A fitting unit, used for taking the middle transaction price in the middle of the window as an independent variable and the values ​​of the two end points of the training interval as dependent variables, and fitting the first function and the second function corresponding to each dependent variable; The interval construction unit is used to input the intermediate transaction price obtained by the prediction module into the first function and the second function respectively to obtain the interval of the transaction price.

Images