KR20220165479A - Method for estimating vehicle price using neural network and system for the method - Google Patents

Abstract

The present invention relates to a method for estimating a vehicle price using a neural network. According to the present invention, the method for estimating a vehicle price using a neural network performed in a vehicle price estimating server includes: a step of receiving the input of inspection information by vehicle evaluation item which is input by a user; a step of calculating an estimated price by inputting the inspection information in the neural network learned in advance with state information vectorized by being aligned according to a vehicle's structural features; and a step of providing the calculated estimated price. The neural network is learned by using learning data in which the estimated price is labeled in continuously provided state information according to the position relation of the structural features of the vehicle. According to the present invention, the method for estimating a vehicle price using a neural network can provide consumers with an accurate price based on the result of vehicle inspection by using Artificial Intelligence.

Description

Method for estimating vehicle price using neural network and system for the method}

The present invention relates to a method for predicting a market price of a vehicle using a neural network, and more particularly, to a method for calculating a rate of change in a market price based on a condition check of a vehicle.

Recently, with the development of artificial intelligence technology, artificial intelligence technologies are being applied to various fields, and methods for generating additional information by extracting features inherent in data through neural network models have been developed and used instead of existing data processing methods.

A neural network model used for artificial intelligence can quickly and accurately detect and recognize features within input data through learning compared to general data processing. In recent years, artificial intelligence technology has been applied not only to tracking and detecting objects, but also to learning past history and deriving current features that reflect future predictions or time-series change information.

In the case of used products, where the value is not fixed according to depreciation or product condition, along with the vitalization of used transactions, there is a need for consumers to calculate the value of used products appropriately, and neural networks are also being applied to these value calculation fields.

In particular, machine learning technologies are being applied to market price calculation in the used car market, which is relatively expensive and has various variables.

As an example, the prior art (Korean Registered Patent Publication No. 10-2218287 (registration date 2021.02.16)) proposes a method of predicting used car prices through machine learning. However, the method presented in the prior art requires various learning data as a separate learning model is used according to conditions such as vehicle grade, model, and manufacturer. It is difficult to continuously acquire learning data of .

Therefore, in order to more realistically predict the market price, it is necessary to develop a neural network capable of configuring learning data by reflecting vehicle characteristics and reflecting time-series changes.

An object of the present invention is to propose a method of predicting market prices based on vehicle inspection results using artificial intelligence.

In addition, the present invention configures the vehicle inspection information by reflecting the structural position of the vehicle, so that the neural network model learns market price influence information according to the positional relationship.

In addition, the present invention provides more accurate market price information in consideration of temporal variables by mutual learning of a correlation between year and mileage information that causes a natural depreciation of a vehicle.

In the vehicle price prediction method using a neural network performed in a vehicle price prediction server according to the present invention to solve the above technical problem, the step of receiving inspection information for each vehicle evaluation item input from a user; Calculating a predicted market price by arranging the inspection information according to structural characteristics of the vehicle and inputting the vectorized state information to a pre-learned neural network; and providing the calculated predicted market price, wherein the neural network is trained using learning data in which the predicted market price is labeled in continuous state information according to the positional relevance of structural features of the vehicle.

Preferably, the neural network learns the market price influence weight of the structural feature of the vehicle according to the increase or decrease in the year or mileage of the vehicle by using sequential learning data according to the year or mileage of the vehicle of the same type as the learning data.

It is preferable that the front part inspection information and engine room inspection information of the vehicle in the state information are consecutive, and the side part inspection information is continuous with the engine room information and internal inspection information.

Preferably, the neural network learns the expected market price for each condition information according to the model year and the mileage, respectively, and defines a difference between the expected market price for the model year and mileage of the same type of vehicle as a loss and learns a weight between the model year and the mileage. Do.

Preferably, the neural network learns by time-sequentially normalizing state information according to different years or mileage of the vehicles of the same type according to the weights.

Preferably, the neural network outputs a depreciation rate according to the state information of the vehicle, and in the calculating step, the predicted market price is calculated by applying a reference market price reflecting the mileage or year of the vehicle to the output depreciation rate.

A vehicle price prediction server according to the present invention for solving the above technical problem includes an inspection information input unit that receives inspection information for each vehicle evaluation item input from a user; a predicted market price calculation unit for arranging the inspection information according to structural characteristics of the vehicle and inputting vectorized state information to a pre-learned neural network to calculate a predicted market price; a unit providing the calculated market price; and a database that normalizes the inspection information according to the year or mileage of the vehicle and stores it in a time-series manner, wherein the neural network uses learning data in which predicted market prices are labeled in continuous state information according to positional relevance of structural features of the vehicle. It is preferable to learn using .

According to the present invention, it is possible to more accurately provide consumers with market prices based on vehicle inspection results using artificial intelligence.

In addition, consumers can determine the possibility of a false sale through the market price, and can sell or purchase a vehicle at a reasonable price.

In addition, the present invention can increase the convenience of maintenance, which is the original task, and increase the learning efficiency of the neural network by using sequential inspection results according to the convenience of mechanics for checking the condition of the vehicle as learning data.

1 is an exemplary view showing a vehicle market price prediction system according to an embodiment of the present invention.
2 is a flowchart illustrating a vehicle price prediction method using a neural network according to an embodiment of the present invention.
3 is an exemplary view showing an example of vehicle inspection according to an embodiment of the present invention.
4 is an exemplary diagram illustrating an example of vehicle inspection information according to an embodiment of the present invention.
5 is an exemplary diagram illustrating a first neural network model according to an embodiment of the present invention.
6 is an exemplary diagram illustrating a second neural network model according to an embodiment of the present invention.
7 is an exemplary diagram illustrating a method for learning a time-sequential influence of a vehicle according to an embodiment of the present invention.
8 is an exemplary diagram illustrating a time-sequential configuration of vehicle inspection information according to an embodiment of the present invention.
9 is a block diagram showing the configuration of a vehicle price prediction server according to an embodiment of the present invention.

The following merely illustrates the principles of the invention. Therefore, those skilled in the art can invent various devices that embody the principles of the invention and fall within the concept and scope of the invention, even though not explicitly described or shown herein. In addition, all conditional terms and embodiments listed in this specification are, in principle, clearly intended only for the purpose of understanding the concept of the invention, and it should be understood that it is not limited to the specifically listed embodiments and conditions. .

The above objects, features and advantages will become more apparent through the following detailed description in conjunction with the accompanying drawings, and accordingly, those skilled in the art to which the invention belongs will be able to easily implement the technical idea of the invention. .

In addition, in describing the invention, if it is determined that a detailed description of a known technology related to the invention may unnecessarily obscure the subject matter of the invention, the detailed description will be omitted. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is an exemplary view showing a vehicle market price prediction system according to an embodiment of the present invention.

Referring to FIG. 1 , a vehicle price prediction system 1000 may include a vehicle price prediction server 100 and a user terminal 200 .

A user, preferably a vehicle inspector, may use the user terminal 200 to check the state of the vehicle and input the inspection result.

The vehicle price prediction server 100 may receive the input inspection result through an application for inspectors installed in the user terminal 200, reconstruct the received inspection information according to the structural characteristics of the vehicle, vectorize it to have location information, and generate a neural network. can be entered in

Specifically, the inspection information input unit may receive inspection information input by a vehicle inspector and convert it into vector information for input to a neural network. That is, the inspection result for each item can be reconstructed into normalized numerical values according to the category. The order of the reconstructed inspection results may be determined on the data according to the characteristics of each item.

In detail, the inspection information may be sorted according to classification such as a front part, a rear part, a side part, a lower part, and an inner engine room according to structural characteristics of the vehicle.

The above sorted inspection information is input to the neural network, and the neural network of the predicted market price calculation unit can predict the market price based on the input inspection information and provide it to the consumer.

Hereinafter, a vehicle price prediction method using the neural network of the vehicle price prediction server 100 will be described with reference to FIG. 2 .

2 is a flowchart illustrating a vehicle price prediction method using a neural network according to an embodiment of the present invention.

The vehicle price prediction server 100 may receive inspection information for each vehicle evaluation item from the user terminal 200 (S100).

Here, the inspection information can be structured into items classified according to the hierarchy, and the large item is the most comprehensive item and can be composed of a total of five items: exterior, interior and function, engine room and powertrain, wheels and tires, and lower part. .

The middle item may refer to a part or function that exists in a position that is distinguished from each other in the large item. A subitem means a part or component called by a direct name within a subitem, and may be a unit that is actually inspected by an inspector.

In other words, for each sub-item, the inspector divides the exterior, interior and function of the car, engine and powertrain, wheels and tires, and the lower part of the car excluding the frame into exterior / safety as performance points, and then goes through a series of inspection processes to score according to the condition. can be calculated.

In addition, the frames are divided into ranks (ranks) according to the severity of the anomaly, and for example, the first super-severe frame with the most serious abnormality, the second-order high-severe frame, the third-order central angle frame, and the fourth rank. It can be itemized as a low severity frame.

The ultra-severe frame may include, for example, a dash panel, a side member, a floor panel, and a package tray as a core component supporting the vehicle, and a high-severity frame may include wheels for supporting spaces such as driving, driver's seat, and front passenger's seat of the vehicle. House, side members, and filler panels may be included.

The center angle frame may include a front panel and an inside panel, and the low angle frame may include a rear panel and a trunk floor side sill panel.

For the classified items, the inspector checks whether or not there is an abnormality, or checks that there is an abnormality or that a detailed confirmation is required for the part that needs confirmation.

This process may be performed through the inspector's user terminal.

Specifically, referring to FIG. 3, the inspector creates a report 30 using an inspector application installed on his/her user terminal, and the report is a value according to a predetermined item-specific characteristic type and data type, and determines whether or not there is an abnormality. It can be set as a numerical value according to

For example, the condition score for each item can be evaluated as 1 point if there is no abnormality, 0.8 point if confirmation is required, and 0.2 point if there is an abnormality. If additional confirmation is required, it is also possible to recalculate the score through data such as evaluations or images that are additionally attached based on the 0.8 point.

At this time, the numerical value of each sub-item can be used for market price prediction through the weighted average process of the rule base according to the importance of appearance or safety. Specifically, the inspection information digitized for each sub-item may be collected for each sub-item and calculated as an average for each sub-item.

The calculated value for each item can be recalculated again by dividing it into an impact index on the exterior of the vehicle and an impact index on safety. That is, the weighted value according to the importance of appearance or safety may be reflected in the inspection state, and the score may be collected as a final score.

The final subitem score is compared with the performance score based on the perfect score to calculate the rate of decline compared to the standard according to the condition, and the performance score that reflects the calculated drop rate is reflected in the final mileage and model year or the residual value rate of the vehicle according to the mileage. price can be predicted.

In addition, it is also possible to compare the mileage compared to the year and classify it into an appropriate level and non-traveling/overtraveling, and when the driving is higher than the appropriate level, a mileage coefficient is calculated to set the remaining cost rate.

Accordingly, in the case of over-driving, a market price to which depreciation according to the driving distance and additional depreciation according to the condition are applied to the market price relative to the model year may be calculated.

In addition, in this embodiment, in addition to the above rule-based method, the vehicle price prediction server 100 according to this embodiment learns the positional relationship between the structural characteristics of the vehicle and the correlation between the year and mileage by using a neural network We further propose a method to have more accurate and robust performance for various environment variables.

To this end, the inspection information calculated in this embodiment may be reconstructed according to the structural location of the vehicle.

Specifically, referring to FIG. 4 , the inspection information 35 may be sequentially reconstructed according to the front, engine room, side, interior, rear, and lower parts of the vehicle in a categorical manner.

That is, the front fender, engine hood, windshield, headlamp, fog lamp, grill, and bumper constituting the front part of the vehicle are sequentially itemized according to safety importance and external appearance.

Next, the engine room configured at the front of the vehicle is itemized successively to the front configuration. In relation to the engine room, it may include front panels, inside panels, wheel houses, cross members, side members, and other functional items related to engine oil or filters.

Then, in the case of the side part, safety and appearance components include items related to pillars, doors, wheel tires, and brake discs.

In the case of the interior, items include whether the seat, dashboard, lighting system, air conditioner, and other convenience devices are operating.

For the next rear part, it includes the rear fender, rear lamp and bumper, the inside of the trunk, and the rear wheel house. include

According to the above classification, the sub-items are categorized according to the structural position of the vehicle, and as a whole, the successive effects on safety according to the abnormality of the fenders and panels constituting the front and engine room can be learned with weights.

In addition, as a feature of the engine room and the side part, it can be configured to learn by linking the safety risk that is aggravated by the abnormality between the inside panel and the side sill panel, and the safety and appearance effects of the abnormality of the wheel house and A-pillar. .

In addition, the inspection information may include numerical information corresponding to the year and mileage of the vehicle, and the depreciation rate of the vehicle price according to the year and mileage is learned as a weight.

Referring to FIG. 5 , the reconstructed check information 35 allows the neural network 50 to learn the connection strength between pieces of information as a weight.

Therefore, the training data labels the inspection information and the expected market price according to the inspection information so that the weight parameters of the inner hidden layer constituting the neural network can be updated.

First, the neural network is composed of fully-connected layers and can learn weights for each input value.

That is, calculations are performed according to weights and biases for each value of the inspection information, and the expected market price is calculated by finally integrating the corresponding values. Compare the calculated expected price with the labeled price and update the weight by back-propagating the error.

In addition, since the inspection information is arranged in consideration of the structural position of the vehicle in this embodiment, the connection strength between adjacent inputs can be set to be updated with mutual correlation. For example, when updating the value of the front item and the calculation weight for the engine room item, the bias can be adjusted according to whether the items are adjacent in the data structure (or the distance between nodes in the neural network layer) to further reflect the characteristics of the relationship. there is.

Furthermore, in this embodiment, it is also possible to further reflect the change in market price influence according to the year and mileage by constructing the inspection information in a time series based on the year and mileage.

Referring to FIG. 6, in this embodiment, time-series learning data of the same or similar (preferably the same grade model line of the manufacturer) is sequentially input according to the year or mileage, and the market price is predicted in order of past time points or mileage. The result of the calculation of the time can be transferred to the next time point in the next order or the price prediction of the mileage.

Therefore, the neural network 50-1 determines the extent to which not only the market price prediction according to the current inspection information 35 but also the inspection results of various items 35-1, 2, 3, and 4 in the past affect the current market price prediction. can be learned, and through this, items that have a large impact on market price changes according to the year are extracted and reflected in market price prediction.

In this regard, the neural network (50-1) for market price prediction can be configured in the form of a recurrent neural network, and by learning the neural network time-sequentially using the widely used LSTM (Long Short-Term Memory), the next market price Separate the information that affects the prediction from the information that is not to be transmitted, and reflect it in the calculation.

In addition, for time-sequential learning, it is necessary to secure inspection information according to regular time for the same or similar vehicle models. However, since it may be difficult to acquire inspection information sufficient for learning in a balanced manner according to the model year or mileage, the inspection information of the corresponding vehicle type is temporally normalized using the model year and mileage.

Referring to FIG. 7 , in this embodiment, learning can be performed to minimize an error by defining the difference between the output results of the expected market price of the neural network as a loss with only the year or mileage as an input for inspection information of the same vehicle type. Through this, it is possible to learn the relationship between the effect of the model year of a specific vehicle and the effect of change in mileage. Based on the relationship between the learned year and mileage, the year and mileage of the inspection information 35 are normalized and input in time series as shown in FIG.

Hereinafter, the configuration of the vehicle price prediction server 100 will be described with reference to FIG. 9 .

9 is a block diagram showing the configuration of a vehicle price prediction server 100 according to an embodiment of the present invention.

Referring to FIG. 9 , the vehicle price prediction server 100 may include an inspection information input unit 110, a predicted market price calculation unit 120, a market price providing unit 130, and a learning data database 140.

The inspection information input unit 110 receives vehicle state information input by an inspector using a user terminal, preferably a vehicle inspection terminal.

An inspector may select condition information for each item determined in advance according to a category of a vehicle for inspection, and the selected condition information may be digitized and used as vector information. Additionally, it is possible to extract feature information through direct image-based convolution operation in addition to state information by allowing the inspector to attach a photo or image data if a state requires detailed confirmation.

The inspection information input unit 110 may sequentially reconstruct the input inspection information according to the structural location of the vehicle. Depending on the habit of the inspector or the characteristics of the inspection application, the inspection order or sub-item classification criteria may vary, so to normalize these values, they can be reclassified based on the structural location of the vehicle.

The inspection information reconstructed through the above process is used to calculate the predicted market price through the neural network.

The predicted market price calculation unit 120 calculates the current predicted market price of the vehicle by inputting inspection information to the neural network. Alternatively, if the neural network is trained to predict the depreciation rate relative to the reference value instead of the predicted market price, the neural network may calculate an additional depreciation rate compared to the model year based on vehicle inspection information.

Specifically, the neural network learns whether there is an additional effect on mutual safety or appearance according to the structural positional relationship along with the market price impact index according to the condition of each item inherent in the inspection information itself, thereby improving safety or performance according to the organic structure of the vehicle. Learn the effects of appearance.

In addition, natural depreciation occurs according to the year and mileage of the vehicle, and by learning this as time-series inspection information, elements that have a greater impact according to time changes are extracted and learned from the past status information for each vehicle type. .

Accordingly, the predicted market price calculation unit 120 more accurately calculates the depreciation rate of the condition-based value of the vehicle.

The market price providing unit 130 may provide the predicted market price calculated according to the above process to the user. That is, it is possible to provide an estimated market price directly calculated from the neural network, or to calculate and provide a market price by applying an additional depreciation rate according to the state value compared to the natural depreciation rate when the depreciation rate is predicted.

Furthermore, the learning data DB 140 can store and manage the inspection information used for market price prediction to additionally use it for learning, and use the inspection information at this time to manipulate vehicle state information or provide additional services. let it be

According to the present invention, it is possible to more accurately provide consumers with market prices based on vehicle inspection results using artificial intelligence.

In addition, consumers can determine the possibility of a false sale through the market price, and can sell or purchase a vehicle at a reasonable price.

In addition, the present invention can increase the convenience of maintenance, which is the original task, and increase the learning efficiency of the neural network by using sequential inspection results according to the convenience of mechanics for checking the condition of the vehicle as learning data.

Furthermore, various embodiments described herein may be implemented in a recording medium readable by a computer or a device similar thereto using, for example, software, hardware, or a combination thereof.

According to hardware implementation, the embodiments described herein include application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), It may be implemented using at least one of processors, controllers, micro-controllers, microprocessors, and electrical units for performing other functions. The described embodiments may be implemented in the control module itself.

According to software implementation, embodiments such as procedures and functions described in this specification may be implemented as separate software modules. Each of the software modules may perform one or more functions and operations described herein. The software code may be implemented as a software application written in any suitable programming language. The software code may be stored in a memory module and executed by a control module.

The above description is merely an example of the technical idea of the present invention, and those skilled in the art can make various modifications, changes, and substitutions without departing from the essential characteristics of the present invention. will be.

Therefore, the embodiments disclosed in the present invention and the accompanying drawings are not intended to limit the technical idea of the present invention, but to explain, and the scope of the technical idea of the present invention is not limited by these embodiments and the accompanying drawings. . The protection scope of the present invention should be construed according to the following claims, and all technical ideas within the equivalent range should be construed as being included in the scope of the present invention.

Claims (8)

In the vehicle price prediction method using a neural network performed in a vehicle price prediction server,
Receiving inspection information for each vehicle evaluation item input from a user;
Calculating a predicted market price by arranging the inspection information according to structural characteristics of the vehicle and inputting the vectorized state information to a pre-learned neural network; and
Including providing the calculated predicted market price,
The method of predicting vehicle market price using a neural network, characterized in that the neural network is learned using learning data in which predicted market prices are labeled in continuous state information according to positional relevance of structural features of the vehicle. According to claim 1,
The neural network uses the learning data sequentially according to the year or mileage of the same type of vehicle to learn the market price influence weight of the structural feature of the vehicle according to the increase or decrease in the year or mileage of the vehicle using a neural network. How to predict market price. According to claim 1,
Vehicle price prediction method using a neural network, characterized in that the front part inspection information and engine room inspection information of the vehicle in the state information are continuous, and the side inspection information is continuous with the engine room information and internal inspection information. According to claim 2,
The neural network learns the expected market price for each state information according to the model year and the state information according to the mileage, and defines the difference between the expected market price for the model year and the mileage of the same type of vehicle as a loss to learn the weight between the model year and the mileage. A vehicle price prediction method using a neural network. According to claim 4,
The method of predicting vehicle price using a neural network, characterized in that the neural network learns by time-sequentially normalizing state information according to different years or mileage of the same type of vehicle according to the weight. According to claim 1,
The neural network outputs a depreciation rate according to the state information of the vehicle,
Wherein the calculating step calculates the predicted market price by applying a reference market price reflecting the mileage or year of the vehicle to the output depreciation rate. In a vehicle price prediction server using a neural network,
an inspection information input unit that receives inspection information for each vehicle evaluation item input from a user;
a predicted market price calculation unit for arranging the inspection information according to structural characteristics of the vehicle and inputting vectorized state information to a pre-learned neural network to calculate a predicted market price;
a unit providing the calculated market price; and
A database that normalizes the inspection information according to the year or mileage of the vehicle and stores it in time series;
The vehicle price prediction server using a neural network, characterized in that the neural network is learned using learning data in which predicted market prices are labeled in continuous state information according to positional relevance of structural features of the vehicle. In the vehicle price prediction method using a neural network performed in a vehicle price prediction server,
Step of receiving inspection information for each vehicle evaluation item input from the user
Calculating a predicted market price by arranging the inspection information according to structural characteristics of the vehicle and inputting the vectorized state information to a pre-learned neural network;
Including providing the calculated predicted market price,
The neural network is a computer readable recording medium for performing a vehicle market price prediction method using a neural network, characterized in that the neural network is learned using learning data in which the predicted market price is labeled in continuous state information according to the positional relevance of the structural feature of the vehicle. stored program.

Images