KR102218287B1 - Method and system for a used car through machine learning - Google Patents

Abstract

The present disclosure relates to a method for predicting the market price of a used car through machine learning. The method for predicting the market price of a used car comprises the steps of: receiving a data set about a plurality of vehicles for which a transaction has been completed; classifying the data set about the plurality of vehicles, for which a transaction has been completed according to a predetermined condition, into a plurality of data subsets; generating a plurality of prediction models, which are configured to deduce the price of a used car regarding the plurality of classified data subsets; receiving a request for predicting the market price of a target used car and extracting one or more prediction models corresponding to the target used car from the plurality of generated prediction models; and calculating the market price of the target used car on the basis of one or more extracted prediction modules in response to the request for predicting the market price of the target used car. The present invention enables a user to easily and quickly find and purchase a desired product at a desired price.

Description

Used car price prediction method and system through machine learning {METHOD AND SYSTEM FOR A USED CAR THROUGH MACHINE LEARNING}

The present disclosure relates to a method and system for predicting used car market prices through machine learning, and more specifically, to a used car price prediction method and system capable of calculating the market price of a used car by generating a predictive model through ensemble learning. .

In general, a used car refers to a car that has been operated for a certain period after a new car is released. These used cars are actively traded through electronic commerce with the development of communication and network technologies. Recently, as the market's interest in AI-related companies has exploded, the used car industry is also actively introducing AI-based services. Along with this trend, as the trend of sharing cars rather than owning them among consumers is becoming more common, the used car trading volume is rapidly increasing, and the used car trading market is growing rapidly.

However, compared to this rapid growth, the reliability of the used car trading market is not high for users who use used car trading. Due to the nature of the used car market, vehicle information such as accident history and parts replacement is not transparently disclosed. Due to such market opacity, asymmetry of information between a consumer who wants to buy a used car and a seller who wants to sell occurs, and a lot of damage is caused by a false sale.

In addition, the price of a used car can be calculated from various factors such as year, mileage, accident history, vehicle condition, options, type of transmission, color, usage, fashion, and region. It is difficult for a buyer of a used car to accurately judge the appropriateness of the price of a used car suggested by a seller of a used car or to predict a change in market price. Accordingly, there is a need for a method that enables consumers to determine the adequacy of prices by predicting market price fluctuations based on 　 used cars 　 prices actually traded in the used car trading market.

The present disclosure provides a method, a system, and a computer program stored in a recording medium to solve the above problems.

Classify data on a plurality of vehicles for which transactions have been completed into a plurality of data subsets, generate a plurality of predictive models configured to infer used car prices for a plurality of data subsets, and respond to a target used car among the plurality of generated predictive models A used car market price prediction method, a system and a computer program stored in a recording medium for calculating the market price of a target used car by extracting one or more prediction models are provided.

The present disclosure may be implemented in a variety of ways, including a method, system, apparatus or computer readable storage medium storing instructions.

A method for predicting a used car market price according to an embodiment of the present disclosure includes receiving data sets for a plurality of vehicles for which transactions have been completed, classifying data sets for a plurality of vehicles for which transactions have been completed according to a predetermined condition into a plurality of data subsets Generating a plurality of predictive models configured to infer the price of a used car for a plurality of classified data subsets, receiving a request for price prediction for a target used car, and among the plurality of generated prediction models, a target used car And extracting at least one prediction model corresponding to and in response to a request for price prediction for the target used car, calculating a market price for the target used car based on the extracted at least one prediction model.

According to an embodiment, the predetermined condition includes a plurality of conditions, and each of the plurality of data subsets is stored in association with each of the plurality of conditions, and data on a plurality of vehicles that have completed a transaction according to the predetermined condition Classifying the set into a plurality of data subsets corresponds to each of the data sets corresponding to each of the at least some conditions satisfied when each of the data sets for the plurality of vehicles for which the transaction has been completed satisfies at least some of the plurality of conditions. And classifying it as a subset of data.

According to an embodiment, the plurality of conditions include vehicle grade conditions according to year, vehicle grade conditions, detailed model conditions, model conditions, manufacturer conditions, and all sales conditions, and each of the plurality of data subsets is a predetermined value. It has more than one parameter.

According to an embodiment, generating a plurality of predictive models configured to infer a market price of a used car for a plurality of classified data subsets includes, for each of the plurality of classified data subsets, through a machine learning algorithm. And generating a predictive model configured to infer a market price of a used car.

According to an embodiment, generating a plurality of predictive models configured to infer a market price of a used car for a plurality of classified data subsets includes: each of the plurality of data subsets corresponding to the vehicle class condition according to the year Generating a plurality of first prediction model candidates through ensemble learning using, generating a plurality of second prediction model candidates through ensemble learning using each of a plurality of data subsets corresponding to vehicle class conditions, Generating a plurality of third prediction models through ensemble learning using each of the plurality of data subsets corresponding to the detailed model conditions, and multiple through ensemble learning using each of the plurality of data subsets corresponding to the model conditions Generating a fourth prediction model candidate of, generating a plurality of fifth prediction model candidates through ensemble learning using each of a plurality of data subsets corresponding to the manufacturer's condition, and data subs corresponding to all selling conditions And generating a plurality of sixth prediction model candidates through ensemble learning using the set.

According to an embodiment, each of the plurality of classified data subsets includes a plurality of training data subsets and test data subsets, and for each of the classified plurality of data subsets, a used car through a machine learning algorithm. The step of generating a prediction model configured to infer the price of each prediction model using a plurality of training data subsets so that each of the plurality of prediction model candidates corresponding to each of the plurality of data subsets has a minimum error Extracting the candidate hyperparameters and selecting a prediction model having the smallest error when verifying using the test data subset from among a plurality of prediction model candidates corresponding to each of the plurality of data subsets.

According to an embodiment, calculating a market price for a target used car includes selecting at least some of a plurality of conditions that the target used car satisfies, through a prediction model selected for each of a plurality of selected data subsets. And calculating a market price for the target used car and calculating a final market price for the target used car through ensemble learning by applying a weight to each of the calculated market prices for the used car.

According to an embodiment, the request for price prediction for the target used car includes a future prediction condition for the target used car, and the step of calculating the market price for the target used car satisfies at least some of the plurality of conditions that the target used car satisfies. Selecting, calculating a future market price for a target used car through a predictive model selected for each of a plurality of data subsets corresponding to the selected at least some conditions, and applying a weight to each of the calculated future market prices for the target used car. And calculating the final future market price for the target used car through ensemble learning.

According to an embodiment, the receiving of a data set for a plurality of vehicles for which the transaction has been completed includes, among data sets for a plurality of vehicles, a data set having the same property, a data set having a vehicle price error, and a driving distance error. Filtering at least one data set of the data set.

In an embodiment of the present disclosure, a computer program stored in a computer-readable recording medium is provided in order to execute the method for predicting used car market prices described above in a computer.

A used car market price prediction system according to an embodiment of the present disclosure receives data sets for a plurality of vehicles that have been traded, and classifies data sets for a plurality of vehicles that have been traded into a plurality of data subsets according to a predetermined condition. , For a plurality of classified data subsets, a learning unit configured to generate a plurality of predictive models configured to infer a market price of a used car, and a target used car receiving a request for price prediction, and among the plurality of generated prediction models, a target And a market price calculation unit configured to extract one or more prediction models corresponding to the used car, and calculate a market price for the target used car based on the extracted one or more prediction models in response to a request for price prediction for the target used car.

According to some embodiments of the present disclosure, a plurality of prediction models may be generated using data on a plurality of vehicles for which transactions have been completed received from an external device, and an ensemble model may be performed using the generated prediction models. By generating the used car market price fluctuations more accurately, the reliability of the calculated market price can be improved, and thus, consumers who want to purchase a used car can determine the appropriateness of the price for the used car.

According to some embodiments of the present disclosure, in response to a market price prediction request for a target used car, a market price prediction system having continuity in market price fluctuations by extracting one or more prediction models from a plurality of prediction models to calculate a market price for a target used car Can provide.

According to some embodiments of the present disclosure, in response to a market price prediction request for a target used car, a weight is applied to the price of each of the extracted prediction models to calculate the final market price for the target used car through ensemble learning, Fluctuations can be detected quickly.

The effects of the present disclosure are not limited to the effects mentioned above, and other effects that are not mentioned are obvious to those of ordinary skill in the technical field belonging to the present disclosure from the description of the claims (hereinafter, ``ordinary technician''). Can be understood.

Embodiments of the present disclosure will be described with reference to the accompanying drawings described below, where like reference numerals denote like elements, but are not limited thereto.
1 is a schematic diagram illustrating a configuration in which a used car price prediction system according to an embodiment of the present disclosure is connected to enable communication with an external device.
2 is a block diagram showing a detailed configuration of a used car price prediction system according to an embodiment of the present disclosure.
3 is an exemplary diagram illustrating a process of classifying data on a plurality of vehicles for which transactions have been completed into a plurality of data subsets according to an embodiment of the present disclosure.
4 is an exemplary view showing a process of generating a plurality of prediction models capable of inferring a market price of a used car according to an embodiment of the present disclosure.
5 is an exemplary view showing an example of calculating a market price for a target used car in response to a market price prediction request according to an embodiment of the present disclosure.
6 is a flowchart illustrating a method for predicting a used car market price according to an embodiment of the present disclosure.

Hereinafter, with reference to the accompanying drawings, specific details for the implementation of the present disclosure will be described in detail. However, in the following description, when there is a possibility that the subject matter of the present disclosure may be unnecessarily obscure, detailed descriptions of widely known functions or configurations will be omitted.

In the accompanying drawings, the same or corresponding elements are assigned the same reference numerals. In addition, in the description of the following embodiments, overlapping descriptions of the same or corresponding components may be omitted. However, even if description of a component is omitted, it is not intended that such component is not included in any embodiment.

Advantages and features of the disclosed embodiments, and a method of achieving them will become apparent with reference to the embodiments described below together with the accompanying drawings. However, the present disclosure is not limited to the embodiments disclosed below, but may be implemented in a variety of different forms, and only these embodiments make the present disclosure complete, and those skilled in the art to which the present disclosure pertains. It is provided only to fully inform the person of the scope of the invention.

The terms used in the present specification will be briefly described, and the disclosed embodiments will be described in detail. The terms used in the present specification have selected general terms that are currently widely used as possible while considering functions in the present disclosure, but this may vary according to the intention or precedent of a technician engaged in a related field, the emergence of new technologies, and the like. In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning of the terms will be described in detail in the description of the corresponding invention. Therefore, the terms used in the present disclosure should be defined based on the meaning of the term and the contents of the present disclosure, not the name of a simple term.

In this specification, expressions in the singular include plural expressions, unless the context clearly specifies that they are singular. In addition, plural expressions include expressions in the singular unless clearly specified as plural in context. When a part of the specification is said to "include" a certain component, it means that other components may be further included rather than excluding other components unless otherwise stated.

In addition, the terms "module" or "unit" used in the specification mean software or hardware components, and "module" or "unit" performs certain roles. However,'module' or'unit' is not meant to be limited to software or hardware. The'module' or'unit' may be configured to be in an addressable storage medium, or may be configured to reproduce one or more processors. Thus, as an example, the'module' or'unit' refers to components such as software components, object-oriented software components, class components and task components, processes, functions, properties, It may include at least one of procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, or variables. Components and the functions provided within the'module' or'unit' may be combined into a smaller number of components and'module' or'unit', or additional components and'module' or'unit'. Can be further separated.

According to an embodiment of the present disclosure, a'module' or'unit' may be implemented with a processor and a memory. 'Processor' should be interpreted broadly to include general purpose processors, central processing units (CPUs), microprocessors, digital signal processors (DSPs), controllers, microcontrollers, state machines, etc. In some circumstances, a'processor' may refer to an application specific semiconductor (ASIC), programmable logic device (PLD), field programmable gate array (FPGA), and the like. 'Processor' refers to a combination of processing devices such as, for example, a combination of a DSP and a microprocessor, a combination of a plurality of microprocessors, a combination of one or more microprocessors in combination with a DSP core, or any other such combination of configurations. You may. In addition, the'memory' should be broadly interpreted to include any electronic component capable of storing electronic information. 'Memory' refers to random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRAM), programmable read-only memory (PROM), erase-programmable read-only memory (EPROM), It may refer to various types of processor-readable media such as electrically erasable PROM (EEPROM), flash memory, magnetic or optical data storage, registers, and the like. The memory is said to be in electronic communication with the processor if it can read information from and/or write information to the memory. The memory integrated in the processor is in electronic communication with the processor.

In the present disclosure, the'system' may include at least one of a server device and a cloud device, but is not limited thereto. For example, the system may consist of one or more server devices. As another example, the system may consist of one or more cloud devices. As another example, in the system, a server device and a cloud device may be configured and operated together.

In the present disclosure, a'prediction model' may refer to a'prediction model candidate', and conversely, a'prediction model candidate' may refer to a'prediction model'.

In the present disclosure, the'hyperparameter of the prediction model' or the'hyperparameter of the prediction model candidate' may refer to a parameter representing or characterizing a tree represented by a prediction model or a prediction model candidate. For example, the parameters may include a depth of a tree, an average of the number of trees, the number of leaf nodes, and a branch criterion of a tree, but are not limited thereto.

1 is a schematic diagram illustrating a configuration in which a used car price prediction system 130 according to an embodiment of the present disclosure is connected to enable communication with external devices 110_1 and 110_2. The used car price prediction system 130 may be configured to communicate with an external device through the communication network 120. The used car price prediction system 130 may communicate with an external device through the communication network 120 to receive data on a plurality of vehicles for which transactions have been completed. Here, the data on the plurality of vehicles for which the transaction has been completed may include a data set for the plurality of vehicles for which the transaction has been completed. The data set for the plurality of vehicles for which the transaction has been completed may include new car price data and performance check data for used cars, but is not limited thereto. For example, the used car price prediction system 130 may be included in one or more server devices and/or cloud devices. As another example, the used car price prediction system 130 may be included in the user terminal.

According to an embodiment, the used car price prediction system 130 is a plurality of vehicles that have been periodically or aperiodically traded from an external device such as a new car price DB 110_1 and a performance test DB 110_2 through the communication network 120 The data set for can be received and stored in a storage unit. In FIG. 1, the new car price DB 110_1 and the performance test DB 110_2 are shown as examples of external devices, but are not limited thereto, and wired and/or wireless communication with the used car price prediction system 130 is possible, and a plurality of transactions have been completed. It may be any computing device capable of receiving a data set for a vehicle of a vehicle.

The communication network 120 may be configured to enable communication between the used car price prediction system 130 and an external device. Depending on the installation environment, the communication network 120 is, for example, a wired network such as Ethernet, a wired home network (Power Line Communication), a telephone line communication device, and RS-serial communication, a mobile communication network, or a wireless LAN (WLAN). , Wi-Fi, Bluetooth, and wireless networks such as ZigBee, or a combination thereof.

According to an embodiment, the used car market price prediction system 130 may store, provide, and execute computer-executable programs and data related to used car price prediction, and one or more server devices and/or databases, or one or more cloud computing services-based Distributed computing devices and/or distributed databases. For example, the used car price prediction system 130 is a computing device capable of communicating with other devices through a wired or wireless network, and includes a central processing unit (CPU), a graphic processing unit (GPU), and a digital signal processor (DSP). It may be a computing device capable of performing an operation using a processing device such as, but is not limited thereto.

The used car market price prediction system 130 classifies data sets for a plurality of vehicles that have been transaction-completed received from an external device into a plurality of data subsets, and provides a plurality of prediction models for inferring the price of used cars for the plurality of data subsets. Can be generated. The used car market price prediction system 130 may more accurately predict the market price fluctuation of the used car by receiving a market price prediction request for a target used car, extracting one or more prediction models from among a plurality of prediction models and calculating the market price for the used car. Here, the market price prediction request for the target used car may be received from one or more user terminals and/or the used car sales system, and the calculated market price for the used car is transmitted to the user terminal and/or the used car sales system that requested the price prediction. 120) can be provided. A process of predicting the market price of a used car using data provided from an external device will be described in detail below with reference to FIGS. 2 to 6.

2 is a block diagram showing a detailed configuration of a used car price prediction system 130 according to an embodiment of the present disclosure. Hereinafter, a specific method of more accurately predicting market price fluctuations of used cars by generating a prediction model through ensemble learning using data sets of a plurality of vehicles for which the used car market price prediction system 130 has completed transactions will be described. It will be described in detail. In an embodiment, the used car price prediction system 130 may be configured to include a communication unit 210, a storage unit 220 and a processor 230, as shown in FIG. 2.

The communication unit 210 may be configured to receive data sets for a plurality of vehicles for which transactions have been completed from an external device through a communication network. According to an embodiment, the communication unit 210 may receive a data set for a vehicle in which a transaction has been completed among used vehicles from an external device every preset period. Here, the data set for the vehicle for which the transaction has been completed may include data related to the price of a new vehicle, a performance check of a used vehicle, the type of fuel of the vehicle, the vehicle driving distance, the vehicle operation date, and whether an accident has occurred, but is not limited thereto. It may further include various information related to the used car that has been traded, such as data (eg, actual transaction price data according to a detailed vehicle model). In addition, the communication unit 210 may provide the received data to the processor 230, and transmit information on the market price of the used car predicted by the processor 230 and/or the information analyzed or processed through the communication network. Can be configured to transmit via. For example, the processor 230 may transmit a quote for a used car calculated to a user terminal or a system that has transmitted the prediction request through the communication unit 210 in response to a price prediction request for a target used car.

The processor 230 includes at least one of a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a field programmable gate array (FPGA), and an application specific integrated circuit (ASIC). It may be configured to perform an operation and manage, process, and/or store a data set received from an external device and/or a data set obtained by analyzing or processing the received data set in the storage unit 220. In one embodiment, the processor 230 may be configured to include a communication unit 210, a learning unit and a price calculation unit 234.

The processor 230 may store a data set of a plurality of vehicles for which transactions have been completed received from an external device in the form of a database in the storage unit 220. According to an embodiment, the processor 230 may filter data sets of a plurality of vehicles for which transactions have been completed, and store the filtered data sets of a plurality of vehicles in the storage unit 220 in the form of a database. For example, the processor 230 is configured to filter at least one data set of a data set having the same property, a data set having a vehicle price error, and a data set having a driving distance error, among data sets for a plurality of vehicles. Can be. In FIG. 2, the storage unit 220 is shown to be included in the used car price prediction system 130, but is not limited thereto, and the storage unit 220 communicates with the used car price prediction system 130 by wire and/or wirelessly. It can be included in possible external devices.

The learning unit 232 may classify data sets for a plurality of vehicles for which transactions have been completed according to a predetermined condition into a plurality of data subsets. Here, the predetermined condition may include a plurality of conditions, and each of the plurality of classified data subsets may be stored in association with each of the plurality of conditions. According to an embodiment, the plurality of conditions may include vehicle grade conditions, detailed model conditions, model conditions, manufacturer conditions, and all sale conditions according to a year.

The learning unit 232 classifies each of the data sets as a data subset corresponding to each of the satisfied at least some conditions when each of the data sets for the plurality of vehicles for which the transaction has been completed satisfies at least some of the plurality of conditions. Can be configured to In addition, each of the plurality of data subsets may include a subset of training data for training a plurality of prediction model candidates generated through each subset of data and a subset of test data for applying a plurality of learned prediction model candidates. have. According to an embodiment, the learning unit 232 may classify each classified data subset into a training data subset and a test data subset. For example, the learning unit 232 may sample at least a portion of the data subset corresponding to a specific condition and classify it as training data for each data subset, and the remaining unsampled data among each data subset Can be classified as test data. Here, the learning unit 232 may sample a subset of data corresponding to a specific condition several times to classify and store a plurality of subsets of training data and a subset of unsampled test data.

Here, the training data may also be used as verification data for extracting hyperparameters of each of a plurality of prediction models generated from each data subset.

The learning unit 232 may be configured to include a prediction model generator 240 and a prediction model verification unit 242. The prediction model generation unit 240 may be configured to generate a plurality of prediction models configured to infer a market price of a used car for a plurality of classified data subsets. According to an embodiment, for each of a plurality of classified data subsets, it may be configured to generate a plurality of predictive models configured to infer a market price of a used car through a machine learning algorithm. In addition, a plurality of generated predictions may be trained by the prediction model generator 240. In this case, the machine learning algorithm may be ensemble learning. For example, the prediction model generation unit 240 generates a plurality of prediction models for a subset of data corresponding to each of a plurality of conditions using an ensemble learning algorithm, and generates an ensemble model using the generated prediction model. can do. Here, the ensemble learning algorithm may include a bagging algorithm, a boosting algorithm, and the like. For example, the boosting algorithm may include an XGBoost (Extreme Gradient Boosting) algorithm. A process of generating an ensemble model using the XGBoost algorithm will be described in detail below with reference to FIG. 4.

According to an embodiment, the prediction model generator 240 generates a feature vector by extracting a feature from each of a plurality of classified training data subsets, and inputs each of the generated feature vectors. Multiple prediction models can be generated and trained using values. In this case, values such as the type of fuel, vehicle driving distance, vehicle driving date, new vehicle price, and whether an accident has occurred may be extracted as a feature.

The prediction model verification unit 242 extracts hyperparameters of each prediction model candidate generated using a subset of the training data, and determines each prediction model candidate to which each hyperparameter extracted using the subset of test data is applied. It can be configured to evaluate. According to an embodiment, the prediction model verification unit 242 uses a plurality of training data subsets so that each prediction model candidate corresponding to each of the plurality of data subsets has a minimum error. It can be configured to extract candidate hyperparameters. In addition, the prediction model verification unit 242 may be configured to select a prediction model having the smallest error when verifying using a test data subset from among a plurality of prediction model candidates corresponding to each of the plurality of data subsets. have. For example, the prediction model verification unit 242 may extract hyperparameters of each prediction model candidate using Grid Search. Alternatively, the prediction model verification unit 242 may extract hyperparameters of each prediction model candidate using Random Search.

The price calculation unit 234 may be configured to receive a price prediction request for a target used car from the communication unit 210 and extract one or more prediction models corresponding to the target used car from among a plurality of generated prediction models. In response to a market price prediction request for the target used car, the market price calculation unit 234 may calculate a market price for the target used car based on the extracted one or more prediction models. Here, the market price prediction request for the target used car may include any information or conditions that can indicate or characterize the target used car, and for example, at least one of vehicle class, vehicle year, vehicle model, vehicle detailed model, and manufacturer. One can contain information or conditions.

According to an embodiment, the price calculation unit 234 selects at least some conditions from among a plurality of conditions that the target used car satisfies, and generates a prediction model selected for each of a plurality of data subsets corresponding to the selected at least some conditions. It can be configured to calculate the market price for the target used car. Then, the market price calculation unit 234 may be configured to calculate a final market price for the target used car through ensemble learning by applying a weight to each of the calculated market prices for the target used car. For example, among conditions representing or characterizing a target used car through ensemble learning, a high weight may be applied to a prediction model corresponding to a more specific and detailed condition.

Alternatively, or additionally, the market price calculation unit 234 may be configured to calculate a future market price for the target used car based on one or more extracted prediction models in response to a request for price prediction for the target used car. Here, the market price prediction request for the target used car may further include not only arbitrary information or conditions capable of indicating or characterizing the target used car, but also a future prediction condition for the target used car. Here, the future prediction condition may indicate any information or conditions capable of representing or characterizing the target used car at a certain point in the future (eg, 6 months, 1 year, 2 years, etc.). For example, the future prediction conditions may be arbitrary information or conditions on a target used car related to a temporal sequence, such as a predicted driving distance after one year, a predicted accident rate after one year, and an expected popularity of a target used car after one year.

According to an embodiment, the future prediction condition may calculate a future prediction condition by using the current condition (eg, total driving distance up to now, total driving period, total number of accidents, etc.) for a target used car. For example, the average daily driving distance can be calculated by using the total driving distance and driving period of the target used car, and the estimated driving distance after one year is calculated using the calculated average daily distance to calculate the estimated driving distance after one year. Can be calculated. As another example, an accident rate after a certain point in the future may be calculated based on a driving habit (eg, an accident rate) that drivers of a target used car have driven in a specific period, and the calculated accident rate may be included in a future prediction condition. As another example, the popularity of the target used car after a certain point in the future is calculated based on a predetermined equation for calculating the popularity, and the calculated popularity may be included in a future prediction condition.

According to an embodiment, the price calculation unit 234 selects at least some of a plurality of conditions that the target used car satisfies in response to a market price prediction request for a target used car including a future prediction condition, and selects at least some A future market price for a target used car may be calculated through a prediction model selected for each of a plurality of data subsets corresponding to the condition. Then, the market price calculation unit 234 may be configured to calculate a final future market price for the target used car through ensemble learning by applying a weight to each of the calculated future market prices for the target used car. For example, among conditions representing or characterizing a target used car through ensemble learning, a high weight may be applied to a prediction model corresponding to a condition related to a temporal sequence. A process of calculating the market price for the target used car in response to the market price prediction request for the target used car by the market price calculation unit 234 will be described in detail below with reference to FIG. 5.

3 is an exemplary view showing a process of classifying a data set 310 for a plurality of vehicles for which transactions have been completed into a plurality of data subsets 312, 314, 316, 318, 320, 322 according to an embodiment of the present disclosure. to be. According to an embodiment, the learning unit 232 determines a data set for a plurality of vehicles for which transactions have been completed, according to a vehicle grade data subset 312, a vehicle grade data subset 314, and a detailed model data subset 316 according to the year. ), model data subset 318, manufacturer data subset 320, and all listing data subset 322.

According to an embodiment, the learning unit 232 is a data subset corresponding to the conditions of all sales (for example, data on all domestic and imported vehicles for which transactions have been completed) with respect to the data set 310 received from an external device. 322). In this case, since all of the property data subset 322 includes a data set for all properties, it may be made of one data subset.

In addition, the learning unit 232 may classify each of the data sets for a plurality of vehicles for which transactions have been completed into, for example, a data subset 320 corresponding to each manufacturer's condition, such as Company A, Company B, and Company C. have. At this time, since data subsets corresponding to each of the plurality of manufacturer conditions are classified, the manufacturer data subset 320 may include a plurality of data subsets, as shown.

In addition, the learning unit 232 uses the data set 310 as a vehicle model condition of each manufacturer, such as, for example, A model of company A, B model of company A, C model of company A, D model of company B, C model of company B, etc. That is, it can be classified into a data subset 318 corresponding to each model condition. At this time, since data subsets corresponding to each of the plurality of model conditions are classified, the model data subset 318 may include a plurality of data subsets, as shown.

In addition, the learning unit 232 applies the data set 310 to each manufacturer's detailed model conditions, that is, detailed model conditions, such as company A's A model XG, company A's A model HG, and company A's model IG. It can be categorized into a corresponding data subset 316. In this case, since data subsets corresponding to each of the plurality of detailed model conditions are classified, the detailed model data subset 316 may include a plurality of data subsets, as illustrated.

Similarly, the learning unit 232 converts the data set 310 into a data subset 314 corresponding to each of the vehicle class conditions, such as company A's A model hybrid premium and company A's A model hybrid luxury. Can be classified. In this case, since data subsets corresponding to each of the plurality of vehicle class conditions are classified, the vehicle class data subset 314 may include a plurality of data subsets, as shown.

In addition, the learning unit 232 determines each of the data sets for a plurality of vehicles for which the transaction has been completed, and each class of vehicles according to the year, such as 2016 model A company A model hybrid premium, 2015 model A model hybrid premium, and 2015 B model gasoline luxury. It can be classified into a data subset 312 corresponding to the condition. At this time, since data subsets corresponding to each of the plurality of vehicle class conditions are classified, the vehicle class data subset 314 according to an age may include a plurality of data subsets, as illustrated.

In this case, each of the plurality of data subsets 312 to 322 may have a parameter greater than or equal to a predetermined value. For example, each classified data subset may have 100 or more data parameters.

For example, suppose that one of the data sets 310 is data for a company A's vehicle for sale. Then, this data set may be classified into all of the listing data subset 322, and may be classified into the data subset of Company A among the manufacturer data subset 320. However, since this data set does not contain information on vehicle class according to model, detailed model, vehicle class, and year, it may not be classified into any data subset among the data subsets 312, 314, 316, 318. have.

Each of the plurality of classified data subsets may include a training data subset and a test data subset. According to an embodiment, the learning unit 232 may sample a part of the data subset for all properties and classify it as a training data subset, and classify the remaining data subset that has not been sampled as a test data subset. have. By performing such sampling several times, a plurality of subsets of training data 334_a and a plurality of subsets of test data 334_b corresponding thereto may be generated. In addition, the learning unit 232 may sample a part of each of the plurality of data subsets for manufacturer conditions and classify them as a training data subset, and classify the remaining data subsets that are not sampled as a test data subset. By performing such sampling several times, a plurality of subsets of training data 332_a and a plurality of subsets of test data 332_b corresponding thereto may be generated. For example, a part of the data subset for Company A may be sampled several times and classified into a plurality of training data subsets and a plurality of test data subsets corresponding thereto.

In addition, the learning unit 232 may sample a part of each of the plurality of data subsets for the model condition and classify it as a training data subset, and classify the remaining data subsets that are not sampled as a test data subset. By performing such sampling several times, a plurality of subsets of training data 330_a and a plurality of subsets of test data 330_b corresponding thereto may be generated. For example, a part of the data subset for the model B of Company A may be sampled several times and classified into a plurality of training data subsets and a plurality of test data subsets corresponding thereto.

Similarly, the learning unit 232 may sample a part of each of the plurality of data subsets for detailed model conditions and classify them into a training data subset, and classify the remaining data subsets that are not sampled into a test data subset. have. By performing such sampling several times, a plurality of subsets of training data 328_a and a plurality of subsets of test data 328_b corresponding thereto may be generated. For example, a part of the data subset for Company A's B model HG may be sampled several times and classified into a plurality of training data subsets and a plurality of test data subsets corresponding thereto.

In addition, the learning unit 232 may sample a part of each of the plurality of data subsets for the vehicle class condition and classify them as a training data subset, and classify the remaining data subsets that are not sampled as a test data subset. By performing such sampling several times, a plurality of subsets of training data 326_a and a plurality of subsets of test data 326_b corresponding thereto may be generated. For example, a part of the data subset for the B model HG hybrid of Company A may be sampled several times and classified into a plurality of training data subsets and a plurality of test data subsets corresponding thereto.

In addition, the learning unit 232 samples a part of each of the plurality of data subsets for vehicle grade conditions according to the year and classifies them into a training data subset, and classifies the remaining data subsets that are not sampled into a test data subset. can do. By performing such sampling several times, a plurality of subsets of training data 324_a and a plurality of subsets of test data 324_b corresponding thereto may be generated. For example, a part of the data subset for the 2019 model A's B model HG hybrid may be sampled several times and classified into a plurality of training data subsets and a plurality of test data subsets corresponding thereto.

According to an embodiment, for each data subset, the training data subset may be classified as 70%, and the test data subset may be classified as 30%. Here, each subset of training data may be used to extract each hyperparameter of a plurality of prediction models corresponding thereto, and to configure each prediction model to give a minimum error. In addition, the prediction model generator 240 may generate a plurality of prediction models configured to infer the market price of a used car through a machine learning algorithm using each classified subset of training data. Here, the machine learning algorithm may include ensemble learning (eg, Bagging, Boosting, etc.).

4 is an exemplary view showing a process of generating a plurality of prediction models capable of inferring a market price of a used car according to an embodiment of the present disclosure. According to an embodiment, the prediction model generator 240 may generate an ensemble model configured to infer a market price of a used car using an ensemble learning algorithm. For example, the ensemble learning algorithm may include a boosting algorithm such as XGBoost (Extreme Gradient Boosting).

The prediction model generation unit 240 may generate a plurality of subsets of training data by sampling multiple times from a specific data subset (eg, one data subset among a plurality of data subsets corresponding to a specific condition). . For example, as shown, the prediction model generation unit 240 includes a first subset of training data 412, a second subset of training data 416, ..., nth from the data subset 410 A subset of training data 420 may be generated. In the present disclosure, n may be an integer of 3 or more.

The prediction model generation unit 240 may be configured to generate a prediction model from a subset of training data through a machine algorithm. Here, the machine learning algorithm may be predetermined or automatically selected, and may be one of an artificial neural network, deep learning, and a decision tree, but is not limited thereto. . For example, as shown, the prediction model generation unit 240 includes a first prediction model 414 and a second prediction model 418 from each of the first to nth training data subsets 412, 416, and 420. ), ... The n-th prediction model 422 may be generated.

When generating a plurality of prediction models, the prediction model generator 240 may adjust the weight of the next subset of training data based on the training result of the previous prediction model. For example, as shown, a weight applied to the second subset of training data 416 may be adjusted based on a result of learning through the first prediction model 414. Likewise, the weight of the third subset of training data may be adjusted based on a result of training the second prediction model 418. This process may be repeated until the n-th subset of training data and the n-th prediction model 422 are generated. That is, the weight of the next training data is adjusted from the training result of the previous classifier, and the first prediction model 414, the second prediction model 418, ... n-th prediction model 422 are sequentially performed through the adjusted weight. Can be created with Then, the prediction model generation unit 424 generates a final prediction model 430 based on the generated first prediction model 414, second prediction model 418, ... n-th prediction model 422 can do. Here, the final prediction model 430 may be generated using each of the first to n-th subsets of training data and adjusted weights.

5 is an exemplary view showing an example of calculating a market price for a target used car in response to a market price prediction request according to an embodiment of the present disclosure. According to an embodiment, the market price calculator 234 may extract one or more prediction models corresponding to the target used car based on the received market price prediction request for the target used car. Here, the market price prediction request for the target used car may include arbitrary information indicating or characterizing the target used car, for example, vehicle class, vehicle class according to year, vehicle detailed model (detail model), vehicle model It may include at least one of (model) and manufacturer, but is not limited thereto.

In this disclosure, similar to FIG. 3, from the data set 310, the vehicle grade data subset 312 according to the year, the vehicle grade data subset 314, the detailed model data subset 316, the model data subset 318, manufacturer data subset 320, and all listing data subset 322. Further, for each data subset 312 to 322, a plurality of training data subsets 324_a, 326_a, 328_a, 330_a, 332_a, 334_a and a plurality of test data subsets 324_b, 326_b, 328_b, 330_b corresponding thereto , 332_b, 334_b) may be generated.

According to an embodiment, the prediction model generation unit 240 may generate a plurality of prediction models to infer a market price of a used car for a plurality of data subsets 312 to 322 through a machine algorithm. Here, the machine algorithm may be operated using a decision tree through ensemble learning such as bagging and boosting. For example, the prediction model generation unit 240 may generate a plurality of first prediction model candidates 534 through ensemble learning using each of a plurality of data subsets corresponding to vehicle grade conditions according to an age. . In addition, a plurality of second prediction model candidates 536 may be generated through ensemble learning using each of a plurality of data subsets corresponding to the vehicle class condition. In addition, the prediction model generator 240 may generate a plurality of third prediction model candidates 538 through ensemble learning using each of a plurality of data subsets corresponding to detailed model conditions. Further, a plurality of fourth prediction model candidates 540 may be generated through ensemble learning using each of a plurality of data subsets corresponding to the model conditions. In addition, the prediction model generator 240 may generate a plurality of fifth prediction model candidates 542 through ensemble learning using each of a plurality of data subsets corresponding to the manufacturer's condition. In addition, a plurality of sixth prediction model candidates 544 may be generated through ensemble learning using a data subset corresponding to all property conditions.

The prediction model verification unit 242 may extract hyperparameters of each prediction model candidate by using a plurality of subsets of training data. According to an embodiment, the hyperparameter of each prediction model candidate is a plurality of prediction model candidates 534 corresponding to each of a plurality of data subsets 312 to 322 using a plurality of training data subsets 324_a to 334_a. Each of to 544) may be extracted to have a minimum error. In this case, the prediction model verification unit 242 may use a known algorithm for extracting known hyperparameters such as Grid Search and Random Search.

For example, the prediction model verification unit 242 uses the plurality of training data subsets 324_a to determine that each of the plurality of prediction model candidates 534 corresponding to the vehicle class data subset 312 according to the year is at least Hyperparameters can be extracted to have an error of. The prediction model verification unit 242 may apply a method similar to the hyperparameter extraction method applied to the vehicle class data subset 312 according to the year to each of the other data subsets 314 to 322.

Then, the prediction model verification unit 242 uses the plurality of prediction model candidates 534 to 544 corresponding to each of the plurality of data subsets 312 to 322 by using the corresponding test data subsets 324_b to 334_b. Can be verified. During verification, a prediction model having the smallest error among the plurality of prediction model candidates 534 to 544 may be selected for the corresponding data subset. For example, the prediction model verification unit 242 verifies a plurality of prediction model candidates 534 corresponding to the vehicle class data subset 312 according to the year using the plurality of test data subsets 324_b, The first prediction model 546 may be selected from among the plurality of prediction model candidates 534. Here, the first prediction model 546 may be a prediction model having the smallest error or a prediction model having the highest accuracy among the plurality of prediction model candidates 534. In addition, the prediction model verification unit 242 applies a method similar to the prediction model selection method applied to the vehicle grade data subset 312 according to the year to each of the other data subsets 314 to 322 to each data subset ( Second to sixth prediction models 548, 550, 552, 554, and 556 corresponding to each of 314 to 322 may be selected.

The market price calculation unit 234 extracts one or more prediction models corresponding to the target used car in response to the market price prediction request 558 for the target used car, and the market price 560 for the target used car based on the extracted one or more prediction models. ) Can be calculated. For example, the market price calculation unit 234 selects at least some of a plurality of conditions that the target used car satisfies, and uses a prediction model selected for each of a plurality of data subsets corresponding to the selected at least some conditions. You can calculate the market price for used cars. Here, each of the plurality of data subsets 312 to 322 may be stored in association with each of the plurality of conditions.

According to an embodiment, the price calculation unit 234 applies a weight (for example, A, B, C, D, E, F) to the used car price calculated from the extracted one or more prediction models, and through ensemble learning. You can calculate the final market price for the target used car. These weights may vary depending on the number of extracted prediction models. For example, as the number of extracted prediction models increases, a weight for a later-ranked prediction model (eg, a fifth prediction model, a sixth prediction model, etc.) may decrease. On the contrary, as the number of extracted prediction models decreases, a weight for a later-ranked prediction model may be relatively increased.

For example, suppose that a request for price prediction for a target used car includes a 2019 model A company A model HG hybrid, which is a vehicle class according to the year. In this case, the market price calculation unit 234 extracts the first prediction model generated and learned from the data subset classified as the 2019 model A company A model HG hybrid among the plurality of data subsets for the vehicle detailed model, and A second prediction model generated and trained from the data subset classified as the A model HG hybrid is extracted, and the third prediction model generated and learned from the data subset classified as the A model hybrid of Company A is extracted, and the A model is used. A fourth prediction model generated and learned from the classified data subset, a fifth prediction model generated and learned from the data subset classified as Company A, and a sixth prediction model generated and learned from the data subset for all properties are extracted. can do. At this time, the first prediction model, the second prediction model, the third prediction model, the fourth prediction model, the fifth prediction model, and the sixth prediction model have the most error selected based on the results evaluated from each of the test data subsets. It may be a prediction model that is small and has high accuracy. Accordingly, the market price calculation unit 234 includes a first prediction model, a second prediction model, a third prediction model, and a fourth prediction model extracted based on the 2019 model A company A model HG hybrid included in the market price prediction request for the target used car. Through each of the prediction model, the fifth prediction model, and the sixth prediction model, the market price for the A model HG hybrid of Company A in 2019 can be calculated. Then, the price calculation unit 234, for each calculated price, for example, the weight A of the first prediction model is 35%, the weight B of the second prediction model is 20%, and the weight C of the third prediction model. Is 15%, the weight D of the 4th prediction model is 10%, the weight E of the 5th prediction model is 10%, and the weight F of the 6th prediction model is 5%. The final price for the hybrid can be calculated, and the calculated price can be provided to a user terminal that has transmitted a price prediction request for a target used car.

As another example, suppose that a request for price prediction for a target used car includes a vehicle detailed model (detail model), a company A model hybrid. In this case, the market price calculation unit 234 extracts a third prediction model generated and trained from the data subset classified as A company's A model hybrid among the plurality of data subsets for the detailed model, and the data classified as the A model. A fourth prediction model generated and learned from the subset, a fifth prediction model generated and learned from the data subset classified as Company A, and a sixth prediction model generated and learned from the data subset for all properties may be extracted. . In this case, the third prediction model, the fourth prediction model, the fifth prediction model, and the sixth prediction model may be prediction models having the smallest error and high accuracy selected based on the evaluation result from each test data subset. . Accordingly, the market price calculation unit 234 includes a third prediction model, a fourth prediction model, a fifth prediction model, and a sixth prediction model extracted based on the A model hybrid of Company A included in the price prediction request for the target used car. Through each, it is possible to calculate the market price for Company A's A model hybrid. For each calculated price, the price calculation unit 234, for example, the weight C of the third prediction model is 45%, the weight D of the fourth prediction model is 30%, and the weight E of the fifth prediction model is 15%. , By applying the weight F of the sixth prediction model as 10%, the final price for the hybrid of the A company's A model can be calculated through ensemble learning, and the calculated price is transmitted to a price prediction request 558 for the target used car. It can be provided as a user terminal.

6 is a flowchart illustrating a method 600 for predicting a used car market price according to an embodiment of the present disclosure. In an embodiment, the used car price prediction method 600 may be performed by a used car price prediction system. As shown, the used car market price prediction method 600 may be initiated by receiving data sets for a plurality of vehicles for which the transaction has been completed (S610). Here, the data set for the vehicle for which the transaction has been completed may include data related to the price of a new car, a performance check of a used car, a type of fuel of the vehicle, a vehicle driving distance, a vehicle operation date, and whether an accident has occurred, but is not limited thereto. According to an embodiment, the used car market price prediction system may receive a data set for a vehicle in which a transaction has been completed among used cars from an external device every preset period.

According to an embodiment, the used car market price prediction system filters a data set for a received vehicle by at least one of a data set having the same sale, a data set having a vehicle price error, and a data set having a driving distance error. , Can be saved in the form of a database in the storage unit.

Then, according to a predetermined condition, data sets for a plurality of vehicles for which transactions have been completed may be classified into a plurality of data subsets (S620). Here, the predetermined condition may include a plurality of conditions. In one embodiment, when each of the data sets for the plurality of vehicles for which the transaction has been completed satisfies at least some of the plurality of conditions, the used car market price prediction system includes a plurality of data sets corresponding to each of the at least some conditions satisfied. Can be classified as a subset of the data.

For example, the plurality of conditions may include vehicle grade conditions according to year, vehicle grade conditions, detailed model conditions, model conditions, manufacturer conditions, and all sales conditions, and the learning unit may include data sets for a plurality of vehicles for which transactions have been completed. It can be categorized into one or more subsets of data according to the vehicle class by all sale, make, model, detailed model, vehicle class, and year. In this case, each of the plurality of classified data subsets may have a parameter equal to or greater than a predetermined value, and may be stored in association with each of the plurality of conditions.

Thereafter, with respect to the plurality of data subsets classified in step S630, a plurality of prediction models configured to infer the market price of the used car may be generated. According to an embodiment, the used car price prediction system may generate a prediction model configured to infer the price of a used car through a machine learning algorithm for each of a plurality of classified data subsets.

For example, the used car market price prediction system may generate a plurality of first prediction model candidates through ensemble learning using each of a plurality of data subsets corresponding to the vehicle class condition according to the year A plurality of second prediction model candidates can be generated through ensemble learning using each of the plurality of data subsets corresponding to the grade condition, and ensemble learning using each of the plurality of data subsets corresponding to the detailed model conditions Through this, a plurality of third prediction models may be generated.

In addition, the used car market price prediction system may generate a plurality of fourth prediction model candidates through ensemble learning using each of a plurality of data subsets corresponding to the model conditions, and the plurality of data subsets corresponding to the manufacturer conditions. Using each of them, a plurality of fifth prediction model candidates may be generated through ensemble learning, and a plurality of sixth prediction model candidates may be generated through ensemble learning using a data subset corresponding to all property conditions. Here, the ensemble learning may be learning using an XGBoost (Extreme Gradient Boosting) algorithm.

In one embodiment, each of the plurality of data subsets may include a plurality of training data subsets and test data subsets. The used car price prediction system extracts and extracts hyperparameters of each prediction model candidate so that each of a plurality of prediction model candidates corresponding to each of the plurality of data subsets has a minimum error using a plurality of subsets of training data. Each prediction model candidate can be verified using a subset of test data by applying each of the hyperparameters to each prediction model candidate.

The used car market price prediction system selects a prediction model with the smallest error based on each evaluation result of the prediction model candidate among a plurality of prediction model candidates corresponding to each of a plurality of data subsets as a prediction model of the classified data subset. You can choose each.

Next, the used car price prediction system may receive a price prediction request for the target used car, and extract one or more prediction models corresponding to the target used car from among the generated prediction models (S640). According to an embodiment, the used car price prediction system may receive a price prediction request for a target used car from a user terminal. Here, the price prediction request for the target used car may include at least one of a vehicle year, vehicle class, vehicle detailed model, vehicle model, and manufacturer, but is not limited thereto.

Finally, the used car market price prediction system may calculate a market price for a target used car based on one or more extracted prediction models in response to a request for price prediction for the target used car (S650). According to an embodiment, a market price for a target used car may be calculated through each of the extracted prediction models. The used car price prediction system selects at least some of a plurality of conditions that the target used car satisfies, and calculates the price of the target used car through a prediction model selected for each of a plurality of data subsets corresponding to the selected at least some conditions. Can be calculated. In this case, the used car price prediction system may calculate the final price for the target used car through ensemble learning by applying a weight to each of the calculated price for the used car.

A weight is applied to each calculated used car market price, the final price for a target used car is calculated through ensemble learning, and the calculated final price can be provided to a user terminal through a communication network. Here, the weight for each used car price may be applied according to each prediction model that calculated the used car price.

The above-described method and system for predicting used car market prices may be implemented as computer-readable codes on a computer-readable recording medium. The computer-readable recording medium includes all types of recording devices that store data that can be read by a computer system. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices. In addition, the computer-readable recording medium is distributed over a computer system connected through a network, so that computer-readable codes can be stored and executed in a distributed manner. In addition, functional programs, codes, and code segments for implementing the above-described embodiments can be easily inferred by programmers in the technical field to which the present invention belongs.

The method, operation, or techniques of this disclosure may be implemented by various means. For example, these techniques may be implemented in hardware, firmware, software, or a combination thereof. Those of skill in the art will appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the disclosure herein may be implemented in electronic hardware, computer software, or combinations of both. To clearly illustrate this interchange of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and design requirements imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementations should not be interpreted as causing a departure from the scope of the present disclosure.

In a hardware implementation, the processing units used to perform the techniques include one or more ASICs, DSPs, digital signal processing devices (DSPDs), programmable logic devices (PLDs). ), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, electronic devices, other electronic units designed to perform the functions described in this disclosure , Computer, or a combination thereof.

Accordingly, the various exemplary logic blocks, modules, and circuits described in connection with the present disclosure may include a general purpose processor, DSP, ASIC, FPGA or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or It may be implemented or performed in any combination of those designed to perform the functions described in. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, eg, a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in connection with the DSP core, or any other configuration.

In firmware and/or software implementation, the techniques include random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRAM), PROM ( on a computer-readable medium such as programmable read-only memory, erasable programmable read-only memory (EPROM), electrically erasable PROM (EEPROM), flash memory, compact disc (CD), magnetic or optical data storage device, etc. It can also be implemented as stored instructions. The instructions may be executable by one or more processors, and may cause the processor(s) to perform certain aspects of the functionality described in this disclosure.

When implemented in software, the techniques may be stored on a computer-readable medium as one or more instructions or code or transmitted through a computer-readable medium. Computer-readable media includes both computer storage media and communication media, including any medium that facilitates transfer of a computer program from one place to another. Storage media may be any available media that can be accessed by a computer. By way of non-limiting example, such computer-readable medium may contain RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or the desired program code in the form of instructions or data structures. It may include any other medium that may be used for transfer or storage to and accessible by a computer. Also, any connection is properly termed a computer-readable medium.

For example, if the software is transmitted from a website, server, or other remote source using wireless technologies such as coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or infrared, wireless, and microwave, coaxial cable , Fiber optic cable, twisted pair, digital subscriber line, or wireless technologies such as infrared, wireless, and microwave are included within the definition of the medium. Disks and discs as used herein include CDs, laser disks, optical disks, digital versatile discs (DVDs), floppy disks, and Blu-ray disks, where disks are usually magnetic It reproduces data optically, whereas discs reproduce data optically using a laser. Combinations of the above should also be included within the scope of computer-readable media.

The software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM, or any other known type of storage medium. An exemplary storage medium may be coupled to a processor such that the processor can read information from or write information to the storage medium. Alternatively, the storage medium may be integrated into the processor. The processor and storage medium may also reside within the ASIC. The ASIC may exist in the user terminal. Alternatively, the processor and storage medium may exist as separate components in the user terminal.

Although the above-described embodiments have been described as utilizing aspects of the currently disclosed subject matter in one or more standalone computer systems, the present disclosure is not limited thereto, and may be implemented in connection with any computing environment such as a network or a distributed computing environment. . Furthermore, aspects of the subject matter in this disclosure may be implemented in multiple processing chips or devices, and storage may be similarly affected across multiple devices. Such devices may include PCs, network servers, and portable devices.

In the present specification, the present disclosure has been described in connection with some embodiments, but various modifications and changes may be made without departing from the scope of the present disclosure as understood by those of ordinary skill in the art to which the present disclosure belongs. In addition, such modifications and changes should be considered to fall within the scope of the claims appended to this specification.

As described above, by actively utilizing the payment information of other users to process the reservation purchase request, the user can easily make a purchase without having to continuously invest time to purchase the product he wants at a desired price. . That is, it is possible to find the lowest price product that can be actually traded, not a false sale, using payment information of other users. In addition, it is difficult for the seller to predict the demand of the buyer and it is necessary to manage the inventory. By providing statistical information on the reservation purchase request registered by the users to the seller, the seller can easily grasp the demand and the demand price for the product and strategically It is also possible to use.

110_1: New car price DB
110_2: Performance check DB
120: communication network
130: used car price prediction system
210: communication department
220: storage unit
230: processor
232: Learning Department
234: price calculation unit
240: prediction model generation unit
242: prediction model verification unit

Claims (11)

In the used car market price prediction method through machine learning,
Receiving data sets for a plurality of vehicles for which transactions have been completed;
Classifying the data sets of the plurality of vehicles for which the transaction has been completed into a plurality of data subsets according to a predetermined condition;
Generating a plurality of prediction models configured to infer a market price of a used car for the classified plurality of data subsets, the plurality of prediction models including a first prediction model and a second prediction model;
Receiving a market price prediction request for a target used car, and extracting a plurality of prediction models corresponding to the target used car from among the plurality of generated prediction models; And
In response to a request for price prediction for the target used car, calculating a plurality of market prices for the target used car based on the extracted plurality of prediction models
Including,
For the plurality of classified data subsets, generating a plurality of predictive models configured to infer prices of used cars may include, for each of the plurality of classified data subsets, the price of used cars is calculated through a machine learning algorithm. Generating a plurality of predictive models configured to infer,
The step of calculating a plurality of market prices for the target used car,
Selecting at least some of the plurality of conditions that the target used car satisfies;
Calculating a plurality of market prices for the target used car through a plurality of prediction models selected for each of a plurality of data subsets corresponding to the selected at least some conditions; And
And calculating a final market price for the target used car through ensemble learning by applying a weight to each of the calculated market prices for the target used car.
delete The method of claim 1,
The plurality of conditions include vehicle grade conditions, vehicle grade conditions, detailed model conditions, model conditions, manufacturer conditions, and all sales conditions according to the year,
Each of the plurality of data subsets has a parameter greater than or equal to a predetermined value,
How to predict used car prices.
delete The method of claim 1,
For the plurality of classified data subsets, generating a plurality of predictive models configured to infer a market price of a used car,
Generating a plurality of first prediction model candidates through ensemble learning by using each of a plurality of data subsets corresponding to vehicle grade conditions according to a model year;
Generating a plurality of second prediction model candidates through ensemble learning using each of the plurality of data subsets corresponding to the vehicle class condition;
Generating a plurality of third prediction model candidates through ensemble learning using each of a plurality of data subsets corresponding to detailed model conditions;
Generating a plurality of fourth prediction model candidates through ensemble learning using each of the plurality of data subsets corresponding to the model conditions;
Generating a plurality of fifth prediction model candidates through ensemble learning using each of a plurality of data subsets corresponding to a manufacturer condition; And
Generating a plurality of sixth prediction model candidates through ensemble learning by using a data subset corresponding to all the property conditions
Including, used car market price prediction method.
The method of claim 1,
Each of the classified plurality of data subsets includes a plurality of training data subsets and test data subsets,
For each of the plurality of classified data subsets, generating a predictive model configured to infer a market price of a used car through a machine learning algorithm,
Extracting hyperparameters of each prediction model candidate using the plurality of training data subsets so that each of a plurality of prediction model candidates corresponding to each of the plurality of data subsets has a minimum error; And
Including the step of selecting a prediction model having the smallest error when verifying using the test data subset from among a plurality of prediction model candidates corresponding to each of the plurality of data subsets,
How to predict used car prices.
delete The method of claim 6,
The market price prediction request for the target used car includes a future prediction condition for the target used car,
The step of calculating the market price for the target used car,
Calculating a future market price for the target used car through a plurality of prediction models selected for each of a plurality of data subsets corresponding to the selected at least some conditions; And
Comprising the step of calculating a final future market price for the target used car through ensemble learning by applying a weight to each of the calculated future market price for the target used car,
How to predict used car prices.
The method of claim 1,
Receiving a data set for a plurality of vehicles for which the transaction has been completed,
Filtering at least one data set from among the data sets for the plurality of vehicles, a data set having the same property, a data set having a vehicle price error, and a data set having a driving distance error,
How to predict used car prices.
A computer program stored in a computer-readable recording medium to execute the used car market price prediction method according to any one of claims 1, 3, 5, 6, 8, and 9 on a computer.
In a used car price prediction system through machine learning,
Receiving data sets for a plurality of vehicles for which transactions have been completed, and classifying data sets for a plurality of vehicles for which transactions have been completed according to a predetermined condition into a plurality of data subsets, and for the classified plurality of data subsets, A learning unit configured to generate a plurality of prediction models configured to infer a market price, the plurality of prediction models including a first prediction model and a second prediction model; And
Receiving a market price prediction request for a target used car, extracting a plurality of prediction models corresponding to the target used car from among the generated prediction models, and in response to a market price prediction request for the target used car, the extracted plurality Includes a market price calculation unit configured to calculate a plurality of market prices for the target used car based on the prediction model of,
The learning unit,
For each of the classified plurality of data subsets, further configured to generate a predictive model configured to infer a market price of a used car through a machine learning algorithm,
The price calculation unit,
A plurality of market prices for the target used car are selected through a plurality of prediction models selected for each of a plurality of data subsets corresponding to the selected at least some conditions that are satisfied by the target used car, and And calculating the final price for the target used car through ensemble learning by applying a weight to each of the plurality of market prices for the target used car calculated,
Used car price prediction system.

Images