JP2024011980A - Price prediction system, price prediction method, and price prediction program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a price prediction system, a price prediction method, and a price prediction program for easily and accurately predicting a market price of a secondhand commercial material, such as a secondhand car.

SOLUTION: A price prediction system 1 includes: a class classification model 17 which has machine-learned a relationship between a commercial material mode and three price trend classes to which the commercial material belongs, the commercial material mode including a state and/or maintenance records of a secondhand commercial material, and the price trend classes indicating how the transaction price of the commercial material trends down from the first date of sale with elapsed years; price prediction models 18a-18c for each price transaction class, the models having machine-learned, for each price trend class, a relationship between the commercial material mode of the commercial material and the trend of the transaction price with the elapsed years; and a price prediction unit 15 which decides a price trend class of a target commercial material using the class classification model 17, on the basis of the commercial material mode of the target commercial material for which the price is to be predicted, and predicts a current or future transaction price of the target commercial material using the price prediction model of the decided price trend class.

SELECTED DRAWING: Figure 1

COPYRIGHT: (C)2024,JPO&INPIT

Description

The present invention relates to a price prediction system, a price prediction method, and a price prediction program for predicting the market price of a used product.

Patent Document 1 discloses a basic residual value rate calculation processing unit that calculates a standard residual value rate that is not affected by used car appraisal factors based on used car sales records and residual value rate correction data, and a basic residual value rate calculation processing unit that calculates a standard residual value rate that is not affected by appraisal factors of used cars, and a A residual value decay model generation processing unit that generates a residual value decay model function that expresses the transition of the standard evaluation rate for each used car as an exponential function over time, and substitutes the product characteristics (evaluation points) of the used car. A relational model function generation processing unit that models the relationship between the coefficients of the residual value decay model function and the product characteristics of each used car by learning A product value evaluation device is disclosed that includes a residual value rate prediction processing unit that calculates a residual value rate.

Japanese Patent Application Publication No. 2002-117275

In general, the market price of a used car can follow different price trends over time after the start of sales, depending on the evaluation scores of various assessment items for each used car. For this reason. When using machine learning to generate a model used to predict the market price of a used car, as in the above evaluation device, in order to improve the accuracy of the prediction, it is necessary to assess, for example, the car type, color, grade, equipment, maintenance history, etc. It is necessary to prepare a large number of price prediction models for each of a plurality of groups depending on the size of evaluation points of factors, and the process of model generation and price prediction may become complicated.

An object of the present invention is to easily and accurately predict the market price of used products such as used cars.

One aspect of the present invention is to determine the product characteristics, including the condition and/or maintenance history of the used product, and the decrease in transaction price with respect to the number of years that have passed since the first sale date of the product. A class classification model that machine learns the relationship between the price transition class to which the product belongs, which indicates ease of use, and the transaction for the product type and the elapsed years for each price trend class. The target product is determined using the class classification model based on the price prediction model for each price trend class, which is machine-learning the relationship between the price trend and the product aspect of the target product that is the target of price prediction. A price prediction unit that determines the price transition class of the material and predicts the current or future transaction price of the target product using the price prediction model of the determined price transition class. It is a system.
According to another aspect of the present invention, the price trend class is determined based on a statistical value of the price decrease rate, which is the rate of decrease in the transaction price with respect to the number of years that have passed since the first sale date of the product.
According to another aspect of the present invention, the product aspect includes information on the color, grade, damage, and repair of the product.
According to another aspect of the present invention, the commercial products are products belonging to the same product category.
According to another aspect of the present invention, the merchandise is a vehicle, and the merchandise aspects include information on the vehicle's color, grade, damage, parts replacement, and past vehicle inspection dates.
According to another aspect of the present invention, for a second-hand product, aspect information, which is information about the product aspect including the condition and/or maintenance history of the product, and information about the product aspect from the first sale date of the product, a collection unit that collects price trend information that is information about trends in transaction prices over the number of years that have passed; a classification unit that classifies the product into one of the transition classes; generates a class classification model that performs machine learning on the relationship between the product aspect of the product and the price trend class to which the product belongs; and a model generation unit that generates a price prediction model for each price trend class by machine learning the relationship between the product aspect of the product and the transaction price trend with respect to the elapsed years.
Another aspect of the present invention is a price prediction method performed by a computer of a price prediction system, which includes: a method for predicting a price of a second-hand product, including the condition and/or maintenance history of the product; a step of obtaining a class classification model by machine learning of the relationship between the price trend class to which the commodity belongs, which indicates the ease with which the transaction price decreases with respect to the number of years that have passed since the first sale date of the commodity; , obtaining a price prediction model for each of the price trend classes, which is obtained by machine learning the relationship between the product type and the transaction price trend with respect to the elapsed number of years for the product; Based on the product aspect of a certain target product, the price transition class of the target product is determined using the class classification model, and the price prediction model of the determined price trend class is used to determine the target product. A price prediction method includes the step of predicting the current or future transaction price of a commodity.
Another aspect of the present invention is to enable the computer of the price prediction system to determine the product aspects of used products, including the condition and/or maintenance history of the products, and the information from the first sale date of the products. A class classification model that machine learns the relationship between the price transition class to which the commodity belongs, which indicates the ease with which the transaction price decreases with respect to the number of years that have passed, and the commodity aspect of the commodity for each price transition class. , a price prediction model for each price trend class that is machine-learned for the relationship between the transaction price trend and the number of years that have passed, and the class classification model based on the product aspect of the target product that is the target of price prediction. a price prediction unit that uses the price transition class of the target product to determine the price trend class of the target product, and estimates the current or future transaction price of the target product using the price prediction model of the determined price trend class; It is a functional price prediction program.

According to the present invention, it is possible to easily and accurately predict the market price of used products such as used cars.

FIG. 1 is a diagram showing the configuration of a price prediction system according to an embodiment of the present invention. FIG. 2 is a diagram showing an example of product types and price trend information acquired by the price prediction system. FIG. 3 is a functional block diagram showing the operation of the price prediction unit. FIG. 4 is a flow diagram showing a price prediction method executed by the price prediction system.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a diagram showing the configuration of a price prediction system according to an embodiment of the present invention. The price prediction system 1 estimates or predicts (hereinafter simply referred to as "prediction") the current and/or future market price of a used car, which is an example of a used product (hereinafter also referred to as a used product). Here, the market price refers to the residual value that can be recognized for the used product in the market, and in this embodiment, it refers to the transaction price for buying and selling the used product in the market, as an example. The price prediction system 1 predicts the current or future average transaction price of a used car in the used car market as the market price of a used car, which is a used product. The transaction price as a prediction result can be a price value or a price value range.

The price prediction system 1 acquires transaction information regarding the purchase and sale of used cars from an information providing server 2 that provides information regarding the purchase and sale of used cars in the market via a communication network 3 such as the Internet. The transaction information may include, for example, information about the vehicle type of the used car that has been bought and sold, and information about the transaction price. Vehicle details include, for example, the used car's model, name, year, color, grade, equipment, mileage, date of first sale as a new car, number of years since the date of first sale, and maintenance history. (For example, information such as replaced parts, past vehicle inspection dates, etc.) may be included. The information regarding the vehicle aspect corresponds to the aspect information in the present disclosure, which is information about the aspect of the merchandise including the condition and/or maintenance history of the merchandise, which is a used product.

Further, the transaction price information included in the transaction information may be the actual purchase price (hereinafter referred to as the purchase price), or the sales price that a used car dealer has priced the used car. For example, transaction price information includes the sales price for each year before the used car was actually sold (for example, the sales price in December of each year), and the sales price at the time the used car was actually sold. Information on the purchase price and the purchase price may be included. The above transaction price information corresponds to price trend information in the present disclosure, which is information about the transaction price trend (hereinafter also simply referred to as price trend) with respect to the number of years that have passed since the first sale date of the used product. Here, the "first sale date" refers to the date on which the item, which is currently a used item, is first sold as a new item.

Although one information providing server 2 is shown in FIG. 1, there may be a plurality of information providing servers 2 from which the price prediction system 1 acquires the transaction information. Such one or more information providing servers 2 as a whole can provide the above transaction information to the price prediction system 1 as so-called big data.

The price prediction system 1 includes a processor 10, a memory 11, and a communication device 12. The communication device 12 is a wired communication device and/or a wireless communication device for the price prediction system 1 to communicate with other devices such as the information providing server 2 directly or indirectly via a communication network 3 such as the Internet. .
The memory 11 is composed of nonvolatile and volatile semiconductor memories. The memory 11 stores a class classification model 17 and price prediction models 18a, 18b, and 18c generated by a model generation unit 133, which will be described later.

The processor 10 is, for example, a computer including a processor such as a CPU. The processor 10 may have a configuration including a ROM in which a program is written, a RAM for temporarily storing data, and the like. The processor 10 includes a model acquisition section 13, a target information acquisition section 14, and a price prediction section 15 as functional elements or functional units. The model acquisition section 13 includes a collection section 131, a classification section 132, and a model generation section 133, which are functional elements or functional units.

These functional elements of the processor 10 are realized, for example, when the processor 10, which is a computer, executes the price prediction program 16 stored in the memory 11. In addition to realizing the above functional elements, the price prediction program 16 reads the class classification model 17 and the price prediction models 18a, 18b, and 18c from the memory 11, and causes the processor 10, which is a computer, to function as these models.

Note that the price prediction program 16 can be stored in any computer-readable storage medium. Alternatively, all or part of the functional elements included in the processor 10 may be configured by hardware each including one or more electronic circuit components.

The model acquisition unit 13 uses the collection unit 131, the classification unit 132, and the model generation unit 133 to generate and acquire the class classification model 17 and the price prediction models 18a, 18b, and 18c.

The collection unit 131 collects, for each second-hand item, condition information, which is information about the item condition and/or maintenance history, and price trend information, which is information about price trends. Collect.
Specifically, the collecting unit 131 collects information on the vehicle mode of each used car as the above-mentioned mode information for a plurality of used cars that are used commercial products. The collection unit 131 also collects price trend information of each used car, including the yearly sales price and purchase price, as price trend information of the used car.

FIG. 2 is a diagram showing an example of the contents of used car transaction information collected by the collection unit 131. In the illustrated example, the vehicle type (merchandise type) includes the used car's model, name, year, color, grade, equipment, mileage, first sale date, and elapsed time since the first sale date. Including years and maintenance history. The equipment includes, for example, optional equipment for the vehicle with that car name. The maintenance history includes information on vehicle inspection dates, replaced parts, and replacement dates. In addition, the vehicle condition may include information such as the presence or absence of scratches, repair of scratches, and accident history.

Transaction price information also includes the purchase price on May 12, 2010, which is the first sale date, and the price at a used car dealership after the purchaser traded in the vehicle in 2015. It includes the store price for each year and the sales price when the vehicle was purchased by another buyer in 2019.
In addition, in FIG. 2, it should be understood that the characters in parentheses indicate the specific vehicle name, equipment name, date, part name, price, etc. of the corresponding item.

Based on the price trend information for each used product collected by the collection unit 131, the classification unit 132 classifies the used products using a price trend index that indicates how easily the price decreases with respect to the number of years that have passed since the first sale date. Classify into multiple price trend classes based on. Here, the price trend index is, for example, a statistical value of the rate of decrease in transaction price (hereinafter also referred to as price decrease rate) of the used product over the number of years that have passed since the first sale date. Specifically, the price trend index may be, for example, the average value, maximum value, minimum value, or median value of the annual price decrease rate from the first sale date to the present.

Classification into the above price trend classes is performed using regression analysis, class classification model 17, decision tree (DT), random forest (RF), Gradient Boosting Decision Tree (GBDT), XGBoost, and LightGBM in accordance with conventional techniques. It is possible to use a known classification method such as .
In the present embodiment, the classification unit 132 classifies used merchandise into three price trend classes in which, for example, the degree of ease of price decline with respect to the number of years elapsed is "large,""medium," and "small."

The model generation unit 133 calculates, for second-hand goods, the condition of the goods and/or the condition of the goods, including the maintenance history, and the ease with which the transaction price will decrease with respect to the number of years that have passed since the first sale date of the goods. A class classification model is generated by machine learning the relationship between the product and the price trend class to which the product belongs. In addition, the model generation unit 133 performs machine learning on the relationship between the product type and the price trend, which is the transaction price trend over the elapsed number of years, for the second-hand product for each of the price trend classes. Generate a price prediction model for each price trend class.

The class classification model 17 and the price prediction model 18 are, for example, machine learning models using neural networks. The price prediction model 18 is, for example, an RNN (recurrent model) including LSTM (long short term memory) or GRU (gated regression unit), which can effectively handle price trend data that is time series data. - neural network), and/or one-dimensional CNN (convolutional neural network).

Specifically, the model generation unit 133 uses the aspect information about each used car collected by the collection unit 131 and the price trend class of each used car classified by the classification unit 132 as training data to generate the aspect information. A class classification model 17 is generated by machine learning the relationship between the vehicle aspect (that is, the merchandise aspect) indicated by the information and the price trend class. The model generation unit 133 stores the generated class classification model 17 in the memory 11.

The model generation unit 133 also uses the condition information and price trend information about each used car collected by the collection unit 131 as training data to determine the vehicle condition and first sale date of the used car for each price trend class. Price prediction models 18a, 18b, and 18c are generated for each of the three price trend classes by machine learning the relationship between the price trend and the number of years that have passed since then. Hereinafter, the price prediction models 18a, 18b, and 18c will also be collectively referred to as the price prediction model 18.

For example, the model generation unit 133 uses the generated class classification model 17 based on the vehicle aspects of the plurality of used cars collected by the collection unit 131 to classify the used cars into price trend classes. Then, the model generation unit 133 performs machine learning on the price prediction model 18 to determine the relationship between each vehicle aspect of the used cars and the price change information collected by the collection unit 131 regarding those used cars for each price change class. let For example, the price prediction model 18 may calculate, as a prediction result, one value with the highest likelihood for current and/or future (for example, every year) transaction prices, and/or a likelihood of a predetermined value or more. A range of values with , can be output.
The model generation unit 133 stores the generated price prediction model 18 for each price transition class in the memory 11.

The target information acquisition unit 14 acquires aspect information regarding used merchandise (hereinafter referred to as target merchandise) that is the target of price prediction. For example, a person who wishes to purchase or sell a used product uses a mobile terminal such as a smartphone or a terminal device such as a personal computer to transmit information on the condition of the used product that he or she wishes to purchase or sell over the communication network 3, etc. The information may be transmitted to the price prediction system 1 via the network. The target information acquisition unit 14 receives and acquires the aspect information of those used products through the communication device 12 .
In the present embodiment, the target information acquisition unit 14 acquires information on the vehicle aspect of a used vehicle (hereinafter referred to as the target vehicle) that is the target of price prediction as the aspect information of the target vehicle.

The price prediction unit 15 uses the price prediction model 18 generated by the model generation unit 133 based on the aspect information of the target product acquired by the target information acquisition unit 14 to predict current and/or future transactions for the target product. Predict prices. Specifically, the price prediction unit 15 first determines the price trend class of the target product using the class classification model 17 based on the product aspect of the target product. Then, the price prediction unit 15 uses the price prediction model 18 of the determined price trend class to predict the current or future transaction price of the target product.

In the present embodiment, as an example, the price prediction unit 15 uses the price prediction model 18 generated by the model generation unit 133 to calculate The class classification model 17 is used to determine the price trend class of the target vehicle, and the price prediction model 18 of the determined price trend class is used to estimate the current or future transaction price of the target vehicle.

FIG. 3 shows the operation of the price prediction unit 15 as a functional block diagram. In FIG. 3, a function of selecting the price prediction model 18 according to the price trend class in the price prediction model 18 is shown as a model selection section 151. As described above, the price prediction unit 15 determines the price trend class of the target product using the class classification model 17 based on the aspect information of the target product acquired by the target information acquisition unit 14. Then, the model selection unit 151 selects a price prediction model 18 corresponding to the determined price transition class, and inputs the aspect information of the target product into the selected price prediction model 18. Then, the price prediction unit 15 outputs the predicted value of the transaction price output by the selected price prediction model 18.

The price prediction system 1 having the above configuration predicts the current or future transaction price (market price) of used merchandise using the price prediction model 18 for each price transition class of used merchandise. For this reason, price prediction system 1 reduces the number of models used, compared to a configuration that uses a large number of models depending on the various product aspects of used products, to easily and accurately market used products. Prices can be predicted.

Next, the price prediction method executed by the processor 10, which is the computer of the price prediction system 1, will be described with reference to the flowchart shown in FIG.

First, the price prediction system 1 uses a class classification model 17 that performs machine learning on the relationship between the product type and the price trend class to which the product belongs for each used product (used product). Acquire (S100). Further, the price prediction system 1 acquires a price prediction model 18 for each price transition class, which is obtained by machine learning the relationship between the product type and price transition for used merchandise for each price transition class (S102).

As described above, the price prediction system 1 uses the collection unit 131, the classification unit 132, and the model generation unit 133 to generate and acquire the class classification model 17 and the price prediction model 18 for each price transition class.

Next, the price prediction system 1 uses the target information acquisition unit 14 to acquire aspect information regarding the used product (target product) that is the target of price prediction (S104). Then, the price prediction system 1 uses the class classification model 17 to determine the price trend class of the target product based on the aspect information of the target product acquired by the target information acquisition unit 14 (S106). Then, the price prediction unit 15 uses the price prediction model 18 of the determined price trend class to predict the current or future price of the target product (S108), and ends the process.

After predicting the current or future transaction price of one target product in step S108, the price prediction system 1 returns to step S104 and repeats the process to predict the transaction price of other target products. I can do things.

[Other embodiments]
In the embodiment described above, the price prediction system 1 predicts the transaction price of a used car as a used product (used product). However, the target for which the price prediction system 1 predicts the transaction price is not limited to used cars, but may be any commercial product for which a used product can be bought or sold.

Such merchandise may be, for example, electrical appliances such as air conditioners and refrigerators, so-called brand products such as furniture and bags, or buildings such as houses.

In the embodiment described above, the price trend index is a statistical value such as the average value, maximum value, minimum value, or median value of the annual price decrease rate, but is not limited to this. The price trend index can be any value that indicates how easily the price decreases over the number of years that have passed. For example, the price trend index may be a coefficient of a specific term (for example, the term with the highest degree) of a predetermined formula (for example, a polynomial) when the price trend with respect to the number of years elapsed is approximated by the formula (for example, a polynomial).

Although the class classification model 17 and the price prediction model 18 for each price trend class are generated by the model acquisition unit 13 in the above-described embodiment, they may be generated by a similar method in another device. . In this case, the model acquisition unit 13 simply uses the communication device 12 to receive and acquire the class classification model 17 and the price prediction model 18 for each price trend class from the other device. be able to.

Further, in the above-described embodiment, all functional elements of the price prediction system 1 are realized in one device, but the structure is not limited to this. For example, the price prediction system 1 may be configured such that all or part of the functional elements described above are realized in a distributed manner in a plurality of devices, and these devices are communicably connected via the communication network 3.

Note that the present invention is not limited to the configuration of the above-described embodiments, and can be implemented in various forms without departing from the gist thereof.
For example, the number of price trend classes classified by the class classification model 17 is three in the above-described embodiment, but it may be any number greater than or equal to two. In addition, the number of price prediction models 18 for each price trend class acquired by the model acquisition unit 13 is three in the above-described embodiment, but it is the same as the number of price trend classes classified by the class classification model 17. It can be any number.

[Configurations supported by the above embodiment]
The embodiments described above support the following configurations.

(Structure 1) For second-hand products, it shows the condition of the product, including its condition and/or maintenance history, and the ease with which the transaction price will decrease with respect to the number of years that have passed since the first sale date of the product. , the price transition class to which the commodity belongs, and a class classification model that machine-learns the relationship between the commodity and the price transition class to which the commodity belongs; The price of the target product is calculated using the class classification model based on the price prediction model for each price trend class, which is machine-learning the relationship between , and the product aspect of the target product that is the target of price prediction. A price prediction system comprising: a price prediction unit that determines a transition class and predicts a current or future transaction price of the target product using the price prediction model of the determined price transition class.
According to the price prediction system of configuration 1, the current or future transaction price of used products is predicted using a price prediction model for each price trend class of used products, so that it is possible to predict the current or future transaction price of used products. Compared to a configuration that uses a large number of models depending on the situation, the number of models used can be reduced and the market price of used products can be easily and accurately predicted.

(Configuration 2) The price prediction system according to Configuration 1, wherein the price trend class is determined based on a statistical value of the price decrease rate, which is the rate of decrease in the transaction price with respect to the number of years that have passed since the first sale date of the product. .
According to the price prediction system of configuration 2, price trend classes can be determined objectively.

(Structure 3) The price prediction system according to Structure 1 or 2, wherein the product aspect includes information on color, grade, damage, and repair of the product.
According to the price prediction system of configuration 3, it is possible to predict a transaction price that appropriately reflects the condition of used merchandise.

(Configuration 4) The price prediction system according to Configuration 1, wherein the commercial materials are products belonging to the same product category.
According to the price prediction system of Configuration 4, it is possible to accurately predict the market price of used products in the same category by eliminating the influence of price trend factors that depend on the characteristics of the product category.

(Configuration 5) The price according to configuration 1 or 2, wherein the commercial product is a vehicle, and the product aspect includes information on the color, grade, damage, parts replacement, and past vehicle inspection date of the vehicle. Prediction system.
According to the price prediction system of configuration 5, the market price of a used vehicle can be predicted easily and accurately.

(Configuration 6) For second-hand goods, aspect information that is information about the condition of the goods and/or aspects of the goods including maintenance history, and transaction price for the number of years that have passed since the first sale date of the goods. a collection unit that collects price trend information that is information about the trend of the product; and a collection unit that collects price trend information that is information about the trend of the product; generates a class classification model that performs machine learning on the relationship between the product aspect of the product, and the price trend class to which the product belongs; Any one of configurations 1 to 5, comprising a model generation unit that generates a price prediction model for each of the price trend classes based on machine learning of the relationship between the product type of the material and the transaction price transition with respect to the elapsed number of years. Price prediction system described in Kani.
According to the price prediction system of configuration 6, it is possible to generate an appropriate class classification model and a price prediction model for each price transition class, and perform highly accurate market price prediction for used merchandise.

(Configuration 7) A price prediction method performed by a computer of a price prediction system, which includes the condition and/or maintenance history of the used product, including the product condition and/or maintenance history, and the first sale of the product. acquiring a class classification model that is machine-learned to determine the relationship between the commodity and the price transition class to which the commodity belongs, which indicates the ease with which the transaction price decreases with respect to the number of years that have elapsed since obtaining a price prediction model for each of the price trend classes, which is obtained by machine learning the relationship between the product type and the transaction price trend with respect to the elapsed years; and the target product that is the target of price prediction. Based on the product aspect, the price trend class of the target product is determined using the class classification model, and the current price trend class of the target product is determined using the price prediction model of the determined price trend class. or a step of predicting future transaction prices.
According to the price prediction method of configuration 7, the market price of used items can be easily and accurately calculated using a smaller number of models than in a configuration that uses a number of models depending on the various types of used items. Can be predicted.

(Configuration 8) The computer of the price prediction system is used for transactions regarding second-hand goods, including the condition and/or maintenance history of the goods, and the number of years that have passed since the first sale date of the goods. A class classification model that machine learns the relationship between the price trend class to which the product belongs, which indicates how easily the price will fall, and the product type and the number of years that have passed for the product for each price trend class. A price prediction model for each price trend class is machine-learned to determine the relationship between the transaction price trend and the transaction price. A price prediction unit that determines the price trend class of a commercial product and uses the price prediction model of the determined price trend class to estimate the current or future transaction price of the target product. program.
According to the price prediction program of Configuration 8, it is possible to realize a price prediction system that can easily and accurately predict the market price of used merchandise using a small number of models.

DESCRIPTION OF SYMBOLS 1...Price prediction system, 2...Information provision server, 3...Communication network, 10...Processor, 11...Memory, 12...Communication device, 13...Model acquisition section, 131...Collection section, 132...Classification section, 133...Model generation Part, 14...Target information acquisition unit, 15...Price prediction unit, 151...Model selection unit, 16...Price prediction program, 17...Class classification model, 18, 18a, 18b, 18c...Price prediction model.

Claims (8)

For second-hand products, the product characteristics including the condition and/or maintenance history of the product, and the tendency for the transaction price to decrease with respect to the number of years that have passed since the first sale date of the product. A class classification model that machine learns the relationship between the price trend class to which it belongs, and
a price prediction model for each of the price trend classes, which performs machine learning on the relationship between the product type and the transaction price trend with respect to the elapsed number of years for the product, for each of the price trend classes;
Determine the price trend class of the target product using the class classification model based on the product aspect of the target product that is the target of price prediction, and use the price prediction model for the determined price trend class. a price prediction unit that uses the technology to predict the current or future transaction price of the target product;
Equipped with
Price prediction system. The price trend class is determined based on the statistical value of the price decrease rate, which is the rate of decrease in the transaction price with respect to the number of years that have passed since the first sale date of the product.
The price prediction system according to claim 1. The product aspect includes information on the color, grade, damage, and repair of the product.
The price prediction system according to claim 1. The products belong to the same product category;
The price prediction system according to claim 1. The commercial product is a vehicle,
The product aspect includes information on the vehicle's color, grade, damage, parts replacement, and past vehicle inspection date.
The price prediction system according to claim 1. For second-hand products, condition information, which is information about the condition and/or maintenance history of the product, and trends in transaction prices over the number of years that have passed since the first sale date of the product. a collection unit that collects price trend information, which is information;
a classification unit that classifies the commercial product into one of a plurality of price trend classes indicating how likely the price is to decrease over the number of years that have passed, based on the price trend information;
A class classification model is generated by machine learning of the relationship between the product aspect of the product and the price transition class to which the product belongs, and the product aspect of the product is determined for each price transition class. a model generation unit that generates a price prediction model for each of the price trend classes by machine learning the relationship between and the transaction price trend with respect to the elapsed number of years;
Equipped with
The price prediction system according to any one of claims 1 to 5. A price prediction method performed by a computer of a price prediction system,
For second-hand products, the product characteristics including the condition and/or maintenance history of the product, and the tendency for the transaction price to decrease with respect to the number of years that have passed since the first sale date of the product. a step of obtaining a class classification model that is machine-learned for the relationship between the price trend class to which the price transition class belongs;
obtaining a price prediction model for each of the price trend classes, which is obtained by machine learning the relationship between the product type and the transaction price trend with respect to the elapsed number of years for the product; ,
Determine the price trend class of the target product using the class classification model based on the product aspect of the target product that is the target of price prediction, and use the price prediction model for the determined price trend class. predicting the current or future transaction price of the target product using the method;
has,
Price prediction method. price prediction system computer,
For second-hand products, the product characteristics including the condition and/or maintenance history of the product, and the tendency for the transaction price to decrease with respect to the number of years that have passed since the first sale date of the product. A class classification model that machine learns the relationship between the price trend class to which it belongs, and
a price prediction model for each of the price transition classes, which performs machine learning on the relationship between the product type and the transaction price trend with respect to the elapsed number of years for the product for each of the price trend classes;
Determine the price trend class of the target product using the class classification model based on the product aspect of the target product that is the target of price prediction, and use the price prediction model for the determined price trend class. a price prediction unit that estimates the current or future transaction price of the target product using the
A price prediction program that works as a.

Images