JP5380203B2 - Price estimation apparatus, method and program - Google Patents

Description

  The present invention relates to a technique for estimating an appropriate price of a used product traded in an auction or a used market.

  Currently, there are increasing opportunities for individuals to buy and sell goods directly through online auction transactions using the Internet. An exhibition user who sells a product accesses an auction site on the Internet using a terminal device such as a personal computer (hereinafter referred to as a PC), posts information on the target product on this auction site, and this product Wait for a successful bid. At this time, the exhibitor can set a successful bid condition for the target product. For example, the exhibitor sets the seller desired price as a successful bid condition. When a purchaser wishes to make a successful bid for a product for which such a successful bid condition is set, the purchaser cannot make a successful bid unless a price higher than the seller's desired price is designated.

  By using such a successful bid condition, the exhibitor can prevent the successful bid for his / her item with an unsatisfactory amount. On the other hand, if the seller's desired price is set too high, bids may decrease and the product may not be sold. Therefore, it is important for the exhibitor to appropriately determine the price when setting the seller desired price.

JP 2002-329097 A

  However, conventionally, the seller wants to set the seller's desired price as the highest price within the range that can be sold, and sets the seller's desired price with a sense, such as referring to the price set by other sellers. There was only. Such a situation was the same not only at the time of auction transaction but also at the time of transaction in the used market. Hereinafter, the highest price in the range that can be sold for second-hand goods traded in auctions and second-hand markets is referred to as an appropriate price.

  By the way, such a reasonable price fluctuates depending on the market and the trend of the world (needs, topics, etc.), so it cannot be accurately determined only by the list price or the past transaction amount. In particular, it is difficult to determine an appropriate price for second-hand goods that have not been traded in the past because there is no information on the transaction amount.

  In view of such problems, an object of the present invention is to provide a technique for estimating an appropriate price of a used product.

  In the aspect of the present invention, the following configurations are employed in order to solve the above-described problems.

This aspect relates to a price estimation device. The price estimation device according to the present aspect includes product information including list price information regarding the target product, transaction information based on transaction results regarding the target product, and information reflecting at least one of needs and topicality regarding the target product In the estimation model showing the relationship between the acquisition means for acquiring the trend information and the product information, the transaction information and the peripheral trend information as described above, and the product price, the product information acquired by the acquisition means, the transaction performance information, and Estimation means for estimating the price of the target product as an objective variable by using peripheral trend information as an explanatory variable.

  In this aspect, with respect to the product subject to price estimation, not only transaction information based on transaction results but also product information including price information, and peripheral trend information including information reflecting at least one of needs and topicality are estimated. Estimate the price of the target product by using it as an explanatory variable for the model.

  Therefore, according to this aspect, in addition to the transaction information based on the transaction results, the list price is considered, and furthermore, it is possible to estimate the price following the price fluctuation affected by at least one of needs and topicality. A reasonable price can be estimated with high accuracy.

  The price estimation device according to this aspect is preferably configured so that each time a transaction price of a product is determined, each piece of information used as an explanatory variable of the estimation model acquired by the acquisition unit regarding the product, and the transaction The history storage unit that stores the transaction price, which is the objective variable of the estimation model, and the value of each estimation parameter of the estimation model by multiple regression analysis using each information stored for each transaction stored in the history storage unit Updating means for updating.

  Thereby, according to this aspect, every time the transaction price is determined, the estimation parameter is updated in a form including the information at the time of the transaction, so the accuracy of the estimation model can be improved, The price can be estimated with high accuracy.

  In the price estimation device according to this aspect, preferably, the updating unit weights the new information in the history among the pieces of information stored in the history storage unit when updating the value of each estimation parameter. A least square method using coefficients is used.

  In the price estimation apparatus according to this aspect, preferably, the product information includes a list price and a state of the product, and the transaction information includes an average transaction price based on a transaction record of the same product as the target product. The peripheral trend information is configured to include the number of searches to the content server regarding the target product and the number of publications on the content server.

  Note that another aspect of the present invention may be a method for realizing any of the above configurations, a program, or a computer-readable storage medium that records such a program. There may be.

  ADVANTAGE OF THE INVENTION According to this invention, the technique which estimates the appropriate price of a used article can be provided.

The figure which shows the whole structure of an auction system and its related system. 1 is a conceptual diagram illustrating a configuration of an auction server 9 according to Embodiment 1. FIG. The figure which shows the example of an exhibition goods information table. The figure which shows the example of a transaction information table. 1 is a conceptual diagram illustrating a configuration of a price estimation server 10 in Embodiment 1. FIG. The figure which shows the example of the parameter table 57. FIG. The figure which shows the example of the explanatory variable information table. 5 is a flowchart showing price estimation processing of the price estimation server in the first embodiment. 5 is a flowchart illustrating an estimation model generation process of the price estimation server according to the first embodiment.

  Hereinafter, a specific example of a price estimation server device (hereinafter simply referred to as a price estimation server) as an embodiment for carrying out the present invention will be described. The following example is an example in which the price estimation server according to the present embodiment is applied to an online auction system (hereinafter simply referred to as an auction system). However, the price estimation server is not intended only for an online auction system, and may be applied to a general auction that is not online. Further, the price estimation server may target not only products that are traded in an auction form but also products that are traded in a general second-hand market. Each of the following examples is illustrative, and the present invention is not limited to the configurations of the following examples.

  Hereinafter, Example 1 of the price estimation server as an embodiment will be described. In addition, in order to make explanation easy to understand, in the first embodiment, a used book is given as an example of a product traded in an online auction. However, the present invention does not limit the types of such target products.

〔System configuration〕
First, an overall configuration of an auction system to which the price estimation server according to the first embodiment is applied and a system related thereto will be described with reference to FIG. FIG. 1 is a diagram showing the overall configuration of an auction system and a system related thereto.

  The content server 3 is a general server group that provides various information contents. The content server 3 is realized as, for example, a WEB server or a file server. The content server 3 according to the first embodiment includes a book information server that provides information about each book. In the content server 3, as information about each book, a book name, an author name, a list price, a publisher name, a publication year, a series name, and the like are provided. This content server 3 manages each book by ISBN (International Standard Book Number). In addition, the content server 3 includes a news server that provides news information, a blog server, a WEB server launched by various corporations and individuals, and the like.

  The search server 4 has a search engine for searching for information content provided on the network 1 by the content server 3. The search server 4 provides the address information of the content server 3 that provides information content including the keyword by causing the search engine to search based on the keyword input by the user. This address information is, for example, a URL (Uniform Resource Locator).

  The search server 4 holds a search history by the user. The search server 4 calculates a search burst degree for each search keyword based on the search history, and also holds the calculated search burst degree. This search burst degree means the gradient of the number of searches from the present to a predetermined period (N days ago). For example, 4 days is used for the predetermined period (N). The search server 4 provides information related to the search history in response to a request from another device.

The exhibitor terminal 6 and the purchaser terminal 7 are user terminals such as PCs and mobile phones, for example. For convenience of explanation, a terminal used by a user who sells a product at an auction is referred to as an exhibitor terminal 6, and a terminal used by a user who makes a successful bid (purchase) a product at an auction is referred to as a purchaser terminal 7. . The exhibitor terminal 6 and the purchaser terminal 7 have a general communication function capable of accessing the auction system 8 and receiving data, a general user interface function capable of displaying and operating a screen based on the provided data, and the like. The present invention does not limit the hardware configuration and functional configuration of the exhibitor terminal 6 and the purchaser terminal 7.

The auction system 8 is connected to the content server 3, the search server 4, the exhibitor terminal 6, the purchaser terminal 7, etc. via the network 1. The auction system 8 provides an online auction site to the exhibitor terminal 6 and the purchaser terminal 7 that are accessed via the network 1. The auction system 8 includes an auction server 9 and a price estimation server 10 that are connected to each other via an internal network (not shown). The network 1 and the internal network of the auction system 8 may be a public network such as the Internet, or an in-house network such as a LAN (Local Area Network) or a WAN (Wide Area Network). The present invention does not limit such a network, nor does it limit the protocol that flows over the network.

〔Device configuration〕
Hereinafter, each apparatus structure of the auction server 9 and the price estimation server 10 which comprise the auction system 8 in Example 1 is each demonstrated.

<Auction server>
FIG. 2 is a conceptual diagram illustrating a configuration of the auction server 9 according to the first embodiment.

As shown in FIG. 2, the auction server 9 includes a control unit 20, a storage unit 21, a communication unit 28, and the like connected by a bus 25 as a hardware configuration. The storage unit 21 is, for example, a hard disk, and stores various types of information used in processing executed by the control unit 20. The control unit 20 includes one or a plurality of processors such as a CPU (Central Processing Unit) and peripheral circuits (ROM (Read Only Memory), RAM (Radom Access Memory), interface circuit, etc.) used for processing of the processor. . The communication unit 28 is connected to the network 1 and an internal network (not shown) of the auction system 8, and communicates between the auction server 9 and the exhibitor terminal 6 and the purchaser terminal 7, the auction server 9 and the price estimation server 10. To communicate with each other. The present invention does not limit the communication mode of these auction servers 9, and for example, IP (Internet Protocol) communication is used. The auction server 9 may be constructed by a general-purpose computer such as a PC or a dedicated computer. The present invention does not limit the hardware configuration of the auction server 9.

  The auction server 9 implements an online auction function by the application program stored in the storage unit 21 being executed by the control unit 20. Hereinafter, only processing related to the present invention in the auction server 9 will be described in detail, and a general online auction function will be omitted or simply described.

  The auction server 9 manages users who can be sellers or buyers, and accepts listings for access from the seller terminals 6 and the buyer terminals 7 used by authorized users to be managed, bids Provide a screen that accepts For example, a WEB system is used as an interface between the auction server 9 and the exhibitor terminal 6 and the purchaser terminal 7.

The auction server 9 manages information about each product for which the listing has been received in the listing product information table 22. FIG. 3 is a diagram illustrating an example of the exhibition product information table. As shown in FIG. 3, in the first embodiment, since used books are targeted, the exhibition product information table includes a book ID, a book name, an author name, a publisher name, a publication year, a series name, and a state for each book. , Price, ISBN, etc. are stored. The book ID is an ID for specifying each exhibition book as an object, and different IDs are assigned to the same book if they are different as an object. The book ID is given by the auction server 9 at the time of accepting an exhibition. The ISBN is a number assigned to the same book and is managed by the book information server as described above.

  When the auction server 9 accepts the exhibition from the exhibitor terminal 6, the auction server 9 acquires each of the above-described information about the accepted book. For example, the auction server 9 is a content server in which the book name, author name, publisher name, publication year, series name, and list price for the book are used as the book information server based on the ISBN attached to the accepted book. Get from 3. In addition, the auction server 9 allows the user to input information regarding the state of the book via the WEB screen. The auction server 9 may allow the user to input book information other than the state. As the state of the book, for example, the score (1 to 5 etc.) indicating the state of good, scratched, bad, or inferior like the new article is stored in the exhibition product information table 22.

  In addition, the auction server 9 includes a general successful bid process, a successful bid notification process to the exhibitor terminal 6 and the purchaser terminal 7, and the like. In this successful bid process, the auction server 9 enables the seller to set a successful bid condition for the exhibited product. This successful bid condition includes the seller's desired price. The auction server 9 does not determine a successful bid for an exhibition product for which the seller desired price is set as the successful bid condition unless a bid is made at a price higher than the seller desired price. This successful bid condition is also managed for each item for sale. Hereinafter, when a successful bid is determined in the auction server 9, the auction transaction is assumed to have been established.

  The auction system 8 according to the first embodiment provides the seller with a recommended amount of the seller's desired price as a successful bid condition. The price estimated by the price estimation server 10 is used as the recommended amount of the seller desired price. Thus, the exhibitor can know the appropriate price for the exhibited product by referring to the recommended price provided from the auction system 8. With this recommended price, the exhibitor can set the highest price in the range that can be sold as the seller's desired price.

  Further, the auction server 9 stores the transaction history in the online auction in the transaction information table 23. FIG. 4 is a diagram illustrating an example of a transaction information table. The transaction information table stores a successful bid date, a successful bid price, a book ID, a fixed price, and an ISBN for each item for which an online auction transaction has been performed. Each time a successful bid is determined, the auction server 9 stores the book ID, ISBN, fixed price, successful bid price, and successful bid date indicating the successful bid item in the transaction DB, and at least of the information. The transaction information including the successful bid price and the successful bid date is transmitted to the price estimation server 10 together with the completion of the transaction. Further, the auction server 9 sends the transaction history thus held to the price estimation server 10 in response to a request from the price estimation server 10.

<Price estimation server>
FIG. 5 is a conceptual diagram illustrating a configuration of the price estimation server 10 according to the first embodiment.

As shown in FIG. 5, the price estimation server 10 includes a control unit 31, a storage unit 32, a communication unit 33, and the like connected by a bus 35 as a hardware configuration. The storage unit 32 is, for example, a hard disk, and stores various types of information used in processing executed by the control unit 31. The control unit 31 includes one or more processors such as a CPU and peripheral circuits (ROM, RAM, interface circuit, etc.) used for processing of the processors. The communication unit 33 is connected to the network 1 and an internal network (not shown) of the auction system 8, and communicates between the auction server 9 and the price estimation server 10, the price estimation server 10, the content server 3, and the search server 4. Communication between the two. The present invention does not limit the communication mode of the price estimation server 10, and is realized by, for example, an IP packet. The price estimation server 10 may be constructed by a general-purpose computer such as a PC or a dedicated computer. The present invention does not limit the hardware configuration of the price estimation server 10.

  The price estimation server 10 implements each processing block shown in FIG. 5 by executing the application program stored in the storage unit 32 by the control unit 31. As such processing blocks, the price estimation server 10 includes an estimation process control unit 50, a product information acquisition unit 51, a transaction information acquisition unit 52, a peripheral trend information acquisition unit 53, a price estimation unit 54, a parameter learning unit 55, and the like. . As a database referenced by these processing blocks, the storage unit 32 stores a parameter table 57 and an explanatory variable information table 58. Hereinafter, each processing block of the price estimation server 10 will be described.

  The estimation process control unit 50 controls the start and end of the process in each of the process blocks, thereby estimating the appropriate price for the exhibited product (hereinafter referred to as price estimation process), and the estimation process. The process which produces | generates or updates the estimation model utilized by (it is hereafter described as an estimation model production | generation process) is performed.

  When the estimation process control unit 50 receives a request from the auction server 9, the estimation process control unit 50 starts a price estimation process for the exhibited item specified by the book ID included in the request. When the price estimation process is started, the estimation process control unit 50 gives the book ID of the price estimation target to each of the product information acquisition unit 51, the transaction information acquisition unit 52, and the peripheral trend information acquisition unit. The book information, transaction information, and peripheral trend information related to the ID are acquired. The estimation processing control unit 30 sends the book information, transaction information, and peripheral trend information to the price estimation unit 54 to cause the price estimation unit 54 to perform price estimation, and receives the estimated price information from the price estimation unit 54. . The estimation process control unit 30 sends the estimated price to the auction server 9.

  When receiving the notification that the transaction is completed from the auction server 9, the estimation processing control unit 50 starts the estimation model generation process based on the book ID, the winning bid price, and the winning bid date included in the notification. When the estimation process control unit 50 starts the estimation model generation process, each of the product information acquisition unit 51, the transaction information acquisition unit 52, and the peripheral trend information acquisition unit is given a book ID indicating the product for which the transaction has been established. Book information, transaction information, and peripheral trend information related to the book ID are acquired from the processing unit. The estimation processing control unit 50 newly stores the book information, transaction information, and peripheral trend information in the explanatory variable information table 58, and stores the newly stored information and the explanatory variable information table 58 so far. The parameter learning unit 55 is instructed to calculate the parameter value of the estimation model using the information. The estimation process control unit 50 updates the parameter table 57 with the parameter values calculated by the parameter learning unit 55.

  FIG. 6 is a diagram illustrating an example of the parameter table 57. As shown in FIG. 6, the parameter table 57 stores the values of the parameters of the estimation model used by the price estimation unit 54. In the first embodiment, since an estimation model using nine explanatory variables is taken as an example, the parameter table 57 stores values of ten parameters.

FIG. 7 is a diagram illustrating an example of the explanatory variable information table 58. As shown in FIG. 7, the explanatory variable information table 58 includes, for each transaction, a book ID, a fixed price, an average successful bid price, a search count, a search burst rate, a news appearance count, a blog appearance count, an author-specific difference, and a series difference. , State, successful bid date, and successful bid price are stored. Information acquired as book information is stored in each field of price and state. Information acquired as transaction information is stored in each field of the average successful bid price, successful bid date, and successful bid price. In each of the other fields, information acquired as peripheral trend information is stored. Among the information stored in the explanatory variable information table 58, the list price, the average successful bid price, the number of searches, the number of search bursts, the number of news appearances, the number of blog appearances, the author-specific difference, the series difference, and the state are explanatory variables of the estimation model This is information to be used, and the successful bid price is information to be used as an objective variable of the estimation model. Details of the average successful bid price, the number of searches, the degree of search burst, the number of news appearances, the number of blog appearances, author-specific differences, and series differences will be described later.

  The merchandise information acquisition unit 51 acquires book information related to a book ID that is an instruction target from the estimation processing control unit 50. The product information acquisition unit 51 acquires the book information of the book ID stored in the exhibition product information table 22 of the auction server 9 by sending a book information request designating the book ID to the auction server 9. The acquired book information includes at least a list price, a state, and ISBN. The merchandise information acquisition unit 51 sends the book information acquired in this way to the estimation processing control unit 50.

  When the transaction information acquisition unit 52 receives an instruction from the estimation processing control unit 50, based on the ISBN of the target book included in the instruction, the transaction information acquisition unit 52 sends transaction history information about the same book specified by the ISBN to the auction server 9. Request. In response to this request, the auction server 9 extracts, from the transaction information table 23, transaction information having the same ISBN, that is, all transaction information relating to the same book. Thereby, all the transaction information regarding the book (thing) to which the same ISBN was given about the thing with different book ID is extracted. When the transaction information acquisition unit 52 acquires all transaction information related to the same book from the auction server 9, the transaction information acquisition unit 52 calculates an average successful bid price for the acquired all transaction information. The calculated average successful bid price indicates a transaction record for the same book. The transaction information acquisition unit 52 sends the average successful bid price calculated in this way to the estimation processing control unit 50.

  In response to an instruction from the estimation processing control unit 50, the peripheral trend information acquisition unit 53 acquires peripheral trend information regarding the book that is the target of the instruction. This peripheral trend information is information stored in the explanatory variable information table 58, and includes the number of searches, the degree of search burst, the number of news appearances, the number of blog appearances, author-specific differences, and series differences. Hereinafter, a method for acquiring each piece of information will be described.

  The peripheral trend information acquisition unit 53 requests the auction server 9 for a book name, an author name, and a series name related to the book ID that is the target of the instruction from the estimation processing control unit 50. In response to this request, the peripheral trend information acquisition unit 53 specifies the book name and author name acquired from the auction server 9 and requests the search server 4 for information on the number of searches and the search burst. The search server 4 calculates each search count using the book name and author name specified in the request based on the stored search history, and each of the calculated search counts is a peripheral trend information acquisition unit 53. Send to. The peripheral trend information acquisition unit 53 averages the number of searches based on the book name acquired in this way and the number of searches based on the author name, and acquires this average value as the number of searches.

  Further, the search server 4 acquires the latest search burst degrees using the book name and author name specified in the request, and sends each search burst degree information to the peripheral trend information acquisition unit 53. The peripheral trend information acquisition unit 53 averages the search burst degree based on the book name thus acquired and the search burst degree based on the author name, and acquires the average value as the search burst degree.

  The number of searches and the degree of search burst obtained as the peripheral trend information indicate the search status of information related to the target book, and can be said to reflect the needs for the target book. For example, books with high needs from consumers are likely to search for information related to the books, and the number of searches and the degree of search burst will increase.

  Further, the peripheral trend information acquisition unit 53 designates the book name and author name acquired from the auction server 9 and performs a search for the news site (content server 3) and the blog site (content server 3) to the search server 4. Ask. In response to this request, the search server 4 executes each search for the news site and blog site by the book name by the search engine. Similarly, the search server 4 executes each search for the news site and the blog site by the author name by the search engine. The search server 4 sends each search result by the book name and author name for the news site and each search result by the book and author name for the blog site to the peripheral trend information acquisition unit 53.

  The peripheral trend information acquisition unit 53 uses each search result acquired in this way to calculate the average of the number of news sites on which the book title is posted and the number of news sites on which the author name is posted. The calculated average value is acquired as the number of news appearances. Similarly, the peripheral trend information acquisition unit 53 uses each search result to calculate the average of the number of blog sites that publish the book name and the number of blog sites that publish the author's name. Get the average value as the number of blog appearances.

  Since the number of news appearances and the number of blog appearances acquired as the peripheral trend information indicate the number of contents that provide information related to the target book, it can be said that the topicality of the target book is reflected. For example, for books with high topicality and books written by highly topical authors, the number of sites that post related information is likely to increase.

  The peripheral trend information acquisition unit 53 designates the author name and the series name acquired from the auction server 9, and displays all transaction information related to books with the same author name and all transaction information related to books with the same series name. Further request to 9. In response to this request, the auction server 9 extracts each record having the author name and series name from the exhibition product information table 22. Here, the books having the same series name include, for example, comic books and novels composed of a plurality of volumes. The auction server 9 extracts each transaction information having the same ISBN or book ID from the transaction information table 23 by using the ISBN or book ID of each extracted record as a search key.

  When the peripheral trend information acquisition unit 53 obtains all transaction information related to books with the same author name from the auction server 9, the peripheral trend information acquisition unit 53 calculates the difference between the regular price and the successful bid price for each transaction. The peripheral trend information acquisition unit 53 averages each difference by the number of transactions related to books with the same author name, and acquires this average value as a difference by author. Similarly, when the peripheral trend information acquisition unit 53 obtains all transaction information related to books of the same series name from the auction server 9, the peripheral trend information acquisition unit 53 calculates the difference between the regular price and the successful bid price for each transaction. The peripheral trend information acquisition unit 53 averages each difference based on the number of transactions related to books with the same series name, and acquires the average value as the series difference amount. The author-specific difference is a value indicating statistics of transaction performance between books of the same author, and the series difference is a value indicating statistics of transaction performance between books of the same series.

  The price estimation unit 54 estimates the appropriate price for the book by applying the book information, transaction information, and peripheral trend information related to a certain book sent from the estimation processing control unit 50 to the price estimation model. This price estimation model shows the causal relationship between the winning bid price of a book and the book information, transaction information, and peripheral trend information related to the book, and using the parameter values stored in the parameter table 57, ). The book information includes the list price and status, the transaction information includes the average successful bid price, and the peripheral trend information includes the number of searches, the number of search bursts, the number of news appearances, the number of blog appearances, author-specific differences, Series difference is included.

価格推定部５４は、ｘ_１に定価を、ｘ_２に平均落札価格を、ｘ_３に検索回数を、ｘ_４に検索バースト度を、ｘ_５にニュース出現数を、ｘ_６にブログ出現数を、ｘ_７に著者別差額を、ｘ_８に同シリーズ差額を、ｘ_９に状態をあてはめることにより、上記（式１）で示す回帰式から適正価格ｙを推定する。以降、ｙはこの推定モデルにより説明したい変数であるため目的変数と表記し、ｘ_１からｘ_９はこの目的変数を説明するための変数であるため説明変数と表記する。目的変数に掛け合わされる各パラメータ値は、各説明変数の目的変数に対する因果関係の大きさをそれぞれ示す。価格推定部５４は、このように推定された適正価格を推定処理制御部５０へ返信する。

Price estimation unit 54, the price to _{x 1,} the average bid price on the _{x 2,} the number of searches in _{x 3,} the search burst degree _{x 4,} the news number of occurrences in _{x 5,} the number of blogs appeared in _{x 6} , X ₇ , the difference by author, the series difference to x ₈ , and the state to x ₉ , the reasonable price y is estimated from the regression equation shown in (Expression 1) above. Later, y is written as target variable for a variable to be explained by the estimation model, x ₉ from x ₁ is referred to as explanatory variables for a variable for describing this objective variable. Each parameter value multiplied by the objective variable indicates the magnitude of the causal relationship of each explanatory variable to the objective variable. The price estimation unit 54 returns the appropriate price estimated in this way to the estimation process control unit 50.

  The parameter learning unit 55 receives an instruction of the estimation model generation process from the estimation process control unit 50 every time an auction transaction is established, and uses the information stored in the explanatory variable information table 58 according to the regression equation (Equation 1). Each parameter value of the estimated model shown is calculated. The parameter learning unit 55 acquires each parameter value of the estimation model represented by the regression equation (Formula 1) by multiple regression analysis. A weighted least square method is used for the multiple regression analysis in the first embodiment. That is, the parameter learning unit 55 obtains the actual successful bid price corresponding to the measured value of the objective variable for each record (each transaction) stored in the explanatory variable information table 58 and each objective variable acquired at the time of the transaction. Each parameter value is calculated such that the root mean square of the difference between the estimated price and the estimated price is minimized. The parameter learning unit 55 calculates each parameter that minimizes the following (Equation 2) using the number of records (number of transactions) I stored in the explanatory variable information table 58.

重みｗ_ｉは、新しい取引（説明変数情報テーブル５８に格納されるレコード）程、説明変数となる各情報の重みが高くなるように上記（式３）のように設定される。上記（式３）のｔ_ｃは現在の日を示し、ｔ_ｉはその落札日を示す。この重みｗ_ｉは、予め調整可能にメモリに保持されるようにしてもよい。

The weight w _i is set as in the above (Equation 3) so that the weight of each piece of information that becomes an explanatory variable increases as a new transaction (record stored in the explanatory variable information table 58). T _c of the equation (3) indicates the current day, t _i represents the successful bid day. The weight w _i may be held in the memory in an adjustable manner in advance.

  Each parameter that minimizes (Equation 2) can be obtained by solving simultaneous equations of (Equation 4) to (Equation 13) below, where a value obtained by partially differentiating (Equation 2) in each parameter is zero. The parameter learning unit 55 returns each parameter value thus obtained to the estimation processing control unit 50.

〔動作例〕
以下、実施例１における価格推定サーバの動作例について図８及び９を用いて説明する。図８は、実施例１における価格推定サーバの価格推定処理を示すフローチャートである。

[Operation example]
Hereinafter, an operation example of the price estimation server according to the first embodiment will be described with reference to FIGS. FIG. 8 is a flowchart illustrating the price estimation process of the price estimation server in the first embodiment.

  The price estimation server 10 receives an estimation instruction including the book ID of the book whose price is to be estimated from the auction server 9. In the price estimation server 10, when the estimation processing control unit 50 receives this instruction, the price information server 10 requests the product information acquisition unit 51 to acquire book information related to the book ID that is the target of the instruction.

  Upon receiving this request, the merchandise information acquisition unit 51 requests the auction server 9 for book information related to the book ID to be requested. The product information acquisition unit 51 acquires the list price, state, and ISBN related to the book ID stored in the exhibition product information table 22 from the auction server 9 (S81). The merchandise information acquisition unit 51 sends the information on the list price, state, and ISBN acquired in this way to the estimation processing control unit 50.

  Next, the estimation process control unit 50 sends a transaction information request including the ISBN to the transaction information acquisition unit 52. The transaction information acquisition unit 52 requests the transaction history information for the same book specified by the ISBN included in the request from the auction server 9 to acquire the transaction history information (S82). Thereby, the transaction information acquisition unit 52 calculates an average successful bid price for the same book as the target book for all the acquired transaction information (S83). The transaction information acquisition unit 52 sends the calculated average successful bid price to the estimation process control unit 50.

  Subsequently, the estimation processing control unit 50 sends a peripheral trend information request including the book ID to the peripheral trend information acquisition unit 53. The peripheral trend information acquisition unit 53 requests the auction server 9 for the book name, author name, and series name relating to the book ID included in the request, thereby acquiring the information stored in the exhibition product information table 22 (S84). ).

  The peripheral trend information acquisition unit 53 specifies the book name and author name acquired from the auction server 9 and requests the search server 4 for information on the number of searches and the degree of search burst. In response to this request, the peripheral trend information acquisition unit 53 acquires each number of searches using the book name and author name extracted from the search history held in the search server 4. The peripheral trend information acquisition unit 53 averages the number of searches by the book name acquired in this way and the number of searches by the author name, and acquires this average value as the number of searches (S85).

  Subsequently, the peripheral trend information acquisition unit 53 acquires, from the search server 4, search results for news sites and blog sites using the book name and author name based on the book name and author name acquired from the auction server 9. To do. The peripheral trend information acquisition unit 53 uses the search result acquired from the search server 4 to calculate the average of the number of appearances of the book name and the number of appearances of the author name on the news site, and the calculated average value Is acquired as the number of news appearances (S86). Similarly, the peripheral trend information acquisition unit 53 calculates the average of the number of appearances of the book name and the number of appearances of the author name on the blog site using each search result, and uses the calculated average value as the blog appearance. Obtained as a number (S86).

  Further, the peripheral trend information acquisition unit 53 designates the author name and series name acquired as described above and makes a request to the auction server 9 to thereby make a book with the same series name as all transaction information related to the book with the same author name. And obtain all transaction information (S87). Each transaction information is acquired from the exhibition product information table 22 and the transaction information table 23 in the auction server 9.

The peripheral trend information acquisition unit 53 calculates the difference between the list price and the winning bid price for each transaction based on all transaction information related to books with the same author name, and averages each difference with the number of transactions regarding the book with the same author name. This average value is acquired as the author-specific difference (S88). Similarly, the peripheral trend information acquisition unit 53 calculates the difference between the list price and the winning bid price for each transaction based on all transaction information related to the book with the same series name, and calculates each difference as the number of transactions regarding the book with the same series name. The average is obtained as the series difference (S88).

  The peripheral trend information acquisition unit 53 estimates and controls the peripheral trend information acquired as described above, that is, information regarding the number of searches, the degree of search burst, the number of news appearances, the number of blog appearances, author-specific differences, and series differences. Send to part 50.

  The estimation process control unit 50 relates to the target book, the list price and state acquired by the product information acquisition unit 51, the average successful bid price acquired by the transaction information acquisition unit 52, the number of searches acquired by the peripheral trend information acquisition unit 53, A price estimation request is sent to the price estimation unit 54 together with the search burst degree, the number of news appearances, the number of blog appearances, the author-specific differences and the series differences.

  Upon receiving the price estimation request, the price estimation unit 54 acquires an estimated parameter value from the parameter table 57 (S89). The price estimation unit 54 substitutes the book information, transaction information, and peripheral trend information about the target book into explanatory variables, and applies each estimated parameter acquired from the parameter table 57 to the regression equation (formula (1)). Then, an appropriate price as an objective variable is calculated (S90). The price estimation unit 54 sends the appropriate price of the target book calculated in this way to the estimation processing control unit 50. Finally, the price estimation server 10 returns the appropriate price estimated by the price estimation unit 54 to the auction server 9.

  The auction server 9 presents the estimated price as an appropriate price of the target book to an exhibitor who sells the book. The exhibitor using the exhibitor terminal 6 can set the highest price in the range that can be sold for the book to be displayed as the seller's desired price by referring to the presented price.

  FIG. 9 is a flowchart illustrating an estimation model generation process of the price estimation server according to the first embodiment. When the auction server 9 determines a successful bid for the exhibited product, the auction server 9 sends a notification to the effect that the transaction has been established to the price estimation server 10. In the price estimation server 10, when the estimation processing control unit 50 receives the transaction establishment notification (S91), the estimation model generation processing described below is started based on the book ID, the winning bid price, and the winning bid date included in the notification. To do.

  When the estimation model generation process is started, the estimation process control unit 50 acquires book information, transaction information, and peripheral trend information related to the book ID indicating the book for which the transaction has been established (S81 to S88), similarly to the price estimation process described above. ). Specifically, the estimation processing control unit 50 obtains the list price and state from the product information acquisition unit 51, and the average successful bid price from the transaction information acquisition unit 52. The number of searches, the degree of search burst, the number of news appearances, the number of blog appearances, the author The separate difference and the series difference are acquired from the peripheral trend information acquisition unit 53. The estimation process control unit 50 provides the information obtained in the same manner as during the price estimation process, along with the successful bid date, the successful bid price, and the book ID indicating the book for which the transaction has been established, as explanatory variable information. It adds to the table 58 (S92). Subsequently, the estimation process control unit 50 instructs the parameter learning unit 55 to generate the estimation parameters A to J used in the price estimation process described above including information on the added latest transaction. .

When receiving an instruction from the estimation processing control unit 50, the parameter learning unit 55 extracts information for a predetermined number of records from the latest record among the information stored in the explanatory variable information table 58. Here, since information for each transaction is stored in the explanatory variable information table 58, this one record corresponds to one transaction. The predetermined number of records may be the total number of records, or a predetermined value may be stored in the memory. The parameter learning unit 55 uses the number of records (number of transactions) I in the extracted explanatory variable information table 58 to calculate the value of each parameter A to J that minimizes (Equation 2). At this time, as shown in the above (Formula 2), the parameter learning unit 55 multiplies a new transaction and a weight w _i that increases the weight of each information serving as an explanatory variable. As a result, the parameter learning unit 55 calculates the values of the parameters A to J by solving the simultaneous equations of (Expression 4) to (Expression 13) (S93).

  The parameter learning unit 55 updates the values in the parameter table 57 with the calculated values of parameters A to J (S94). Note that the order of step S81 to step S88 in the price estimation process and the estimated model generation process shown in FIGS. 8 and 9 may be switched. The present invention does not limit the order in which the book information, transaction information, and peripheral trend information of the target book are acquired, and may be executed in parallel.

[Operation and Effect in Example 1]
As described above, in the price estimation server 10 according to the first embodiment, the price estimation unit 54 digitizes various pieces of information about a target book, and serves as an explanatory variable of an estimation model (regression equation) generated using multiple regression analysis. By using each of these pieces of information, the appropriate price of the book, which is the objective variable of the estimation model, is estimated. This is because the price estimation target is a product that is traded in an auction or a second-hand market, and it is considered that the appropriate price is determined from various information related to the product. Actually, if you consider the price of a book, if the book is featured in many media such as TV and the Internet, or if the author of the book wins a famous prize, Is determined, the winning bid price of the book and the transaction price in the secondhand market tend to rise.

  In the price estimation server 10 according to the first embodiment, such various information that affects the price of the product is used as an explanatory variable used in an estimation model for price estimation. These various types of information are acquired by the product information acquisition unit 51, the transaction information acquisition unit 52, and the peripheral trend information acquisition unit 53. Specifically, as various information, in addition to book information such as the list price and the state of the book, an average successful bid (transaction) price that shows past transaction results for the same book, and a search that reflects the needs related to the book Number of times and search bursts, number of news appearances and blog appearances reflecting topicality, difference by author showing statistics of transaction performance between books of the same author, difference of series showing statistics of transaction performance between books of the same series Is used.

  As a result, the price estimated in the first embodiment is likely to be the highest price in the range that can be sold, that is, the appropriate price. Therefore, the exhibitor can know the appropriate price in the auction or second-hand market for the exhibited product by referring to the price estimated by the price estimation server 10 in the first embodiment.

  Furthermore, in the price estimation server 10 according to the first embodiment, every time a transaction is performed, the transaction performance and various information states used in the explanatory variables at the time of the transaction are stored in the explanatory variable information table 58. The parameter learning unit 55 estimates the optimum values of parameters A to J used in the price estimation model by using past performance information stored in the explanatory variable information table 58. Here, the parameters A to J of the price estimation model indicate how much each information should be combined with the price of the product.

  Thereby, according to Example 1, since the parameter of a price estimation model is updated by the optimal value whenever a transaction is performed, the estimation accuracy of a reasonable price can be improved.

Furthermore, in the price estimation server 10 according to the first embodiment, a weighted least square method is used to estimate the optimum value of the parameter of the price estimation model. The price estimation model is estimated using past transaction results. However, since the transaction price is considered to change with time, the price estimation model needs to follow the change. Although it is considered that recent transactions are more important than past transactions in estimating reasonable prices, estimation using simple multiple regression analysis treats past transactions and recent transactions equally. End up. Therefore, in the first embodiment, a parameter is calculated by multiplying information about each transaction by a weighting factor so that new performance information is emphasized among past transaction results.

  Therefore, according to the first embodiment, it is possible to perform price estimation using a price estimation model generated with a more optimal parameter value. If the estimation accuracy of the appropriate price can be improved, the highest price in the range that can be sold can be set as the seller's desired price, so that the seller's profit can be maximized. Furthermore, if such an appropriate price can be provided to the exhibitor, the satisfaction of the exhibitor will increase, leading to an increase in the ability to attract customers at the auction site.

[Modification]
In the above-described first embodiment, for convenience of explanation, the case of a used book is cited as an example of a product that is traded in an online auction. However, the present invention does not limit the types of such target products.

  When dealing with merchandise that is not a book, for example, a merchandise name or manufacturer name may be used instead of the book name or author name. Further, when the product does not have a standard ID such as ISBN of a book and it is difficult to acquire product information from the content server 3, it may be acquired by causing the user to input the product information. .

  In the above-described first embodiment, the list price, the average successful bid price, the number of searches, the number of search bursts, the number of news appearances, the number of blog appearances, the author-specific differences, the series-specific differences, and the status are all used as explanatory variables. The price is estimated, but some of the information that has a strong causal relationship with the winning bid may be used, or other information that has a strong causal relationship with the winning bid may be used. Good. The present invention does not limit the number of explanatory variables used as a model for estimating a reasonable price and the information thereof.

DESCRIPTION OF SYMBOLS 1 Network 3 Content server 4 Search server 6 Exhibitor terminal 7 Buyer terminal 8 Auction system 9 Auction server 10 Price estimation server 20, 31 Control part 21, 32 Storage part 28, 33 Communication part 50 Estimation process control part 51 Product information acquisition Unit 52 transaction information acquisition unit 53 peripheral trend information acquisition unit 54 price estimation unit 55 parameter learning unit 57 parameter table 58 explanatory variable information table

Claims (6)

Acquisition means for acquiring product information including price information about the target product, transaction information based on transaction results regarding the target product, and peripheral trend information including information reflecting at least one of needs and topicality regarding the target product;
In the estimation model showing the relationship between the product information, the transaction information and the peripheral trend information and the product price, by using the product information, the transaction performance information and the peripheral trend information acquired by the acquisition unit as explanatory variables Estimating means for estimating the price of the target product as an objective variable;
A price estimation device comprising: Each time the transaction price of a product is determined, each piece of information used as an explanatory variable of the estimation model acquired by the acquisition means for the product, and the transaction price that becomes an objective variable of the estimation model for the transaction A history storage unit for storing
Update means for updating the value of each estimation parameter of the estimation model by multiple regression analysis using each information stored for each transaction stored in the history storage unit,
The price estimation apparatus according to claim 1, further comprising:   The updating means uses a least-squares method using a weighting factor that emphasizes new information of history among each information stored in the history storage unit when updating the value of each estimation parameter. The price estimation apparatus according to claim 2, wherein   The product information includes a list price and a state of the product, the transaction information includes an average transaction price based on a transaction record of the same product as the target product, and the peripheral trend information includes a content server related to the target product. The price estimation device according to any one of claims 1 to 3, characterized in that the number of searches for and the number of publications on the content server are included. Computer
An acquisition step of acquiring peripheral trend information including product information including price information about the target product, transaction information based on a transaction result regarding the target product, and information reflecting at least one of needs and topicality regarding the target product;
In the estimation model showing the relationship between the product information, the transaction information and the peripheral trend information and the product price, by using the product information, the transaction performance information and the peripheral trend information acquired by the acquisition unit as explanatory variables An estimation step for estimating the price of the target product as an objective variable;
The price estimation method characterized by performing. On the computer,
An acquisition step of acquiring peripheral trend information including product information including price information about the target product, transaction information based on a transaction result regarding the target product, and information reflecting at least one of needs and topicality regarding the target product;
In the estimation model showing the relationship between the product information, the transaction information and the peripheral trend information and the product price, by using the product information, the transaction performance information and the peripheral trend information acquired by the acquisition unit as explanatory variables An estimation step for estimating the price of the target product as an objective variable;
A price estimation program characterized in that it is executed.

Images