JP2001175686A - Method for searching for article - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a searching method for an article which actualizes a selection structure unique to a person who desires to buy and extracts the best article for the individual who desires to buy. SOLUTION: The individual who desires to buy is made to select more than one level by properties as desirable conditions (step 1), plural profiles are generated by combining the levels by properties included in the desirable conditions (step 2), and evaluations of those profiles by the individual who desires to buy are numerized (step 4); and weights of evaluations characteristic of the individual who desires to buy as to respective levels of properties are calculated from those evaluated values (step 5), evaluated values are estimated for articles who meet the desirable conditions in a database 53 by using the weights of evaluations characteristic of the individual who desires to buy (step 6), and an article having a large estimated evaluated value is extracted as a search result (step 7).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of searching for a product using a database, and more particularly, to a search for a product which extracts a product best suited to the individual who wishes to purchase by making the selection structure unique to the purchaser apparent. It is about the method.

[0002]

2. Description of the Related Art In purchasing a product, when selecting a preferable product from a large number of products, if there are various factors characterizing the product, the selection becomes complicated and difficult. In some cases, it may not be possible for the person who is going to purchase to clearly understand his or her preferences, and in such a case, the selection becomes difficult. In addition, in the case of products that are rarely purchased, selection becomes difficult.

[0003] For example, real estate includes area, layout, location,
Since various factors such as price have to be considered, it takes a lot of effort to decide which property to purchase. Examples of this type of product include, besides real estate, automobiles, travel, financial products, and clothing.

It is convenient to register product data in a database so that a purchaser can search for a desired product from the database. In the database, the elements that characterize a product, such as the area, floor plan, location, and price, are called attributes, and discrimination information based on numerical values and matters for each attribute is called a level. That is, in the database, product data representing various attributes that characterize a product at multiple levels for each attribute is registered. The purchaser is caused to select a level for each attribute as a desired condition, and a product whose level of each attribute matches the desired condition is extracted from the database.

[0005] It is important that a large number of commodities are registered in the database. By increasing the number of types of attributes, it is possible to respond to various purchaser preferences.
Increasing the number of levels allows you to distinguish small differences. However, when the number of products, the number of attributes, and the level increase, the amount of data increases. Therefore, in order to quickly and accurately select a product, an excellent search rule is required. Whether the database is used effectively for product selection depends largely on the search rules.

[0006]

The conventional search rules are as follows:
In many cases, a searcher sequentially inputs conditions and narrows the range of product selection. This search rule requires that the buyer wants to explicitly recognize his or her best desired conditions in advance. Further, when narrowing down while adding conditions, it is necessary that the purchaser's own selection structure of which element (attribute) should be prioritized be explicitly recognized in advance.

However, according to many examples, the desired condition expressed by the purchaser consciously does not necessarily reflect the desire under the consciousness of the purchaser, and the purchaser who wishes to purchase from the beginning is the best for himself / herself. It is rare that the desired conditions and the selection structure are explicitly recognized. Therefore, in order to find a satisfactory product, it is necessary to repeat trial and error of the search. Moreover, there is no guarantee that the finally retrieved product is optimal for the purchaser. In fact, even if the price is high, the area may be small, but for any buyer who wants a new building anyway, a contractor who could not obtain an agreement even after successively selecting and introducing properties that match the desired conditions, In some cases, an agreement was reached immediately after introducing a low-priced, large-area property.

Further, the conventional search rule only extracts commodities that match the desired conditions issued by the purchaser, and does not consider the substitutability between elements (attributes).
For example, "it may be of new construction 60m ² apartment at 30 million yen, condominium of 27 million yen if 100m ² be as good" can not search in the hope conditions including the alternative of, the price from 30 million yen With a width of up to 27 million yen and an area of 60 m ² to 100 m
To have a width up to ^2, and when the building age would be a comprehensive desired conditions that also allowed also used new construction, will be extracted a huge number of goods, purchasing, such as that 60m ² in the Middle Ages is in it Even things that do not meet the wishes of the applicant are included.

In recent years, there is a search rule called case-based reasoning. This method collects examples of selection relationships between various products for many people, and based on the cases of purchaser's preferences and past selection experience, says, "People with similar tastes to you It seems that they like such a product (they have selected such a product in the past). " For example, a bag has the A brand,
If there are cases where people who like cars like car B often buy type C wristwatches, those who want to buy watches have brand A for their bags and if they like car B,
Recommend a type of watch. However, case-based reasoning suggests that buyers may have chosen in the past,
Although it is effective for those who can easily judge likes and dislikes, it is difficult to apply to those with little selection experience and those whose selection structure is complicated and it is difficult to judge likes and dislikes. Therefore, case-based reasoning is not suitable for real estate, automobiles, travel, financial products, products related to marriage, and the like.

Accordingly, an object of the present invention is to solve the above-mentioned problems and to provide a product search method for extracting a product optimal for an individual purchaser by realizing a selection structure unique to the purchaser. .

[0011]

SUMMARY OF THE INVENTION In order to achieve the above-mentioned object, the present invention provides a method of purchasing a purchaser from a database in which product data representing various attributes characterizing a product at a plurality of levels are registered for each attribute. When searching for a desired product, a purchaser is asked to select one or more levels for each attribute as desired conditions, and a plurality of profiles composed by combining the levels for each attribute included in the desired conditions are created. The evaluation of the purchaser with respect to the profile is quantified, the weight of the evaluation specific to the purchaser with respect to each level of each attribute is calculated from these evaluation values, and the purchaser with respect to the product satisfying the desired condition in the database is calculated. An evaluation value is estimated using a unique evaluation weight, and a product having a large estimated evaluation value is extracted as a search result.

A desired condition selecting step of presenting the purchaser with one or more levels of options for each attribute and selecting an option of each attribute as a desired condition; and an optional level included in the desired condition for the predetermined attribute. A representative product presenting step of retrieving representative products matching the levels of the respective attributes from the database from the database and presenting the representative products to the purchaser, A numerical evaluation step of quantifying the evaluation of the purchaser with respect to the representative product, and a combination of the evaluation value of the representative product and the level of each attribute of the profile. A utility coefficient deriving step of calculating a utility coefficient representing a size as a weight of the evaluation specific to the purchaser; The products satisfying the Nozomu condition retrieved from said database, a search step for estimating the evaluation value from the levels and utility coefficient for each attribute for the searched product is extracted as the search result items in the descending order of the estimated evaluation values,
A search product presentation step of presenting a product with a large estimated evaluation value to a purchaser.

A plurality of commodities may be presented to a purchaser in descending order of the estimated evaluation value.

[0014] For the plurality of representative commodities, the purchaser may be evaluated by scoring each of the representative commodities, and the score may be used as the evaluation value.

[0015] A small number of representative commodities selected at random from the plurality of representative commodities are presented, the purchaser evaluates the superiority of the presented representative commodities, and the superiority is converted into an evaluation value. Is also good.

[0016]

An embodiment of the present invention will be described below in detail with reference to the accompanying drawings.

FIG. 1 shows the basic procedure of the commodity search method according to the present invention. Here, the target product is rental real estate. Therefore, the procedure for searching for rental real estate is shown in Figure 2.
As shown in FIG.

First, as shown in FIG.
In the above, a database is constructed in which product data, which is expressed in a plurality of levels for each attribute of various attributes characterizing a product, is registered. This step 0 corresponds to the procedure of FIG. That is, as shown in FIG. 2, the property of the rental real estate is registered in the database. For simplicity of explanation, the examples are limited to five properties A to E. Property data includes type, area, age, rent, station distance, floor plan,
It consists of level data for seven types of attributes in the area and specific individual data (not shown) for each attribute.
The individual data represents the value or matter of the property itself. For example, if the area of the property is 76 m ² ,
The value m ² is the individual data. The level data is information for distinguishing each attribute by a numerical value or a matter. The type attribute has two levels of levels for families and individuals. The attributes of the area, less than ^{60m 2, 60~80m 2, 80~100m 2} , 100~12
4 stages levels of 0 m ² is provided. The attributes of the age of the building are provided in four levels: new construction, 1-5 years, 6-10 years, 11 years or more. There are four levels of rent attributes: less than 100,000 yen, 100,000 to 150,000 yen, 15 to 200,000 yen, and 20 to 250,000 yen. Station distance attributes include
There are three levels within 5 minutes, 5 to 10 minutes, and 10 to 15 minutes. 2DK, 2LD
There are three or more levels such as K and 3LDK.
The regional attributes are provided with a plurality of levels, such as in the 23 wards of Tokyo, Kanagawa Prefecture, and Chiba Prefecture. Of course, the type of the attribute, the number of levels, and the values for the levels are not limited to the above examples. In addition, image information such as a floor plan and a location map may be stored in the database in addition to the property data which is character information. Attributes relating to the layout such as the number of windows and storage arrangement constitute property data using levels divided into some typical arrangement patterns, but a sketch drawing can be used when presenting the property data.

The search is started from step 1. In step 1, options are presented to the purchaser, and a desired condition is selected. This step 1 corresponds to the procedure of FIG. As shown in FIG. 2, by presenting options for each element in a question format to a purchaser, it is easier to select a desired condition. In the example shown, one room,
Apartment for families, three choices of detached house, rent,
Less than 100,000 yen, 100,000-190,000 yen, 20-290,000 yen, 30
Four choices of 10,000 yen or more, and as a region, a plurality of choices of 23 wards in Tokyo, Kanagawa, Chiba, Saitama, ... are presented. The numerical value of the option delimiter may be different from the value that grades the level for each attribute of the database. By selecting an option as a desired condition, one or more levels can be selected. The purchaser answers desired conditions as absolute conditions. That is, options are selected on the basis that properties other than the selected condition are unnecessary.

In step 2, a profile is formed by using attributes constituting a utility function described later. The profile is a pattern configured by appropriately combining the levels of each attribute included in the desired condition. This step 2 corresponds to the procedure of FIG. Instead of forming a new profile each time a desired condition is obtained from a purchaser, a profile that matches the desired condition may be extracted from a group of profiles prepared in advance. Also, theoretically, there are profiles that match the desired conditions, differing only by multiplying the number of levels of each attribute constituting the utility function, but based on the orthogonal table of the experimental design method, The minimum number of profiles required for calculating the utility function can be obtained. For example, when an attribute constituting a utility function has the level of each of the four stages were in each attribute a 4, 4 ⁴
Although 16 profiles are conceivable, in practice, it is sufficient to prepare 16 profiles. FIG. 2 shows only profile A, profile B, and profile C.
Here, there are six types of attributes related to the profile: area, age, commuting time, rent, station distance, and floor plan.
In the description of the procedure, commuting time is not illustrated as an attribute, but it is assumed that commuting time data is also registered in the database.

In step 3, only one property whose level of each attribute matches for each profile is searched from the database as a representative property. Then, the individual data of all the searched representative properties is presented to the purchaser. This step 3 corresponds to the procedure of FIG. As shown in the figure, individual data of the property A, the property B, and the property C are presented. These properties are profile A,
It corresponds to the representative property of profile B and profile C. Illustration of representative articles of other profiles is omitted.

In step 4, the evaluation of the purchaser for the presented representative article is quantified, and the numerical value is used as the evaluation value of the profile. This step 4 corresponds to the procedure of FIG.

There are various methods of digitizing, but a profile that is preferable for a purchaser is set to show a high numerical value. Hereinafter, two digitization methods will be described.

The first method is a direct evaluation method, in which all representative articles to be listed are displayed in a list, and a purchaser gives a score to each article while comparing the whole. In the illustrated example, 10 points are closest to the ideal, 7 points are slightly closer to the ideal,
5 points for neither, 3 points for slightly far from ideal
Although the score is based on a scoring format of 0 points for a very far point, the scoring format is not limited to this. However, it is important that the buyers themselves score. According to the direct evaluation method, a purchaser can give an evaluation in consideration of an important element (attribute). This method is suitable for a target product whose specifications (functions / performances) are easy to select, such as a camera or a personal computer.

The second method is an indirect evaluation method, in which only a small number of representative properties, for example, two, are presented, and a purchaser evaluates the superiority or inferiority of the two presented properties. By repeating the evaluation by changing the combination of the two items to be presented, it is possible to grasp the rank of the representative property. It should be noted that, in theory, in order to determine the rank of all representative properties, it is necessary to evaluate all two combinations, but it is necessary to estimate the overall rank from the evaluation of less than all combinations. It is possible. Also, if three or four cases are presented at a time to evaluate the superiority, the number of evaluations can be reduced. Once you understand the hierarchy of representative properties in this way,
A property with a high ranking is given a high rating value, and a property with a low ranking is given a low rating value. This method is suitable for products that need to consider various factors at the same time, such as real estate and automobiles. Details of the indirect evaluation method will be described later.

In step 5, a utility coefficient is calculated from the evaluation value obtained in the previous step and a combination of the levels of the attributes of the profile. The utility coefficient represents the magnitude of the utility that each level of each attribute has on the evaluation, and means the weight of the evaluation unique to the purchaser. This step 5 corresponds to the procedure of FIG. By calculating all the utility coefficients, a utility function is obtained. Here, the utility function is a linear function, and a utility value Y is obtained by summing a product of a variable ζjk specifying a level for each attribute and a utility coefficient Xjk provided for each variable for all attributes.

[0027]

(Equation 1)

However, the variable ζjk can take any value of 1, 0, and for a specific attribute j (j = 1, 2,...) And a specific level k (k = 1, 2,...) Take the value 1,
Otherwise, it is a variable taking the value 0. If the attribute j is represented by an attribute name instead of a numerical value, 1 is the area and 2 is the age. When using the attribute name for the variable ζjk, the level k
Write the attribute name to the right of the value of. As an example of the value of the variable ζjk, the ζ1 area is a value 1 if the area is 60 m ² , the value 0 otherwise, and the ζ2 area is a value 1 if the area is 60 to 80 m ² , and a value 0 otherwise. The $ 4 rent has a value of 1 if the rent is 200,000 yen or more, and a value of 0 otherwise.
α is a constant term that is common to all profiles and does not depend on the variable ζjk, and can be determined as appropriate.

In one profile, only one variable has a value of 1 for one attribute. Therefore, a matrix of variables に 関 す る jk for all profiles can be represented as shown in FIG.

The calculation of the utility function is to derive the value of each utility coefficient Xjk using the least squares method so that the utility value Y in the above equation is the measured value, that is, the evaluation value of each profile. It is. The utility coefficient Xjk represents the magnitude of the utility that each level k of each attribute j has on the evaluation (utility value Y), and corresponds to a quantification of the evaluation weight unique to the purchaser. In this way, the selection structure of the buyer who has not been explicitly recognized by the purchaser is expressed by a mathematical expression called a utility function. FIG. 3 shows an example of the obtained utility function. In this example, the area of the utility coefficient X1 is -1.2.
5. The area of the utility coefficient X2 is -0.75, and the age of the utility coefficient X4 is -1.38.

The procedure may be performed following the procedure. In the procedure, the utility coefficient Xjk is graphed to present the selection structure of the purchaser. From the graph of FIG. 3, it can be seen that the purchaser places the highest importance on the age of the building, and in particular, has a remarkably strong commitment to new construction. on the other hand,
There is also a tendency to attach importance to the area.
If it is less than 0 m ² , it can be predicted that the evaluation will decrease.

In step 6, a product which satisfies the desired condition is searched from the database, an evaluation value is estimated from the level and utility coefficient for each attribute of the searched product, and a product having a large estimated evaluation value is extracted as a search result. I do. This step 6 corresponds to the procedure of FIG. That is, only properties that satisfy the desired conditions are searched from the database. As a result, properties that the purchase applicant does not need are excluded, and the search target is narrowed down to only properties that satisfy the desired conditions. Next, a utility function is applied to the searched property to calculate a utility value Y, and the obtained utility value Y is used as an estimated evaluation value. The properties are sorted in descending order of the estimated evaluation value.

In step 7, a product having a large estimated evaluation value is presented to a purchaser. That is, as in the procedure, a plurality of properties having a high sorting order are presented to the purchaser as “your ideal property”. At this time, it is assumed that the order can be specified by the purchaser.

As shown in the figure, the search results of the properties are as follows: Property B, Property C, Property A,.
However, the properties A, B, C and the representative properties A,
B and C are not necessarily the same property. From the search results, it can be seen that the property B is not at the level where the building age and area are the highest, but is at a level that does not cause any dissatisfaction.

A property having a high estimated evaluation value as described above is considered to be a property having a high probability that a purchaser gives a high evaluation. In other words, it is the most suitable property for buyers. For a purchaser, a property with high likability will be presented, and the trust in the property search service will increase. The potential selection structure of the purchaser, which was unknown by the conventional method, has been clarified, and the utility function can be calculated numerically. Furthermore, the substitutability between elements (attributes) that could not be considered by the conventional method is incorporated in the utility function. In this way, according to the present invention, it is possible to make the selection structure unique to the purchaser obvious and extract the most suitable product for the individual purchaser.

In the procedure, the object of the present invention can be achieved even if a profile is presented without searching and presenting a representative property of the profile. In this case, the profile is directly evaluated, but if the evaluation value is the same as that of the case where the representative property is presented, the same utility function is obtained. When a representative property is presented as in the example of FIG. 2, the purchaser evaluates a specific property, so the truth is generated, and scoring can be performed as if the user is actually purchasing.

Next, a description will be given of an apparatus configuration for executing the product search method of the present invention.

As shown in FIG. 5, a server 52 is connected to a wide area communication network 51, and the server 52 has a database 53. Any terminal 5
4 can access the server 52 via the communication network 51 and receive an information service using the database 53. A purchaser uses a personal computer at home or a device at a store as the terminal 54. The server 52 is equipped with the search engine according to the present invention. When this search engine executes the procedure (1), a screen corresponding to the procedure is displayed on the terminal 54.

In step 1 (procedure) for presenting options, options for a plurality of elements are displayed on one page in the window portion 61 of the screen of the terminal 54, as shown in FIG. Here, as the area options 62, the prefectures and 23 wards of Kanto are individually indicated, and the other prefectures are indicated collectively. There is also an option to select all. If all are selected, there is no condition for that element. Eight types of layout options 63 are displayed including an option of selecting all. Nine options are displayed as area options 64 including an option to select all. As for the option 65 of the station distance, options for each mode of transportation are shown. Each option is selected by an operation such as mouse click, and it is expressed that the selection is made by discoloration, blinking, or the like. The rent option 66 is not a fixed option but a lower limit and an upper limit are displayed as numerical values, and the numerical values can be arbitrarily changed by an operation such as mouse click. Here, in addition to the options, a work address can be input. The search engine calculates the commuting time on the shortest route from the property using the office information. The options may be displayed over a plurality of pages.

In step 3 (procedure) of presenting a representative property, as shown in FIG. 7, the name 72 of the representative property, each element 73, and a sketch 74 are displayed in a window portion 71 of the screen of the terminal 54. Is done. Here, an indirect evaluation method of presenting three cases and evaluating the superiority by ranking among the three cases is adopted, and the ranking is made by an operation such as a mouse click. Since not only the character information but also the sketch 74 is displayed, the superiority evaluation becomes easy.
When presenting a product as described above, image information of the product may be stored in a database in advance, and the image information may be displayed along with character information. If the product is real estate, the layout of the room is an important factor, so a sketch is displayed.
The data relating to the layout is difficult to understand even when presented with character information, but can be readily understood with image information. Therefore,
For example, the attribute level that the balcony faces south is element 7
3 and will be presented in a sketch drawing 74.
Similarly, when the product is a car, the appearance is an important factor, and therefore, a picture of the car may be displayed. For example, the level of red in the attribute of the paint color is presented in the color of the picture.

In the step (procedure) of presenting the selection structure, as shown in FIG. 8, a window portion 81 of the screen of the terminal 54 includes a bar graph 82 for attributes such as area, age, commuting time, and rent. Is displayed. The length of the bar is
It shows the magnitude of the utility coefficient for each level. Since such a bar graph is displayed, the purchaser can clearly understand his / her own selection structure that he did not realize until then.

In step 7 (procedure) for presenting a property with a high estimated evaluation value, a list of ranked properties is displayed in the window portion 91 of the screen of the terminal 54 as shown in FIG. Detailed information of each property can be further opened and displayed by an operation such as mouse click.

Next, the indirect evaluation method will be described in detail.

As shown in FIG. 10, when presenting a representative property, one sheet (a screen page in the case of a terminal) 1
List two representative properties in 1 and the buyers will evaluate the superiority with a circle. When the round cross evaluation is completed for each combination, a round cross evaluation table 12 that describes the superiority of the property that is the subject of the evaluation with the partner property is completed. Based on the round cross evaluation table 12, a matrix 13 is created as regression analysis data, in which the subject is 1, the partner is -1, and the superiority is ± 1 (excellent 1 and inferior -1). Using the matrix 13, each property is used as an explanatory variable of the regression model, and the superiority or the inferiority is used as a dependent variable, and a regression coefficient for the variable is determined by the least squares method. As a result of this regression analysis, a regression coefficient for each property is obtained. The value of this regression coefficient is used as the evaluation value of the profile including the representative property.

As described above, according to the present invention, the potential selection structure of the purchaser is revealed as a utility function (or utility coefficient) that can be numerically calculated from the evaluation value of the purchaser with respect to the profile. The search engine can estimate the evaluation value and extract products that are likely to be highly evaluated by purchasers.

The search target is not limited to rental real estate.
It goes without saying that the present invention can be widely applied to transferred real estate, automobiles, travel, financial products, clothing and the like.

[0047]

The present invention exhibits the following excellent effects.

(1) Since the potential selection structure (weight of evaluation) of the purchaser is revealed as a utility function (utility coefficient) that can be numerically calculated, a product preferable for the purchaser can be effectively extracted from the database. Can be extracted.

(2) In the utility function (utility coefficient), the substitutability between elements is also quantified, so that alternatives can be extracted at the same time.

[Brief description of the drawings]

FIG. 1 is a basic procedure diagram of a product search method of the present invention.

FIG. 2 is a flowchart illustrating a procedure for searching for a property to be rented according to the present invention.

FIG. 3 is a flowchart illustrating a procedure for searching for a property to be rented according to the present invention.

FIG. 4 is a diagram of a matrix of a variable ζjk according to the present invention.

FIG. 5 is a configuration diagram of an apparatus that executes a product search method of the present invention.

FIG. 6 is a configuration diagram of a screen for presenting options according to the present invention.

FIG. 7 is a configuration diagram of a screen for presenting a representative property in the present invention.

FIG. 8 is a configuration diagram of a screen that presents a selection structure according to the present invention.

FIG. 9 is a configuration diagram of a screen for presenting a property having a high estimated evaluation value according to the present invention.

FIG. 10 is a flowchart of the indirect evaluation method in the present invention.

[Explanation of symbols]

 51 Communication Network 52 Server 53 Database 54 Terminal

Claims (5)

[Claims] When searching for a product desired by a purchaser from a database in which product data representing various attributes characterizing the product at a plurality of levels are registered for each attribute, the purchaser requests the purchaser as a desired condition. One or more levels are selected for each attribute, a plurality of profiles are created by combining the levels for each attribute included in the desired condition, the evaluation of purchasers for these profiles is quantified, and these evaluation values are calculated. Calculate the weight of the purchaser-specific evaluation for each level of each attribute from, and estimate the evaluation value using the weight of the purchaser-specific evaluation for the product that satisfies the desired condition in the database. A product search method characterized by extracting a product having a large evaluation value as a search result. 2. A desired condition selection step of presenting a purchase candidate with one or more levels of options for each attribute and selecting an option of each attribute as a desired condition, and an optional condition included in the desired condition for a predetermined attribute. A profile forming step of forming a plurality of different profiles by combining the levels of, and a representative product presenting step of searching the database for representative products matching the level of each attribute with these profiles and presenting them to purchasers, A numerical evaluation step of quantifying the evaluation of the purchaser with respect to the presented representative product; and a combination of the evaluation value of the representative product and the level of each attribute of the profile, the level of each attribute affecting the evaluation. A utility coefficient deriving step of calculating a utility coefficient representing the magnitude of the utility as a weight of the purchaser-specific evaluation; and A search step of searching for a product satisfying the desired condition from the database, estimating an evaluation value from the level and utility coefficient for each attribute of the searched product, and extracting products as search results in descending order of the estimated evaluation value; 2. A method for presenting a product according to claim 1, further comprising a search product presenting step of presenting the product having a large estimated evaluation value to a purchaser. 3. The method according to claim 1, wherein a plurality of commodities are presented to the purchaser in descending order of the estimated evaluation value. 4. A method for searching for a product according to claim 2, wherein the purchaser is allowed to score and evaluate each of the plurality of representative products, and the score is used as an evaluation value. . 5. Presenting a small number of representative products randomly selected from the plurality of representative products, letting a purchaser evaluate the superiority of the presented representative products, and converting the superiority rank into an evaluation value. The method for searching for a product according to claim 2, wherein the search is performed.