KR102200340B1 - Method and apparatus for recommending items based on standardized ratings - Google Patents

Abstract

A method of recommending items based on a standardized rating is disclosed. In the standardized rating-based item recommendation method according to an embodiment of the present invention, a method of recommending an item using initial evaluation information, which is information including an evaluation result value obtained by evaluating at least one of a plurality of items by each of a plurality of users. The method of claim 1, further comprising: calculating an estimated evaluation value using evaluation result values of neighboring users, the remaining users, for each of the plurality of users, based on the initial evaluation information and a collaborative filtering algorithm; Calculating standard evaluation information obtained by standardizing the initial evaluation information based on the initial evaluation information, the estimated evaluation value, and a predetermined weight; And recommending a predetermined number of items from among the plurality of items to a recommendation target person, one of the plurality of users, based on the standard evaluation information and the collaborative filtering algorithm.

Description

Item recommendation method and device based on standardized rating {METHOD AND APPARATUS FOR RECOMMENDING ITEMS BASED ON STANDARDIZED RATINGS}

The present invention relates to a method and apparatus for recommending items based on a standardized rating, and more particularly, to a method and apparatus for recommending items based on a collaborative filtering algorithm using a standardized rating by mitigating differences in rating criteria between users. .

The recommendation system is essentially used in most online shopping malls today to promote users' purchases, and is being actively studied in academia. The most representative of several approaches to the recommendation algorithm is the collaborative filtering (CF) algorithm. The collaborative filtering algorithm analyzes the user's taste and propensity for items (products) based on the user's rating, and then the ratings to be given to items that the user has not yet purchased or the top N items worth purchasing (top-N). Predict the item of.

In this case, the user's personal preference for the item is reflected in the user's rating, and the criteria for giving a rating are generally different for each user. Accordingly, a case in which the predicted rating according to the collaborative filtering algorithm does not meet the criteria for assigning a rating of the user to be recommended may occur, which may lead to a result of item recommendation with low accuracy.

Accordingly, there is a need for a method and apparatus capable of improving the accuracy of recommendation results of the top N items by performing a collaborative filtering algorithm by matching ratings given by different criteria as one criterion.

As a related prior art, there is Korean Laid-Open Patent Publication No. 10-2014-0046108 (title of the invention: product recommendation system through association analysis between products in an online shopping mall, publication date: April 18, 2014).

An object of the present invention is to provide a method and apparatus for recommending items based on a collaborative filtering algorithm using a standardized rating in which a difference in criteria for assigning ratings between users is reduced.

In addition, an object of the present invention is to provide a method and apparatus for recommending items in which the data scarcity problem is mitigated through effective data imputation on items of indifference and items of interest.

The problem to be solved by the present invention is not limited to the problem(s) mentioned above, and another problem(s) not mentioned will be clearly understood by those skilled in the art from the following description.

The standardized rating-based item recommendation method according to an embodiment of the present invention for solving the above problem uses initial evaluation information, which is information including an evaluation result value obtained by evaluating at least one of a plurality of items by each of a plurality of users. A method of recommending an item according to the method, comprising: calculating an estimated evaluation value using evaluation result values of neighboring users, which are remaining users, for each of the plurality of users, based on the initial evaluation information and a collaborative filtering algorithm; Calculating standard evaluation information obtained by standardizing the initial evaluation information based on the initial evaluation information, the estimated evaluation value, and a predetermined weight; And recommending a predetermined number of items among the plurality of items to a recommendation target person, one of the plurality of users, based on the standard evaluation information and the collaborative filtering algorithm.

Preferably, the standard evaluation information is used as the new initial evaluation information until a predetermined termination condition is satisfied, and then the steps of calculating the estimated evaluation value and calculating the standard evaluation information may be repeatedly performed. have.

Preferably, the termination condition may be a condition in which the number of repetitions is equal to a predetermined threshold number.

Preferably, prior to the step of calculating the estimated evaluation value, setting an evaluation result value of the initial evaluation information corresponding to an indifference item selected according to a predetermined criterion among the plurality of items as a predetermined indifference evaluation value It may further include.

Preferably, the step of calculating the standard evaluation information comprises setting an evaluation result value of the standard evaluation information corresponding to an item of interest selected according to a predetermined criterion among the plurality of items as a predictive evaluation value using a collaborative filtering algorithm. It may include.

Preferably, the step of setting the predictive evaluation value may be performed only when the step of calculating the standard evaluation information repeatedly performed is first performed.

Preferably, the step of calculating the estimated evaluation value may use one of a user-based collaborative filtering algorithm, an item-based collaborative filtering algorithm, and an SVD-based collaborative filtering algorithm. have.

In addition, the standardized rating-based item recommendation device according to an embodiment of the present invention for solving the above problem includes initial evaluation information, which is information including an evaluation result value obtained by evaluating at least one of a plurality of items by each of a plurality of users. An apparatus for recommending an item by using, comprising: an estimating unit for calculating an estimated evaluation value using evaluation result values of neighboring users, which are remaining users, for each of the plurality of users, based on the initial evaluation information and a collaborative filtering algorithm; A standardization unit for calculating standard evaluation information obtained by standardizing the initial evaluation information based on the initial evaluation information, the estimated evaluation value, and a predetermined weight; And a recommendation unit that recommends a predetermined number of items from among the plurality of items to a recommendation target person, one of the plurality of users, based on the standard evaluation information and the collaborative filtering algorithm.

Preferably, the standard evaluation information may be used as the new initial evaluation information until a predetermined termination condition is satisfied, and then a control unit for controlling the estimation unit and the standardization unit to be repeatedly performed.

Preferably, the termination condition may be a condition in which the number of repetitions is equal to a predetermined threshold number.

Preferably, it may further include a setting unit for setting the evaluation result value of the initial evaluation information corresponding to the indifferent item selected according to a predetermined criterion among the plurality of items as a predetermined indifferent evaluation value.

Preferably, the setting unit may further set an evaluation result value of the standard evaluation information corresponding to an item of interest selected according to a predetermined criterion among the plurality of items as a predictive evaluation value using a collaborative filtering algorithm.

Preferably, the setting unit may further set the predicted evaluation value only when the repetitively performed standardization unit is first performed.

Preferably, the estimation unit may use one of a user-based collaborative filtering algorithm, an item-based collaborative filtering algorithm, and an SVD-based collaborative filtering algorithm.

The present invention provides a method and apparatus for recommending items based on a collaborative filtering algorithm using a standardized rating that mitigates differences in criteria for assigning ratings between users, thereby improving the accuracy of item recommendation.

In addition, the present invention provides an efficient data imputation method for items of indifference and items of interest, thereby reducing the data scarcity problem that reduces the accuracy of the collaborative filtering algorithm.

1 is a flowchart illustrating a standardized rating-based item recommendation method according to an embodiment of the present invention.
2 is a flowchart illustrating a standardized rating-based item recommendation method according to another embodiment of the present invention.
3 is a block diagram illustrating an apparatus for recommending items based on a standardized rating according to an embodiment of the present invention.
4 is a block diagram illustrating an apparatus for recommending items based on a standardized rating according to another embodiment of the present invention.
5 and 6 are pseudocodes of a standardized rating-based item recommendation method according to an embodiment of the present invention.
7 is a diagram for explaining a process of performing a standardized rating-based item recommendation method according to an embodiment of the present invention.

In the present invention, various modifications may be made and various embodiments may be provided, and specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to a specific embodiment, it is to be understood to include all changes, equivalents, and substitutes included in the spirit and scope of the present invention. In describing each drawing, similar reference numerals have been used for similar elements.

Terms such as first, second, A, and B may be used to describe various elements, but the elements should not be limited by the terms. These terms are used only for the purpose of distinguishing one component from another component. For example, without departing from the scope of the present invention, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element. The term and/or includes a combination of a plurality of related listed items or any of a plurality of related listed items.

When a component is referred to as being "connected" or "connected" to another component, it is understood that it may be directly connected or connected to the other component, but other components may exist in the middle. Should be. On the other hand, when a component is referred to as being "directly connected" or "directly connected" to another component, it should be understood that there is no other component in the middle.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In the present application, terms such as "comprise" or "have" are intended to designate the presence of features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, but one or more other features. It is to be understood that the presence or addition of elements or numbers, steps, actions, components, parts, or combinations thereof, does not preclude in advance.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms as defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and should not be interpreted as an ideal or excessively formal meaning unless explicitly defined in this application. Does not.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

In the present invention, the initial evaluation information may be information including an evaluation result value obtained by evaluating at least one of a plurality of items by each of a plurality of users. In this case, each user has evaluated at least one of the plurality of items, and the type and number of items evaluated for each user may be different.

In addition, the collaborative filtering algorithm (CF) is based on the assumption that users who like the same item will have similar preferences and tastes of the item, and is an algorithm for predicting the item the user prefers. In other words, it is a technique to identify users with similar patterns in preferences and interests based on users' preferences and expressions of interest. This may be used to provide a service in the form of cross-recommending items that have not yet purchased each other to users with similar tastes, or recommending related items according to the classified user's taste or lifestyle.

1 is a flowchart illustrating a standardized rating-based item recommendation method according to an embodiment of the present invention.

In step S110, the item recommendation device calculates an estimated evaluation value for each of the plurality of users using the evaluation result values of the neighboring users, the remaining users, based on the initial evaluation information and the collaborative filtering algorithm.

That is, the item recommendation device determines the rating criteria for each of the neighboring users from the evaluation result values of the neighboring users for each of the plurality of users, and then uses the estimated evaluation value, which is the rating that each user is likely to give to each item. Can be calculated. In this case, the item recommendation device may use a collaborative filtering algorithm to calculate the estimated evaluation value.

Meanwhile, among the plurality of users, a user and a neighboring user thereof are not in a physical neighboring relationship with each other, but may have a relationship with which they can evaluate the same item.

In another embodiment, the item recommendation device may use one of a user-based collaborative filtering algorithm, an item-based collaborative filtering algorithm, and an SVD-based collaborative filtering algorithm.

For example, the item recommendation apparatus may calculate an estimated evaluation value, which is a rating that each user would like to give to each item by using an item-based k-nearest neighbor algorithm.

)을 산출할 수 있다. More specifically, referring to FIG. 5, for a user (u _s ), a set C of the 14th line is a set of neighboring users who have evaluated the k-th and m-th items (t _k , t _m ). , The PCC function in line 15 evaluates the similarity using the Pearson correlation coefficient. In addition, the item recommendation device estimates the estimated evaluation value from the similarity-based weighted average value for the user (u _s ) in the 17th line.

) Can be calculated.

Meanwhile, in FIG. 5, U is a set of users, T is a set of items, R is initial evaluation information, and R _c may be standard evaluation information reflecting a global consensus between users.

In step S120, the item recommendation apparatus calculates standard evaluation information obtained by standardizing the initial evaluation information based on the initial evaluation information, the estimated evaluation value, and a predetermined weight.

That is, the item recommendation device adjusts the initial evaluation information, including the evaluation result values originally assigned by the user to the item for each of a plurality of users, using the estimated evaluation value reflecting the evaluation criteria of neighboring users and the weight, thereby initial evaluation. Standard evaluation information that standardizes information can be calculated.

In this case, the weight is a value that reflects the reliability of each of the evaluation result value and the estimated evaluation value included in the initial evaluation information, and may be determined through an experiment to confirm the accuracy of recommendation of top-N items with respect to predetermined data.

), 초기평가정보(R_i[s,k]) 및 가중치(ω)를 이용하여 표준평가정보(R_{i +1}[s,k])를 산출할 수 있다. 즉, 표준평가정보는 아래의 수학식 1을 통해 산출될 수 있다.More specifically, referring to FIG. 5, the item recommending device is an estimated evaluation value (

), the standard evaluation information (R _{i +1} [s,k]) can be calculated using the initial evaluation information (R _i [s,k]) and the weight (ω). That is, the standard evaluation information may be calculated through Equation 1 below.

[Equation 1]

At this time, R _{i + 1} [s,k] is the standard evaluation information on the item t _k of the user u _s , w is the weight, and R _i [s,k] is the initial evaluation on the item t _k of the user u _s Information.

In this way, the item recommendation device can adjust the evaluation result value of a specific user using the evaluation result value of neighboring users, and this is viewed as a process of standardizing the criteria for giving evaluation results through continuous consensus among users. I can.

Finally, in step S130, the item recommendation device recommends a predetermined number of items from among a plurality of items to a recommendation target person, one of a plurality of users, based on the standard evaluation information and a collaborative filtering algorithm.

For example, the item recommendation apparatus may recommend top-N items to a recommendation target using standardized standard evaluation information and a collaborative filtering algorithm. In this case, the collaborative filtering algorithm in step S130 is not necessarily the same as the collaborative filtering algorithm used in step S110.

As described above, the standardized rating-based item recommendation method according to an embodiment of the present invention provides an item recommendation method based on a collaborative filtering algorithm using a standardized rating that mitigates the difference in rating criteria between users. There is an effect that can improve accuracy.

2 is a flowchart illustrating a standardized rating-based item recommendation method according to another embodiment of the present invention.

In step S210, the item recommendation device sets an evaluation result value of initial evaluation information corresponding to an indifferent item selected according to a predetermined criterion among a plurality of items as a predetermined indifferent evaluation value.

Here, when the degree of preference for a user's item before use is referred to as a pre-use preference, there are various methods for inferring such a prior preference. That is, regardless of the assigned evaluation result value, it can be assumed that the item for which the evaluation result exists once was selected (used) because the user's preference for the item was very high before use. Accordingly, the item recommendation device may construct a prior preference matrix in which 1 is substituted for an item having an existing evaluation result value, and then infer the prior preference from the viewpoint of OCCF (Single Class Collaborative Filtering).

More specifically, from the perspective of OCCF, the item recommendation device may determine that the item that the user who has used many items and given the evaluation result has not yet used and has not evaluated is an item with a low prior preference of the user. . Therefore, the user-oriented weighting scheme can be predicted through the wALS technique after appropriate weights are assigned to all items in such a way that higher weights are assigned to these items. .

In this case, the item recommendation device may select an item having a relatively lower prior preference, that is, an item belonging to the lower θ _x % range of the prior preference criterion as an indifferent item. On the other hand, the parameter θ _x value used for selecting an indifferent item may be numerically determined through an experiment to confirm the accuracy of recommending top-N items.

In addition, the item recommendation device has a relatively low value of indifference evaluation value (a relatively low value of indifference evaluation value) because the item recommendation device is an item that is highly likely to receive poor evaluation (low preference after use) from the user even if it is used, For example, a rating of 0) can be substituted (zero-injection) to be excluded from the nomination candidate. In this case, a rating of 0 on a 1-5 rating scale may mean the lowest preference that a user may have for an item. As described above, the standardization method of the present invention can alleviate the scarcity problem by applying the zero-injection method of the existing research during the standardization process.

In step S220, the item recommendation device calculates an estimated evaluation value using the evaluation result values of the neighboring users, the remaining users, for each of the plurality of users, based on the initial evaluation information and the collaborative filtering algorithm.

In this case, the item recommendation device may calculate an estimated evaluation value using initial evaluation information in which an indifferent evaluation value is set.

In step S230, the item recommendation apparatus calculates standard evaluation information obtained by standardizing the initial evaluation information based on the initial evaluation information, the estimated evaluation value, and a predetermined weight.

In another embodiment, the item recommendation apparatus may set an evaluation result value of standard evaluation information corresponding to an item of interest selected according to a predetermined criterion among a plurality of items as a predictive evaluation value using a collaborative filtering algorithm.

That is, the item recommendation device may select an item having a relatively higher prior preference, that is, an item belonging to the upper θ _y % range based on the prior preference as the item of interest. In this case, like the item of interest, the parameter θ _y value used for selecting the item of interest may be numerically determined through an experiment to confirm the accuracy of recommending top-N items.

On the other hand, among the items for which the evaluation result does not exist, an item that does not belong to either item of interest or item of interest, that is, does not fall within the range of the lower θ _x % and the upper θ _y % based on the prior preference may be referred to as a neutral item.

As described above, in the process of calculating the standard evaluation information, the present invention can further alleviate the data scarcity problem by performing an appropriate imputation (replacement of missing values) for the item of interest as well as zero-injection for the item of interest.

In addition, according to the existing research, since the item in which the evaluation result from the user already exists is part of the item of interest, the evaluation result value to be newly assigned to the item of interest (there is no evaluation result) is It should be similar to the rating value given to the item. However, since the previously assigned evaluation result has various values (eg, 1 to 5), the evaluation result to be assigned to the item of interest cannot be composed of only one type of value, such as 0, unlike the item of interest.

At this time, existing studies (AutAI, AdaM, etc.) exist as a method of substituting the user's predicted evaluation results for items that the user has not yet evaluated. Although there are some differences in the process of determining the key set that will be the target of the implantation, both methods have a common point in that the predictive evaluation result calculated by collaborative filtering is substituted into the key set.

That is, the item recommendation device may use the predictive evaluation result calculation method by collaborative filtering introduced in the existing research in order to calculate the evaluation result value (ie, predictive evaluation value) to be substituted for the item of interest. However, unlike previous studies, in order to obtain a more accurate predictive evaluation result, an indifference evaluation value (eg, 0) substituted for an indifferent item can be used together in the calculation process.

As described above, the item evaluation apparatus performs data imputation in the process of collecting the ratings of neighboring users for both the item of interest and the item of no interest, thereby mitigating the data scarcity problem.

More specifically, the indifference evaluation value assigned to the item of interest alleviates the data scarcity problem in the process where the predicted evaluation value to be assigned to the item of interest is calculated by collaborative filtering, and the predicted evaluation value assigned to the item of interest is, In the process of recommending top-N items after standardization of ratings is completed, data scarcity problem can be alleviated.

In another embodiment, the item recommendation apparatus may set the evaluation result value of the initial evaluation information corresponding to the item of interest as the predicted evaluation value only when step S230 is first performed.

That is, the item recommendation device may initially set the evaluation result value of the standard evaluation information as the predicted evaluation value only once, and may not set the evaluation result value of the standard evaluation information after that. This is because when step S230 is repeatedly performed according to the termination condition described in step S240, the standard evaluation information to be standardized may be distorted by repeatedly resetting the evaluation result value of the standard evaluation information to the predicted evaluation value. to be.

In this case, referring to FIG. 5, it can be seen that data imputation for setting a prediction evaluation value is performed only when i is 0 (ie, when first performed) in the 21st line. In this case, the DATA_IMPUTATE function in line 22 may correspond to an imputation that sets the evaluation result value of the standard evaluation information described above as a predictive evaluation value.

In step S240, the item recommendation device determines whether a predetermined termination condition is satisfied.

That is, the item recommendation device may perform step S250 when the end condition is satisfied, and may perform step S220 if the end condition is not satisfied.

In this case, when the item recommendation device performs step S220, the standard evaluation information may be set as new initial evaluation information and step S220 may be performed.

In another embodiment, the termination condition may be a condition in which the number of repetitions is equal to a predetermined threshold number.

That is, referring to FIG. 5, the item recommendation apparatus may repeatedly perform steps S220 and S230 by a predetermined threshold number of nums. In this case, the critical number of times required for global consensus between neighboring users may be determined through an experiment or the like.

Finally, in step S250, the item recommendation device recommends a predetermined number of items from among a plurality of items to a recommendation target person, one of a plurality of users, based on the standard evaluation information and a collaborative filtering algorithm.

In this case, referring to FIG. 5, the item recommendation device repeatedly performs steps S220 and S230 by a predetermined threshold number of nums, and updates standard evaluation information R _{i+1 for} each execution. In addition, referring to FIG. 6, the item recommendation device maximizes the recommendation accuracy of top-N items (ie, F1_SCORE function value) among a plurality of standard evaluation information updated for each execution of the standard evaluation information and collaborative filtering. A predetermined number of items can be recommended to a recommendation target using an algorithm.

On the other hand, referring to FIG. 7, the item recommendation device substitutes the indifference evaluation value (Z) to the indifference item from the initial evaluation information (R), and then generates the first standard evaluation information (R ₁ ), and predicts evaluation in the item of interest Value (P) can be substituted. And, it is possible to generate standard evaluation information (R _c ) for which the global agreement is completed through repetition of nums times.

As described above, the standardized rating-based item recommendation method according to another embodiment of the present invention provides an efficient data implantation method for indifferent items and items of interest, thereby reducing the data scarcity problem. .

3 is a block diagram illustrating an apparatus for recommending items based on a standardized rating according to an embodiment of the present invention.

Referring to FIG. 3, the standardized rating-based item recommendation device 300 according to an embodiment of the present invention may include an estimation unit 310, a standardization unit 320, and a recommendation unit 330.

Meanwhile, the item recommendation device 300 may be mounted and used in a computing device such as a desktop computer, a notebook computer, a server computer, a smart phone, and a tablet PC.

The estimating unit 310 calculates an estimated evaluation value using evaluation result values of neighboring users, the remaining users, for each of the plurality of users, based on the initial evaluation information and the collaborative filtering algorithm.

In another embodiment, the estimating unit 310 may use one of a user-based collaborative filtering algorithm, an item-based collaborative filtering algorithm, and an SVD-based collaborative filtering algorithm.

The standardization unit 320 calculates standard evaluation information obtained by standardizing the initial evaluation information based on the initial evaluation information, the estimated evaluation value, and a predetermined weight.

The recommendation unit 330 recommends a predetermined number of items from among a plurality of items to a recommendation target person, who is one of a plurality of users, based on standard evaluation information and a collaborative filtering algorithm.

4 is a block diagram illustrating an item recommendation device based on a standardized rating according to another embodiment of the present invention.

Referring to FIG. 4, a standardized rating-based item recommendation device 300 according to an embodiment of the present invention includes an estimation unit 310, a standardization unit 320, a recommendation unit 330, and a control unit 340. can do. In addition, it may optionally further include a setting unit (not shown).

The estimating unit 310 calculates an estimated evaluation value using evaluation result values of neighboring users, the remaining users, for each of the plurality of users, based on the initial evaluation information and the collaborative filtering algorithm.

The standardization unit 320 calculates standard evaluation information obtained by standardizing the initial evaluation information based on the initial evaluation information, the estimated evaluation value, and a predetermined weight.

The recommendation unit 330 recommends a predetermined number of items from among a plurality of items to a recommendation target person, who is one of a plurality of users, based on standard evaluation information and a collaborative filtering algorithm.

The controller 340 uses the standard evaluation information as new initial evaluation information until a predetermined termination condition is satisfied, and then controls the estimation unit 310 and the standardization unit 320 to be repeatedly performed.

In another embodiment, the termination condition may be a condition in which the number of repetitions of the control of the estimation unit 310 and the standardization unit 320 is equal to a predetermined threshold number.

The setting unit (not shown) sets an evaluation result value of initial evaluation information corresponding to an indifferent item selected according to a predetermined criterion among a plurality of items as a predetermined indifferent evaluation value.

In another embodiment, the setting unit (not shown) may further set an evaluation result value of standard evaluation information corresponding to an item of interest selected according to a predetermined criterion among a plurality of items as a predictive evaluation value using a collaborative filtering algorithm.

In another embodiment, the setting unit (not shown) may further set the predicted evaluation value only when the repetitively performed standardization unit 320 is first performed.

Meanwhile, the above-described embodiments of the present invention can be written as a program that can be executed in a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium.

The computer-readable recording medium includes a magnetic storage medium (for example, a ROM, a floppy disk, a hard disk, etc.), and an optical reading medium (for example, a CD-ROM, a DVD, etc.).

So far, the present invention has been looked at around its preferred embodiments. Those of ordinary skill in the art to which the present invention pertains will be able to understand that the present invention can be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered from an illustrative point of view rather than a limiting point of view. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope equivalent thereto should be construed as being included in the present invention.

Claims (14)

In a method for recommending an item using initial evaluation information, which is information including an evaluation result value obtained by evaluating at least one of a plurality of items by each of a plurality of users,
Setting an indifferent evaluation value to an evaluation result value of the initial evaluation information corresponding to an indifferent item among the plurality of items;
Calculating an estimated evaluation value for each of the plurality of users, based on the indifference evaluation value, initial evaluation information, and a collaborative filtering algorithm, using evaluation result values of neighboring users, which are other users;
Calculating standard evaluation information obtained by standardizing the initial evaluation information based on the initial evaluation information, the estimated evaluation value, and a predetermined weight; And
And recommending a predetermined number of items from among the plurality of items to a recommendation target person, one of the plurality of users, based on the standard evaluation information and a collaborative filtering algorithm,
The indifference evaluation value is set based on prior preference according to one class collaborative filtering (OCCF),
A user who has used and evaluated the number of items equal to or greater than a preset first threshold value is set to be less than a preset second threshold value,
The standardized rating-based item recommendation method, characterized in that the indifference evaluation value for the indifference item in which a preset preference is less than a third threshold value is set to 0. The method of claim 1,
Standardization, characterized in that, until a predetermined termination condition is satisfied, the standard evaluation information is used as the new initial evaluation information, and then the steps of calculating the estimated evaluation value and calculating the standard evaluation information are repeatedly performed. Rating-based item recommendation method. The method of claim 2,
The above termination conditions are
The standardized rating-based item recommendation method, characterized in that a condition in which the number of repetitions is equal to a predetermined threshold number. delete delete The method of claim 2,
The step of calculating the standard evaluation information
The standardized rating-based item recommendation method, characterized in that it is performed only when the step of calculating the standard evaluation information that is repeatedly performed is performed for the first time. The method of claim 1,
The step of calculating the estimated evaluation value
A standardized rating-based item recommendation method, characterized by using one of a user-based collaborative filtering algorithm, an item-based collaborative filtering algorithm, and an SVD-based collaborative filtering algorithm. In an apparatus for recommending an item using initial evaluation information, which is information including an evaluation result value obtained by each of a plurality of users evaluating at least one of a plurality of items,
A setting unit for setting an evaluation result value of the initial evaluation information corresponding to an indifferent item among the plurality of items as an indifferent evaluation value;
An estimating unit for calculating an estimated evaluation value using evaluation result values of neighboring users, the remaining users, for each of the plurality of users, based on the indifference evaluation value, the initial evaluation information, and a collaborative filtering algorithm;
A standardization unit for calculating standard evaluation information obtained by standardizing the initial evaluation information based on the initial evaluation information, the estimated evaluation value, and a predetermined weight; And
And a recommendation unit for recommending a predetermined number of items among the plurality of items to a recommendation target person, one of the plurality of users, based on the standard evaluation information and a collaborative filtering algorithm,
The indifference evaluation value is set based on prior preference according to one class collaborative filtering (OCCF),
The setting unit uses the number of items equal to or greater than a preset first threshold, and sets a prior preference of an item not used and not evaluated by a user who evaluated it to be less than a preset second threshold, and the preset preference is less than a third threshold. The standardized rating-based item recommendation device, characterized in that the indifference evaluation value for the indifference item is set to 0. The method of claim 8,
A control unit controlling the estimation unit and the standardization unit to be repeatedly performed after the standard evaluation information is used as the new initial evaluation information until a predetermined termination condition is satisfied.
Item recommendation device based on a standardized rating, characterized in that it further comprises. The method of claim 9,
The above termination conditions are
The standardized rating-based item recommendation device, characterized in that the condition is a condition in which the number of repetitions is equal to a predetermined threshold number. delete delete The method of claim 9,
The setting unit
The standardized rating-based item recommendation device, characterized in that the repetitively performed standardization unit is further set as a predictive evaluation value only when the first is performed. The method of claim 8,
The estimation unit
A standardized rating-based item recommendation device, characterized by using one of a user-based collaborative filtering algorithm, an item-based collaborative filtering algorithm, and an SVD-based collaborative filtering algorithm.

Images