KR102266853B1 - Method and apparatus for recommending items based on multi-type pair-wise preference - Google Patents

Abstract

A multi-type pairwise-based item recommendation method is disclosed. A method of recommending an item using an evaluation result that is a result of evaluation of at least one item among a plurality of items by a plurality of users according to an embodiment of the present invention is a method of recommending an item by using a collaborative filtering technique. generating a comprehensive result by synthesizing a prediction result that is a result of predicting an evaluation result of an unrated item and the evaluation result; classifying the plurality of unrated items into a plurality of uninformed items and a plurality of indifferent items in the comprehensive result; learning model parameters based on the comprehensive result and a plurality of selected item pairs selected from among item pairs consisting of two of an evaluation item, an unevaluated item, an uninformed item, and an indifferent item included in the comprehensive result; and recommending a predetermined number of items to at least one recommendation target among the plurality of users by using a final prediction result that is a result of predicting the evaluation results for the plurality of unrated items based on the learned model parameter. includes

Description

METHOD AND APPARATUS FOR RECOMMENDING ITEMS BASED ON MULTI-TYPE PAIR-WISE PREFERENCE

The present invention relates to a method and apparatus for classifying an item to be evaluated into an evaluated item, an unrated item, an uninformed item, and an indifferent item, and recommending an item to a user using multiple types of item pairs based thereon will be.

Collaborative filtering (CF) is a technology that recommends the top N items to users by using past user actions such as explicit evaluation results such as ratings and implicit evaluation results such as clicks. It is a technology that has already been applied to the system.

Recently, as the demand for processing datasets related to single-class settings increases, a one-class collaborative filtering (OCCF) technique has been proposed. The OCCF technique can be divided into two types according to the method of learning the user's preference (ie, the evaluation result).

First, it is a method of minimizing the error between the actual evaluation result and the predicted evaluation result of individual users in units of individual items (point-wise). Second, it is a method of maximizing a difference between an evaluation result of an item evaluated in an item pair of an individual user and an evaluation result of a predicted item on a per-item pair-wise basis.

Here, the second method is called a single-type pair-wise approach, and it is evaluated that it shows superior performance in predicting the evaluation result of an item than the first method.

However, the single-type pairwise-based approach assumes that the user has the same degree of negative preference for all unrated items (P1) and assumes that the user prefers rated items over all unrated items (P2). is based on However, since such assumptions do not always correspond to reality, it is true that there is a possibility of causing errors in item recommendation.

Accordingly, there is a need for a method and apparatus for recommending items based on a multi-type pairwise to solve this problem.

R. Steffen, F. Christoph, G. Zeno, and S.-T. Lars. 2009. BPR: Bayesian Personalized Ranking from Implicit Feedback. In Proc. of the 25th Int'l Conf. on Uncertainty in Artificial Intelligence (UAI). 452-461.

The present invention classifies an item into an evaluated item, an unrated item, an uninformed item, and an indifferent item, and recommends an item to a user with higher accuracy by using a model parameter learned based on multiple types of item pairs. A method and apparatus are provided.

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.

In a multi-type pairwise-based item recommendation method according to an embodiment of the present invention for achieving the above object, a method of recommending an item using an evaluation result that is a result of evaluating at least one item among a plurality of items by a plurality of users In the method, generating a comprehensive result by synthesizing a prediction result that is a result of predicting an evaluation result of a plurality of unrated items included in the evaluation result using a collaborative filtering technique and the evaluation result; classifying the plurality of unrated items into a plurality of uninformed items and a plurality of indifferent items in the comprehensive result; learning model parameters based on the comprehensive result and a plurality of selected item pairs selected from among item pairs consisting of two of an evaluation item, an unevaluated item, an uninformed item, and an indifferent item included in the comprehensive result; and recommending a predetermined number of items to at least one recommendation target among the plurality of users by using a final prediction result that is a result of predicting the evaluation results for the plurality of unrated items based on the learned model parameter. includes

Preferably, the learning of the model parameter comprises: generating an objective function based on a difference in evaluation results between the types of items included in each of the plurality of selected item pairs; obtaining a plurality of evaluation result information including evaluation results for each item type included in the plurality of selected item pairs from the comprehensive result for one selected user selected from among the plurality of users; updating the model parameter value so that the value of the objective function is maximized by using the objective function and the plurality of evaluation result information; and repeating the steps of obtaining the plurality of evaluation result information and updating the model parameter values until a predetermined completion condition is satisfied.

Preferably, the updating of the model parameter value comprises: calculating a gradient value of the objective function by using the plurality of evaluation result information; and updating the model parameter value by applying the gradient value based on a stochastic gradient descent (SGD) technique.

Preferably, the plurality of evaluation result information may include one evaluation result for each item type included in the plurality of selected item pairs.

Preferably, the plurality of evaluation result information includes a group evaluation result that is a plurality of evaluation results for each item type included in the plurality of selected item pairs, and the objective function may be based on a difference between the group evaluation results. .

Preferably, the objective function is that the evaluation result of the evaluation item is higher than the evaluation result of the unrated item, the non-information item, and the indifferent item, and the evaluation result of the indifferent item is higher than the evaluation result of the indifferent item It can be home based.

Preferably, the plurality of selected item pairs include an item pair composed of the evaluation item and the unrated item and an item pair composed of the non-information item and the indifferent item, or the evaluation item and the non-information item It may include an item pair composed of , an item pair composed of the evaluation item and the indifferent item, and an item pair composed of the non-information item and the indifferent item.

In addition, the multi-type pairwise-based item recommendation apparatus according to an embodiment of the present invention for achieving the above object recommends an item using an evaluation result that is a result of evaluating at least one item among a plurality of items by a plurality of users What is claimed is: 1. An apparatus comprising: a generator for generating a comprehensive result by synthesizing a prediction result, which is a result of predicting an evaluation result of a plurality of unevaluated items included in the evaluation result, and the evaluation result using a collaborative filtering technique; a classification unit for classifying the plurality of unrated items into a plurality of uninformed items and a plurality of indifferent items in the comprehensive result; a learning unit for learning model parameters based on the comprehensive result and a plurality of selected item pairs selected from among item pairs consisting of two of an evaluation item, an unevaluated item, an uninformed item, and an indifferent item included in the comprehensive result; and a recommendation for recommending a predetermined number of items to at least one recommendation target among the plurality of users by using a final prediction result that is a result of predicting the evaluation result for the plurality of unrated items based on the learned model parameter. includes wealth.

Preferably, the learning unit generates an objective function based on a difference in evaluation results between the types of items included in each of the plurality of selected item pairs, and for one selected user selected from among the plurality of users, the Obtain a plurality of evaluation result information including evaluation results for each item type included in a plurality of selected item pairs, and use the objective function and the plurality of evaluation result information to maximize the value of the objective function. may be updated, and the process of acquiring the plurality of evaluation result information and the process of updating the model parameter values may be repeated until a predetermined completion condition is satisfied.

Preferably, when the learning unit updates the model parameter value, the gradient value of the objective function is calculated using the plurality of evaluation result information, and the gradient value is calculated based on a stochastic gradient descent (SGD) technique. By applying, it is possible to update the model parameter values.

Preferably, the plurality of evaluation result information may include one evaluation result for each item type included in the plurality of selected item pairs.

Preferably, the plurality of evaluation result information includes a group evaluation result that is a plurality of evaluation results for each item type included in the plurality of selected item pairs, and the objective function may be based on a difference between the group evaluation results. .

Preferably, the objective function is that the evaluation result of the evaluation item is higher than the evaluation result of the unrated item, the non-information item, and the indifferent item, and the evaluation result of the indifferent item is higher than the evaluation result of the indifferent item It can be home based.

Preferably, the plurality of selected item pairs include an item pair composed of the evaluation item and the unrated item and an item pair composed of the non-information item and the indifferent item, or the evaluation item and the non-information item It may include an item pair composed of , an item pair composed of the evaluation item and the indifferent item, and an item pair composed of the non-information item and the indifferent item.

The present invention classifies an item to be evaluated into an evaluated item, an unrated item, an uninformed item, and an indifferent item, and uses a model parameter learned based on multiple types of item pairs to provide users with higher accuracy. It has the effect of recommending an item.

In addition, the present invention has the effect of being able to recommend an item to a user with higher accuracy by grouping and using the evaluation results for the same type of item when learning model parameters using multiple types of item pairs.

1 is a diagram for explaining (a) single-type pairwise preference and (b) multi-type pairwise preference.
2 is a flowchart illustrating a method for recommending a multi-type pairwise based item according to an embodiment of the present invention.
3 is a flowchart illustrating a method of learning model parameters according to an embodiment of the present invention.
4 is a block diagram illustrating a multi-type pairwise-based item recommendation apparatus according to an embodiment of the present invention.
5 is a diagram illustrating pseudocode of a model parameter learning algorithm according to an embodiment of the present invention.
6 and 7 are block diagrams for explaining the effect of a multi-type pairwise-based item recommendation method according to an embodiment of the present invention.

Since the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention. In describing each figure, like reference numerals have been used for like 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. The above terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component. 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 the other component may be directly connected or connected to the other component, but other components may exist in between. it should be On the other hand, when it is mentioned that a certain element is "directly connected" or "directly connected" to another element, it should be understood that the other element does not exist 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. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, but one or more other features It should be understood that this does not preclude the existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

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

In the present invention, when an evaluation result (eg, a matrix) that is a result of evaluating at least one of a plurality of items to be evaluated by a plurality of users exists, the item evaluated by a certain user in the evaluation result is a rated item , and an item not evaluated by the user is defined as an unrated item. Here, among unrated items, an item whose user's evaluation result is expected to be low is defined as an uninteresting item, and the rest of the unrated items, excluding the uninterested item, are items whose preference is difficult to predict. item) is defined.

Meanwhile, the present invention can be mainly applied to a single class environment based on implicit feedback (eg, click). However, the present invention is not limited thereto, and may be applied to a multi-class environment based on explicit feedback (eg, rating) through appropriate data transformation and processing.

1 is a diagram for explaining (a) single-type pairwise preference and (b) multi-type pairwise preference.

First, referring to FIG. 1A , a single type pairwise preference according to a conventional technique (ie, Bayesian Personalized Ranking, BPR) is shown. At this time, in the prior art, a single type of pair is composed of an evaluation item and an unrated item, and under the premise (PP) that the evaluation result on the evaluation item is higher than the evaluation result on the unrated item, model parameters to maximize the difference can be learned

Also, referring to FIG. 1(b) , a multi-type pairwise preference according to the present invention is shown. At this time, in the present invention, not only the pair of the evaluated item and the unrated item, but also the pair of the evaluated item and the non-information item, the pair of the evaluation item and the indifferent item, the pair of the non-information item and the indifferent item, a total of four multi-type pairs appear. have. In addition, in the present invention, the premise (PP1, PP2, PP3) that the evaluation result on the evaluation item is higher than the evaluation result on the unrated item, the non-information item, and the indifferent item, and the evaluation result on the non-information item is the evaluation of the indifferent item Under the premise (PP4) higher than the result, the model parameters can be trained to maximize the difference. In this case, i ₁ , i ₂ , and i ₃ are evaluation items, k ₁ , k ₂ are non-information items, and l ₁ , l ₂ , l ₃ , and l ₄ are indifferent items.

2 is a flowchart illustrating a method for recommending a multi-type pairwise based item according to an embodiment of the present invention.

In step S210, the item recommendation device generates a comprehensive result by synthesizing the prediction result, which is a result of predicting the evaluation result of a plurality of unrated items included in the evaluation result, using the collaborative filtering technique, and the evaluation result.

In this case, the item recommendation apparatus may roughly predict the evaluation result of the user for the unrated item by using the collaborative filtering technique. Preferably, the item recommendation apparatus may use a weighted regularized matrix factorization (WRMF) technique. On the other hand, WRMF is a well-known OCCF technique, and can indicate unvalued items included in the prediction result as 1 (positive) or 0 (negative).

Also, the item recommendation apparatus may generate a comprehensive result by synthesizing the evaluation result and the prediction result. For example, the comprehensive result is a combination of the evaluation result of the evaluation item included in the evaluation result and the evaluation result of the unrated item included in the prediction result, and may be in the form of a matrix.

In step S220, the item recommendation apparatus classifies the plurality of unrated items into uninformed items and indifferent items from the overall result.

For example, the item recommendation apparatus may classify unrated items of a predetermined ratio having a low evaluation result predicted through the collaborative filtering technique as indifferent items. In addition, the item recommendation apparatus may classify the remainder of the plurality of unrated items except for the plurality of indifferent items into the plurality of uninformed items.

In step S230, the item recommendation device based on the comprehensive result and the plurality of selected item pairs selected from among the item pairs consisting of two of an evaluation item, an unrated item, an uninformed item, and an indifferent item included in the comprehensive result, Train model parameters.

That is, the item recommendation apparatus may learn the model parameters used in the item recommendation method by using the synthesis result and a plurality of selected item pairs.

At this time, the item pair is composed of two items selected from an evaluation item, an unrated item, an uninformed item, and an indifferent item. For example, the item pair may be {evaluated item, unrated item}, {evaluated item, uninformed item}, {evaluated item, indifferent item}, {uninformative item, indifferent item}.

Meanwhile, details of how the item recommendation apparatus learns the model parameters will be described later in detail with reference to FIG. 3 .

In another embodiment, the plurality of selected item pairs may include an item pair composed of an evaluated item and an unrated item, and an item pair composed of an uninformed item and an indifferent item.

In addition, the plurality of selected item pairs may include an item pair composed of an evaluation item and an indifferent item, an item pair composed of an evaluation item and an indifferent item, and an item pair composed of an uninformation item and an indifferent item.

That is, the item recommendation apparatus may select a plurality of selected item pairs in order to more efficiently and accurately recommend an item to a user.

To this end, the plurality of selected item pairs may include {evaluated item, unrated item} and {uninformed item, indifferent item}. In addition, the plurality of selection item pairs may include {evaluation item, non-information item}, {evaluation item, indifferent item}, and {no information item, indifferent item}.

Finally, in step S240, using the final prediction result, which is the result of predicting the evaluation result for the plurality of unrated items based on the learned model parameter, the item recommendation device is given to at least one recommendation target among the plurality of users. A number of items are recommended.

That is, the item recommendation apparatus may generate a final prediction result by predicting evaluation results for a plurality of unrated items using the learned model parameter. The final prediction result is a result predicted more accurately than the above-mentioned prediction result, and the item recommendation apparatus can recommend the top predetermined number of items to the recommendation target by using the final prediction result.

As such, the multi-type pairwise-based item recommendation method according to an embodiment of the present invention classifies a plurality of items into an evaluated item, an unrated item, an uninformed item, and an indifferent item, and is based on multiple types of item pairs. By using the learned model parameters, there is an effect that items can be recommended with higher accuracy to the user.

3 is a flowchart illustrating a method of learning model parameters according to an embodiment of the present invention.

In step S310, the item recommendation apparatus generates an objective function based on a difference in evaluation results between the types of items included in each of the plurality of selected item pairs.

For example, it is assumed that the plurality of selected item pairs include {evaluated item, unrated item} and {uninformation item, indifferent item}. In this case, the item recommendation apparatus may generate an objective function representing the sum of the difference between the evaluation results of the evaluated item and the unrated item and the difference between the evaluation results of the non-information item and the indifferent item for a plurality of users.

In another embodiment, the objective function may be based on the assumption that the evaluation result of the evaluation item is higher than the evaluation result of the unrated item, the uninformed item, and the indifferent item, and the evaluation result of the uninformed item is higher than the evaluation result of the indifferent item. .

At this time, referring to Fig. 1(b), the assumption that the evaluation result of the evaluation item is higher than the evaluation result of the unrated item corresponds to PP1, and the assumption that the evaluation result of the evaluation item is higher than the evaluation result of the non-information item is in PP2 and the assumption that the evaluation result of the evaluation item is higher than the evaluation result of the indifferent item corresponds to PP3, and the assumption that the evaluation result of the indifferent item is higher than the evaluation result of the indifferent item corresponds to PP4.

That is, the item recommendation apparatus may learn the model parameters so that the difference in the objective function is maximized based on the above-mentioned assumption.

In step S320, the item recommendation apparatus obtains a plurality of evaluation result information including evaluation results for each item type included in the plurality of selected item pairs from the comprehensive result with respect to one selected user selected from among the plurality of users.

In this case, it may be preferable that the item recommendation apparatus randomly selects one selection user from among the plurality of users. In addition, the item recommendation apparatus may obtain a plurality of evaluation result information by randomly extracting one or a plurality of evaluation results for each item type included in the plurality of selected item pairs from the evaluation result of the selected user included in the comprehensive result. .

For example, it is assumed that the plurality of selected item pairs include {evaluated item, unrated item} and {uninformation item, indifferent item}. In this case, the item recommendation apparatus may obtain a plurality of evaluation result information by extracting evaluation results for an evaluation item, an unrated item, an uninformed item, and an indifferent item from among the evaluation results of the selected user. Also, the evaluation result for the unrated item may be the evaluation result for an item randomly selected from among uninformed items and indifferent items.

In another embodiment, the plurality of evaluation result information may include one evaluation result for each item type included in the plurality of selected item pairs.

That is, the item recommendation apparatus may calculate an objective function value by using the evaluation results of one item for each type of item for the selected user, and include each evaluation result in order to learn the model parameter.

This is a change from the existing Bayesian Personalized Ranking (BPR) based on multiple types of pairwise evaluation results for individual items, and in the present invention, it is called M-BPR (Multi-type pair-wise preference BPR).

In another embodiment, the plurality of evaluation result information may include a group evaluation result that is a plurality of evaluation results for each item type included in the plurality of selected item pairs, and the objective function may be based on a difference between the group evaluation results.

That is, the item recommendation device calculates an objective function value by using the evaluation results (ie, group evaluation results) of a plurality of items for each type of item for the selected user, and uses one group evaluation result to learn the model parameters. can be included In this case, the objective function may be generated based on the group evaluation result difference, not the evaluation result difference of individual items. In addition, the group evaluation result may be calculated using an average value of evaluation results of the plurality of items.

This is a modification of the M-BPR of the present invention described above to be based on a group unit item, and is referred to as M-BPR ^G in the present invention.

In step S330, the item recommendation apparatus uses the objective function and the plurality of evaluation result information to update the model parameter value so that the value of the objective function is maximized.

For example, the item recommendation apparatus may determine a model parameter value that maximizes a difference (ie, an objective function) of an evaluation result between item types by using a plurality of evaluation result information. In this case, the item recommendation apparatus may additionally use a method such as gradient descent.

In another embodiment, after the item recommendation apparatus calculates a gradient value of the objective function using a plurality of evaluation result information, the model parameter value may be updated by applying the gradient value based on the SGD technique.

First, the item recommendation apparatus calculates the gradient value of the objective function by using a plurality of evaluation result information.

For example, the item recommendation apparatus may calculate a gradient value by inputting a plurality of evaluation result information to the gradient of the objective function.

Then, the item recommendation apparatus updates the model parameter value by applying the gradient value based on the stochastic gradient descent (SGD) technique.

That is, the item recommendation apparatus may update the existing model parameter value with the newly calculated model parameter value by applying the gradient value calculated based on the SGD technique.

Finally, in step S340, steps S320 and S330 are repeated until the item recommendation apparatus satisfies a predetermined completion condition.

In this case, the completion condition may be a condition of repeating a predetermined number of times. Alternatively, the completion condition may be a condition in which the updated model parameter value is compared with the existing model parameter value and converges within a predetermined threshold range.

That is, the item recommendation apparatus may continuously update the model parameter value by repeating steps S320 and S330 until the completion condition is satisfied.

Meanwhile, in the M-BPR of the present invention, the objective function in the case where a plurality of selected item pairs include {evaluated item, unrated item} and {uninformation item, indifferent item} is as shown in Equation 1. At this time, referring to FIG. 1(b), Equation 1 is an objective function in the case of using PP1 and PP4.

[Equation 1]

Here, f(θ) is the objective function, θ is the model parameter, R(θ) is the normalization term of the model parameter, U is the set of users, I _u ^rt is the set of evaluation items of user u, and I\I _u ^{rt is the set of unrated} items of user u, I _u unk is the set of ^uninformed _{items of user u, I u} uin is the set of indifferent items of user u, w is the weight, σ is the sigmoid function, and r _uij is r -r _{ui uj, uk} r _ukl is r -r _ul, the evaluation results for _ui r, r _{uj, uk} r, r _ul item eunneun user u i, j, k, l .

In addition, in the M-BPR of the present invention, the objective function when a plurality of selected item pairs include {evaluation item, no information item}, {evaluation item, indifferent item}, {no information item, indifferent item} is the equation Same as 2. At this time, referring to FIG. 1(b), Equation 2 is an objective function in the case of using PP2, PP3 and PP4.

[Equation 2]

Here, α, β and γ are confidence coefficients for controlling the relative learning weights for PP2, PP4, and PP4, respectively.

In addition, each of Equations 3 and 4 below is the result of optimization by applying the SGD algorithm to Equations 1 and 2.

[Equation 3]

[Equation 4]

Here, γ is the learning rate, λθ is a parameter related to a normalized term, and the model parameter θ includes Uu·, a latent factor of user u, Vi·, a latent factor of item i, and bias bi· of item i.

In addition, the following relation as in Equation 5 is additionally established.

[Equation 5]

In addition, in the M-BPR ^G of the present invention, the objective function when a plurality of selected item pairs include {evaluated item, unrated item} and {uninformation item, indifferent item} is expressed as Equation (6). At this time, referring to FIG. 1(b), Equation 1 is an objective function in the case of using PP1 and PP4.

[Equation 6]

In addition, in the M-BPR ^G of the present invention, when a plurality of selected item pairs include {evaluation item, no information item}, {evaluation item, indifferent item}, {no information item, indifferent item}, the objective function is mathematical Same as Equation 7. At this time, referring to FIG. 1(b), Equation 2 is an objective function in the case of using PP2, PP3 and PP4.

[Equation 7]

In addition, each of Equations 8 and 9 below is the result of optimization by applying the SGD algorithm to Equations 6 and 7.

[Equation 8]

[Equation 9]

Here, the following relationship is established as in Equation (10).

[Equation 10]

Meanwhile, referring to FIG. 5 , a pseudocode of a model parameter learning algorithm according to an embodiment of the present invention is shown. That is, the entire process of M-BPR and M-BPR ^{G of the present invention is shown.}

First, the item recommendation apparatus predicts a rough evaluation result with respect to a plurality of unrated items included in the training dataset R = {(u, i)}. For each user u, the item set I is classified _{into I u} ^rt , I _u ^unk , I _u ^{uni .} Next, the model parameter θ is updated while iterating T times. At this time, if it is group-based (G = TRUE), three types of group evaluation results are obtained randomly for users u and u at each iteration. The evaluation results of 3 types of individual items are obtained randomly.

Also, when MODE=A, when the plurality of selected item pairs include {evaluated item, unrated item} and {uninformed item, indifferent item}, an evaluation result regarding the unrated item may be further obtained. Also, when MODE=B, the plurality of selected item pairs include {evaluation item, non-information item}, {evaluation item, indifferent item}, and {no information item, indifferent item}.

Finally, the item recommendation apparatus updates the model parameter θ for the user u. For this, the item recommendation apparatus may use Equations 3, 4, 8 and 9.

Also, referring to FIG. 6 , the accuracy of the BPR of the prior art, the two types of M-BPR of the present invention, and the two types of M-BPR ^{G of the present invention is shown.}

In this case, the two types of M-BPR depend on the type of the selected item pair, and (A) is a case in which the plurality of selected item pairs include {evaluated item, unrated item}, {uninformation item, indifferent item} , (B) is a case in which a plurality of selected item pairs include {evaluation item, non-information item}, {evaluation item, indifferent item}, and {uninformation item, indifferent item}.

Specifically, M-BPR and M-BPR ^G show better performance than the conventional BPR regardless of the number of recommended items (N). In addition, it can be seen that between M-BPR and M-BPR ^G, M-BPR ^G exhibits better performance than M-BPR.

Also, referring to FIG. 7 , the recommendation accuracy of a total of eight methods is shown.

At this time, it can be seen that the M-BPR of the present invention showed excellent performance with a significant difference overall for all datasets, compared to all other methods. More specifically, the GAINs of M-BPR are 12.84% (MovieLens100K), 27.41% (Watcha), 61.88% (Ciao), and 17.82% (Last.fm).

4 is a block diagram illustrating a multi-type pairwise-based item recommendation apparatus according to an embodiment of the present invention.

Referring to FIG. 4 , a multi-type pairwise-based item recommendation apparatus 400 according to an embodiment of the present invention includes a generator 410 , a classifier 420 , a learner 430 , and a recommender 440 . include

Meanwhile, the multi-type pairwise-based item recommendation apparatus 400 according to an embodiment of the present invention may be mounted on a computing device such as a server, a desktop PC, a notebook computer, a smart phone, and a tablet PC.

The generator 410 generates a comprehensive result by synthesizing the prediction result and the evaluation result, which are the results of predicting the evaluation results of a plurality of unevaluated items included in the evaluation result, by using the collaborative filtering technique.

The classification unit 420 classifies the plurality of unrated items into a plurality of uninformed items and a plurality of uninterested items from the overall result.

The learning unit 430 is based on the comprehensive result and a plurality of selected item pairs selected from among item pairs consisting of two of an evaluation item, an unevaluated item, an uninformed item, and an uninterested item included in the comprehensive result, learn parameters.

Finally, the recommendation unit 440 uses a final prediction result, which is a result of predicting the evaluation result for a plurality of unrated items based on the learned model parameter, to give at least one recommendation target among a plurality of users a predetermined number of item is recommended.

In another embodiment, the learning unit 430 generates an objective function based on a difference in evaluation results between item types included in each of a plurality of selected item pairs, and for one selected user selected from among a plurality of users, from the comprehensive result Obtain a plurality of evaluation result information including evaluation results for each item type included in the plurality of selected item pairs, and update the model parameter value so that the value of the objective function is maximized by using the objective function and the plurality of evaluation result information Then, the process of acquiring the plurality of evaluation result information and the process of updating the model parameter values may be repeated until a predetermined completion condition is satisfied.

In another embodiment, when the learning unit 430 updates the model parameter value, the gradient value of the objective function is calculated using a plurality of evaluation result information, and the gradient value is based on the stochastic gradient descent (SGD) technique. By applying , the model parameter value can be updated.

In another embodiment, the plurality of evaluation result information may include one evaluation result for each item type included in the plurality of selected item pairs.

In another embodiment, the plurality of evaluation result information may include a group evaluation result that is a plurality of evaluation results for each item type included in the plurality of selected item pairs, and the objective function may be based on a difference between the group evaluation results.

In another embodiment, the objective function may be based on the assumption that the evaluation result of the evaluation item is higher than the evaluation result of the unrated item, the uninformed item, and the indifferent item, and the evaluation result of the uninformed item is higher than the evaluation result of the indifferent item. have.

In another embodiment, the plurality of selected item pairs include an item pair composed of an evaluation item and an unrated item and an item pair composed of an uninformed item and an indifferent item, or an item pair composed of an evaluation item and non-information item. , may include an item pair composed of an evaluation item and an indifferent item, and an item pair composed of an uninformed item and an indifferent item.

The above-described embodiments of the present invention can be written as a program that can be executed on 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 (eg, ROM, floppy disk, hard disk, etc.) and an optically readable medium (eg, CD-ROM, DVD, etc.).

So far, with respect to the present invention, the preferred embodiments have been looked at. Those of ordinary skill in the art to which the present invention pertains will 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 are to be considered in an illustrative rather than a restrictive sense. The scope of the present invention is indicated 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)

A method of recommending an item using an evaluation result that is a result of evaluation of at least one item among a plurality of users by a plurality of users, the method comprising:
generating a comprehensive result by synthesizing a prediction result that is a result of predicting an evaluation result of a plurality of unrated items included in the evaluation result using a collaborative filtering technique and the evaluation result;
classifying the plurality of unrated items into a plurality of uninformed items and a plurality of indifferent items in the comprehensive result;
learning model parameters based on the comprehensive result and a plurality of selected item pairs selected from among item pairs consisting of two of an evaluation item, an unevaluated item, an uninformed item, and an indifferent item included in the comprehensive result; and
Recommending a predetermined number of items to at least one recommendation target among the plurality of users by using a final prediction result that is a result of predicting the evaluation results for the plurality of unrated items based on the learned model parameter
including,
The classification step is
Classifying the unrated items at a preset ratio of which the prediction result is lower than a predetermined threshold among the plurality of unrated items as the plurality of indifferent items, and classifying the remainder except the indifferent items into the plurality of uninformed items A method for recommending items based on multi-type pairwise, characterized in that. According to claim 1,
The step of learning the model parameters is
generating an objective function based on a difference in evaluation results between item types included in each of the plurality of selected item pairs;
obtaining a plurality of evaluation result information including evaluation results for each item type included in the plurality of selected item pairs from the comprehensive result for one selected user selected from among the plurality of users;
updating the model parameter value so that the value of the objective function is maximized by using the objective function and the plurality of evaluation result information; and
Repeating the steps of obtaining the plurality of evaluation result information and updating the model parameter values until a predetermined completion condition is satisfied
A multi-type pairwise-based item recommendation method comprising a. 3. The method of claim 2,
The step of updating the model parameter value is
calculating a gradient value of the objective function by using the plurality of evaluation result information; and
updating the model parameter value by applying the gradient value based on a stochastic gradient descent (SGD) technique.
A multi-type pairwise-based item recommendation method comprising a. 3. The method of claim 2,
The plurality of evaluation result information is
A multi-type pairwise-based item recommendation method, comprising: one evaluation result for each item type included in the plurality of selected item pairs. 3. The method of claim 2,
The plurality of evaluation result information is
and a group evaluation result, which is a plurality of evaluation results for each item type included in the plurality of selected item pairs,
The objective function is
A multi-type pairwise-based item recommendation method, characterized in that it is based on a difference between the group evaluation results. 3. The method of claim 2,
The objective function is
It is based on the assumption that the evaluation result of the evaluation item is higher than the evaluation result of the unrated item, the non-information item, and the indifferent item, and the evaluation result of the non-information item is higher than the evaluation result of the indifferent item A multi-type pairwise based item recommendation method. According to claim 1,
The plurality of selected item pairs
an item pair composed of the evaluated item and the unrated item and an item pair composed of the uninformed item and the indifferent item;
A multi-type pair comprising an item pair composed of the evaluation item and the non-information item, an item pair composed of the evaluation item and the indifferent item, and an item pair composed of the non-information item and the indifferent item Wise-based item recommendation method. An apparatus for recommending an item using an evaluation result that is a result of evaluating at least one item among a plurality of users by a plurality of users,
a generator for generating a comprehensive result by synthesizing a prediction result, which is a result of predicting an evaluation result of a plurality of unevaluated items included in the evaluation result, and the evaluation result using a collaborative filtering technique;
a classification unit for classifying the plurality of unrated items into a plurality of uninformed items and a plurality of indifferent items in the comprehensive result;
a learning unit for learning model parameters based on the comprehensive result and a plurality of selected item pairs selected from among item pairs consisting of two of an evaluation item, an unevaluated item, an uninformed item, and an indifferent item included in the comprehensive result; and
A recommendation unit for recommending a predetermined number of items to at least one recommendation target among the plurality of users by using a final prediction result that is a result of predicting the evaluation results for the plurality of unrated items based on the learned model parameter
including,
The classification section
Classifying the unrated items at a preset ratio of which the prediction result is lower than a predetermined threshold among the plurality of unrated items as the plurality of indifferent items, and classifying the remainder except the indifferent items into the plurality of uninformed items A multi-type pairwise-based item recommendation device, characterized in that. 9. The method of claim 8,
the learning unit
generating an objective function based on the difference in evaluation results between the types of items included in each of the plurality of selected item pairs,
Obtaining a plurality of evaluation result information including evaluation results for each item type included in the plurality of selected item pairs from the comprehensive result for one selected user selected from among the plurality of users,
using the objective function and the plurality of evaluation result information to update the model parameter value so that the value of the objective function is maximized,
and repeating the process of acquiring the plurality of evaluation result information and the process of updating the model parameter values until a predetermined completion condition is satisfied. 10. The method of claim 9,
When the learning unit updates the model parameter value,
Calculate the gradient value of the objective function by using the plurality of evaluation result information,
A multi-type pairwise-based item recommendation apparatus, characterized in that the model parameter value is updated by applying the gradient value based on a stochastic gradient descent (SGD) technique. 10. The method of claim 9,
The plurality of evaluation result information is
The multi-type pairwise-based item recommendation apparatus, characterized in that it includes one evaluation result for each item type included in the plurality of selected item pairs. 10. The method of claim 9,
The plurality of evaluation result information is
and a group evaluation result, which is a plurality of evaluation results for each item type included in the plurality of selected item pairs,
The objective function is
A multi-type pairwise-based item recommendation apparatus, characterized in that based on a difference between the group evaluation results. 10. The method of claim 9,
The objective function is
It is based on the assumption that the evaluation result of the evaluation item is higher than the evaluation result of the unrated item, the non-information item, and the indifferent item, and the evaluation result of the non-information item is higher than the evaluation result of the indifferent item A multi-type pairwise based item recommendation device. 9. The method of claim 8,
The plurality of selected item pairs
an item pair composed of the evaluated item and the unrated item and an item pair composed of the uninformed item and the indifferent item;
A multi-type pair comprising an item pair composed of the evaluation item and the non-information item, an item pair composed of the evaluation item and the indifferent item, and an item pair composed of the non-information item and the indifferent item Wise-based item recommendation device.

Images