APPARATUS AND METHOD FOR RECOMMENDATION OF POINT-OF-INTEREST

The present invention relates to a device and a method for recommending a point of interest. The device for recommending a point of interest, includes: a category projection unit acquiring an embedding vector in a category is projected by adding a place embedding vector for a place of a specific tim...

Full description

Saved in:
Bibliographic Details
Main Authors LEE KUN WOO, CHUNG YON DOHN
Format Patent
LanguageEnglish
Korean
Published 12.04.2023
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:The present invention relates to a device and a method for recommending a point of interest. The device for recommending a point of interest, includes: a category projection unit acquiring an embedding vector in a category is projected by adding a place embedding vector for a place of a specific time point, with respect to a combination result of a category embedding vector for a category to which the place of a specific time point belongs, and a user embedding vector; a relation building unit acquiring at least one weight by using a place embedding vector for a place visited in the past, and a category embedding vector for a category to which the place visited in the past belongs, and acquiring a relation embedding vector based on the at least one weight; and a learning processing unit performing learning by using the embedding vector in which the category is projected, and the relation embedding vector. Therefore, the present invention is capable of preventing a deterioration in learning performance. 관심 장소 추천 장치 및 방법에 관한 것으로, 관심 장소 추천 장치는 특정한 시점의 장소가 속한 범주에 대한 범주 임베딩 벡터와 사용자 임베딩 벡터의 조합 결과에 대해, 상기 특정한 시점의 장소에 대한 장소 임베딩 벡터를 부가함으로써 범주가 투영된 임베딩 벡터를 획득하는 범주 투영부, 과거 시점에 방문한 장소에 대한 장소 임베딩 벡터와, 과거 시점에 방문한 장소가 속한 범주에 대한 범주 임베딩 벡터를 이용하여 적어도 하나의 가중치를 획득하고, 상기 적어도 하나의 가중치를 기반으로 관계 임베딩 벡터를 획득하는 관계 구축부 및 상기 범주가 투영된 임베딩 벡터 및 상기 관계 임베딩 벡터를 이용하여 학습을 수행하는 학습 처리부를 포함할 수 있다.
Bibliography:Application Number: KR20210131497