A method to generate negative cases for collaborative filtering recommendation system

The present invention relates to a method for generating negative case data for a collaborative filtering recommendation system. More specifically, the present invention relates to a method for generating negative case data for a collaborative filtering recommendation system, which is able to train...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: SONG GIL JAE, SONG HEE SEOK
Format: Patent
Sprache:eng ; kor
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The present invention relates to a method for generating negative case data for a collaborative filtering recommendation system. More specifically, the present invention relates to a method for generating negative case data for a collaborative filtering recommendation system, which is able to train the recommendation system by using training data obtained by mixing positive cases and negative cases and improve the performance of the recommendation system. The present invention is able to configure training data by mixing positive cases and negative cases, make a model of the recommendation system learn by using the training data obtained by mixing positive cases and negative cases, repeat the training, exclude negative cases with a greater error than a reference value, replace the excluded negative cases with other negative cases, configure training data, make the recommendation system learn again, and improve the accuracy of the recommendation system. 본 발명은 협업필터링 추천시스템을 위한 부정사례 데이터 생성방법에 관한 것으로, 보다 구체적으로는 긍정사례와 부정사례를 혼합한 훈련데이터를 이용해 추천시스템을 훈련시켜, 추천시스템의 성능을 향상시킬 수 있는 협업필터링 추천시스템을 위한 부정사례 데이터 생성방법에 관한 것이다. 본 발명은 긍정사례와 부정사례를 혼합하여 훈련데이터를 구성하고, 긍정사례와 부정사례를 혼합한 훈련데이터를 이용해 추천시스템의 모델을 학습하되, 반복하여 훈련하면서 오차가 기준치보다 큰 부정사례를 제외하고, 다른 부정사례로 대체하여 훈련데이터를 구성한 후 추천시스템을 재학습시켜, 추천시스템의 정확도를 보다 향상시킬 수 있는 효과가 있다.