Understanding and Classifying Personal Life Patterns Using Calendar Logs
- 주제(키워드) data mining , text mining
- 발행기관 고려대학교 대학원
- 지도교수 주재걸
- 발행년도 2019
- 학위수여년월 2019. 2
- 유형 Text
- 학위구분 석사
- 학과 대학원 컴퓨터학과(정보대학)
- 원문페이지 36 p
- 실제URI http://www.dcollection.net/handler/korea/000000083378
- UCI I804:11009-000000083378
- DOI 10.23186/korea.000000083378.11009.0000822
- 본문언어 영어
- 제출원본 000045978970
초록/요약
Various activities of a personal life are being logged in many different forms. For example, a check-in service obtains the information about where an individual visit, and an online commerce website collects that about what s/he purchases. These types of information are being actively utilized in numerous applications such as personalized recommendation. As another type of logging data, we collected personal activity data stored in an online calendar and scheduling service and performed an analysis of understanding personal life patterns and classifying their activities into meaningful categories based on the event information. In detail, using 4,016,073 personal event instances of 8,545 users, we construct high-level categories and assign the event instance labels based on the relevant keyword set through the keyword recommendation from word embedding. We then present the classifier among different schedule types based on the multi-labeled deep learning model using the word embedding model along with the user representations. Based on our model, we further provide in-depth analysis about the behaviors behind the personal life patterns.
more목차
Section 1 Introduction 1
Section 2 Related Work 4
2.1 Pattern Mining 4
2.2 Human Activity Prediction and Recommendation 5
Section 3 Proposed Method 6
3.1 Data Description 7
3.2 Category Building 8
3.3 Multi-Labeled Neural Network Model 11
Section 4 Predicting Schedule Categories 15
4.1 Feature Importance Analysis 17
Section 5 Personal Life Pattern Discovery 20
Section 6 Conclusions 27
References 28
