Embedding-based real-time change point detection with application to activity segmentation in smart home time series data
作者：
Highlights：
• Novel embedding-based CPD method. Totally unsupervised. No need of labeled data.
• ε-real-time configurable. Balance the model between delay and accuracy.
• Fast and efficient in terms of computational cost. Stable between executions.
• No feature engineering required. Only discretization of time series data.
• Hybrid approach to improve accuracy. Easy to perform transfer learning.
摘要
•Novel embedding-based CPD method. Totally unsupervised. No need of labeled data.•ε-real-time configurable. Balance the model between delay and accuracy.•Fast and efficient in terms of computational cost. Stable between executions.•No feature engineering required. Only discretization of time series data.•Hybrid approach to improve accuracy. Easy to perform transfer learning.
论文关键词：Activity transition detection,Change point detection,Activity segmentation,Smart homes,Action embeddings,Sensor embeddings
论文评审过程：Received 1 December 2020, Revised 11 June 2021, Accepted 19 July 2021, Available online 24 July 2021, Version of Record 28 July 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.115641
