Inductive Gaussian representation of user-specific information for personalized stress-level prediction
作者：
Highlights：
• Formulate a unified end-to-end model (PSP-IGR) for personalized stress-level prediction.
• Categorize heterogeneous inputs into three categories according to their characteristics.
• Devise Inductive Gaussian representation (IGR) for user-specific information.
• Generalize to out-of-sample users under uncertainty modeling with IGR.
• Evaluate effects of PSP-IGR and IGR on stress-level prediction accuracy.
摘要
•Formulate a unified end-to-end model (PSP-IGR) for personalized stress-level prediction.•Categorize heterogeneous inputs into three categories according to their characteristics.•Devise Inductive Gaussian representation (IGR) for user-specific information.•Generalize to out-of-sample users under uncertainty modeling with IGR.•Evaluate effects of PSP-IGR and IGR on stress-level prediction accuracy.
论文关键词：Context awareness,Knowledge representation,Neural networks,Personalization,Stress measurement,Inductive gaussian representation
论文评审过程：Received 30 October 2019, Revised 1 March 2021, Accepted 14 March 2021, Available online 18 March 2021, Version of Record 16 April 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.114912
