Scholarlink

We propose a generalization of the Temporal Nearest Neighbor Gaussian Process to model high-frequency data in the context of individual physical activity monitoring. To our knowledge, previous applications considered the exponential as a model for the temporal covariance but more flexible alternatives are readily available and could be easily embedded in the same modelling framework. Thus, the idea is to verify the comparative performances of alternative temporal covariance specifications and find the one that best suits applications in such a context. The comparison is pursued on a dataset compiled from the PASTA-LA study, which monitored the physical activity patterns of individuals across space and time using a tri-axial accelerometer.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Temporal Nearest Neighbor Gaussian Process (tNNGP) with Flexible Covariance for Modelling Physical Activity

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Marco Mingione,
  • Pierfrancesco Alaimo Di Loro

摘要

We propose a generalization of the Temporal Nearest Neighbor Gaussian Process to model high-frequency data in the context of individual physical activity monitoring. To our knowledge, previous applications considered the exponential as a model for the temporal covariance but more flexible alternatives are readily available and could be easily embedded in the same modelling framework. Thus, the idea is to verify the comparative performances of alternative temporal covariance specifications and find the one that best suits applications in such a context. The comparison is pursued on a dataset compiled from the PASTA-LA study, which monitored the physical activity patterns of individuals across space and time using a tri-axial accelerometer.