Scholarlink

In recent years, process mining (PM) has observed widespread use across the healthcare, education, logistics, and finance domain. Smart homes employ PM to examine human behaviour, health conditions, and enhance daily living. Existing research uses PM to study human behaviour. However, it failed to provide a comprehensive approach that studied/compared the different mixture models (MM) to determine the best model that closely characterizes human behaviour. As a result, this paper uses the gamma, Weibull and Gaussian MMs to represent the process durations of daily living and facilitate an accurate representation of human behaviour. The Expectation-Maximization (EM) algorithm was employed where the Kolmogorov-Smirnov (KS), Kullback-Leibler (KL) divergence, and Cramer-von Mises (CvM) tests were chosen to determine the best MM. The proposed approach was applied over the Kasteren, UCI and 4TU dataset.

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

Using Mixture Models to Characterise the Process Durations of Daily Living

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Eman Shaikh,
  • Bryan Scotney,
  • Sally McClean,
  • Zeeshan Tariq,
  • Nazeeruddin Mohammad

摘要

In recent years, process mining (PM) has observed widespread use across the healthcare, education, logistics, and finance domain. Smart homes employ PM to examine human behaviour, health conditions, and enhance daily living. Existing research uses PM to study human behaviour. However, it failed to provide a comprehensive approach that studied/compared the different mixture models (MM) to determine the best model that closely characterizes human behaviour. As a result, this paper uses the gamma, Weibull and Gaussian MMs to represent the process durations of daily living and facilitate an accurate representation of human behaviour. The Expectation-Maximization (EM) algorithm was employed where the Kolmogorov-Smirnov (KS), Kullback-Leibler (KL) divergence, and Cramer-von Mises (CvM) tests were chosen to determine the best MM. The proposed approach was applied over the Kasteren, UCI and 4TU dataset.