Scholarlink

Data-driven predictive models in epidemiological studies remain unsatisfactory for a range of reasons. One issue includes a lack of non-parametric data-driven predictive models underpinning research in this field. The Gaussian Process is a state-of-art non-parametric data driven Bayesian framework used widely for Machine Learning tasks from regression, classification to clustering. The main challenge is to design a valid kernel. Cancer is the second leading worldwide cause of death and skin cancer is one of the most common causes of cancer death. Our research uses an aggregated data set composed of skin cancer tumors diagnosed from 1995–2017 (ICD-10 C43x - C44x), collected within the UK. This data set includes counts grouped by age group, sex, and diagnosis year for all melanoma and non-melanoma skin cancers. In this work, the Gaussian Process framework is applied, alongside the introduction of a novel kernel and features including age limit, gender, and type of cancer which resulted in a prediction model able to predict the trend for the upcoming 8 years (up to 2025).

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

A Gaussian Process Framework for Prognostication and Visualization in Dermatological Oncology

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Md Habibur Rahman,
  • Nabilah Hossain Sarker,
  • Md Musfique Anwar,
  • Mufti Mahmud,
  • David J. Brown,
  • Muhammad Arifur Rahman

摘要

Data-driven predictive models in epidemiological studies remain unsatisfactory for a range of reasons. One issue includes a lack of non-parametric data-driven predictive models underpinning research in this field. The Gaussian Process is a state-of-art non-parametric data driven Bayesian framework used widely for Machine Learning tasks from regression, classification to clustering. The main challenge is to design a valid kernel. Cancer is the second leading worldwide cause of death and skin cancer is one of the most common causes of cancer death. Our research uses an aggregated data set composed of skin cancer tumors diagnosed from 1995–2017 (ICD-10 C43x - C44x), collected within the UK. This data set includes counts grouped by age group, sex, and diagnosis year for all melanoma and non-melanoma skin cancers. In this work, the Gaussian Process framework is applied, alongside the introduction of a novel kernel and features including age limit, gender, and type of cancer which resulted in a prediction model able to predict the trend for the upcoming 8 years (up to 2025).