Scholarlink

This study analyses the thyroid disease dataset obtained from Kaggle, originally sourced from Garavan Institute in Sydney, Australia, which consists of 3771 cases, to uncover patterns and understand the consequences of this non-communicable disease. The study carried out a thorough analysis of the prevalence of hypothyroidism across age and gender brackets and built a model that would predict the presence of hypothyroidism. The data is cleaned by removing sparse attributes, addressing outliers using the Winsorization method, and handling missing values using techniques like central tendencies. Biasness was removed using Synthetic Minority Over-sampling Technique, and Age was converted to categories using Data Discretization for easy analysis. It is found that the prevalence of hypothyroidism is higher in females (63.24%) when compared to males (28.44%). And adult women (ages 20–59) had the highest positive test rate (54.70%), followed by elderly women (ages 60–100) at (40.98%). Hormonal imbalances, majorly involving T3 (Triiodothyronine), TT4 (Total Thyroxine), and TSH (Thyroid-stimulating hormone), were predominant factors in hypothyroidism. Elevated TSH and normal T3 and T4U (Thyroxine Uptake) indicated subclinical hypothyroidism, especially in adult and elderly women (79%). Screening is recommended for anyone with abnormal hormone levels, as there is an 82.79% chance of having moderate or secondary hypothyroidism. A hypothyroidism predictive model was built using various classification algorithms, among which the Gradient Booster model performed well with an accuracy of 96.37%. The study highlights the need for early detection, particularly considering the significant impact of gender and age on hypothyroidism prevalence, suggesting targeted healthcare interventions.

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 Machine Learning Approach for Analyzing and Predicting Hypothyroidism Based on Hormonal Change

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • P. G. Sunitha Hiremath,
  • Manohar Madgi,
  • Sameer Mansur,
  • Ishan Kulkami,
  • Vedaant Mathreja,
  • Saisatwik Madalageri

摘要

This study analyses the thyroid disease dataset obtained from Kaggle, originally sourced from Garavan Institute in Sydney, Australia, which consists of 3771 cases, to uncover patterns and understand the consequences of this non-communicable disease. The study carried out a thorough analysis of the prevalence of hypothyroidism across age and gender brackets and built a model that would predict the presence of hypothyroidism. The data is cleaned by removing sparse attributes, addressing outliers using the Winsorization method, and handling missing values using techniques like central tendencies. Biasness was removed using Synthetic Minority Over-sampling Technique, and Age was converted to categories using Data Discretization for easy analysis. It is found that the prevalence of hypothyroidism is higher in females (63.24%) when compared to males (28.44%). And adult women (ages 20–59) had the highest positive test rate (54.70%), followed by elderly women (ages 60–100) at (40.98%). Hormonal imbalances, majorly involving T3 (Triiodothyronine), TT4 (Total Thyroxine), and TSH (Thyroid-stimulating hormone), were predominant factors in hypothyroidism. Elevated TSH and normal T3 and T4U (Thyroxine Uptake) indicated subclinical hypothyroidism, especially in adult and elderly women (79%). Screening is recommended for anyone with abnormal hormone levels, as there is an 82.79% chance of having moderate or secondary hypothyroidism. A hypothyroidism predictive model was built using various classification algorithms, among which the Gradient Booster model performed well with an accuracy of 96.37%. The study highlights the need for early detection, particularly considering the significant impact of gender and age on hypothyroidism prevalence, suggesting targeted healthcare interventions.