Scholarlink

Diabetes has become one of the much widespread metabolic disorders not only in India but also in world with a huge mortality rate. Diabetes disease has numerous medical issues and inauspicious influence on various organs of human beings. Hence, correct detection and before time treatment of diabetic patients is essential to avoid its severity and health complications. If early prediction is done accurately, its complications and adverse effects on the other organs of the body can be reduced or avoided. This research work has proposed new methodology using different machine learning algorithms with different feature scaling techniques for diabetes prediction and its comparison. In this methodology, the pre-processing is done on the dataset for removing the null and duplicate values and related attributes are selected using correlation attribute evaluation method. The standardization, normalization, and min-max techniques are applied for scaling the feature from diabetes dataset. ML algorithms, e.g. Logistic Regression (LR), K-Nearest Neighbour (KNN), Naïve Bayes (NB), Support Vector Machine (SVM), Decision Tree (DT) and Random Forest (RF) are used for the diabetes prediction. Experiment conducted by applying different feature scaling and without features scaling techniques on the PIMA Indian Diabetes Dataset using machine learning libraries in the python programming. This study’s primary goal is to identify the best feature scaling methods that produce the most accurate results for diabetes prediction. The experimental results have shown that the Random Forest with Standardization scalar technique improved the prediction accuracy from 75.97% to 79.87%. Then, KNN with Standardization scalar and Min-Max scalar has improved the prediction accuracy from 72.72% to 76.62%. Decision Tree with Standardization scalar technique improved the prediction accuracy from 69.48% to 72.08%. The Logistic Regression with Standardization scalar technique has improved the prediction accuracy from 75.97% to 77.92%. The Support Vector Machine with Min-Max Scaling technique has given 77.27%. Naïve Bayes shows 77.27% accuracy which remains unchanged after applying features scaling also.

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

Diabetes Prediction Using Machine Learning Algorithms with Different Feature Scaling Techniques

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Jyoti N. Dangat,
  • Santosh P. Shrikhande

摘要

Diabetes has become one of the much widespread metabolic disorders not only in India but also in world with a huge mortality rate. Diabetes disease has numerous medical issues and inauspicious influence on various organs of human beings. Hence, correct detection and before time treatment of diabetic patients is essential to avoid its severity and health complications. If early prediction is done accurately, its complications and adverse effects on the other organs of the body can be reduced or avoided. This research work has proposed new methodology using different machine learning algorithms with different feature scaling techniques for diabetes prediction and its comparison. In this methodology, the pre-processing is done on the dataset for removing the null and duplicate values and related attributes are selected using correlation attribute evaluation method. The standardization, normalization, and min-max techniques are applied for scaling the feature from diabetes dataset. ML algorithms, e.g. Logistic Regression (LR), K-Nearest Neighbour (KNN), Naïve Bayes (NB), Support Vector Machine (SVM), Decision Tree (DT) and Random Forest (RF) are used for the diabetes prediction. Experiment conducted by applying different feature scaling and without features scaling techniques on the PIMA Indian Diabetes Dataset using machine learning libraries in the python programming. This study’s primary goal is to identify the best feature scaling methods that produce the most accurate results for diabetes prediction. The experimental results have shown that the Random Forest with Standardization scalar technique improved the prediction accuracy from 75.97% to 79.87%. Then, KNN with Standardization scalar and Min-Max scalar has improved the prediction accuracy from 72.72% to 76.62%. Decision Tree with Standardization scalar technique improved the prediction accuracy from 69.48% to 72.08%. The Logistic Regression with Standardization scalar technique has improved the prediction accuracy from 75.97% to 77.92%. The Support Vector Machine with Min-Max Scaling technique has given 77.27%. Naïve Bayes shows 77.27% accuracy which remains unchanged after applying features scaling also.