Machine Learning Algorithm-Based Prediction of Diabetes Among Female Population Using PIMA Dataset.

Area of Science:

• Medical Informatics
• Computational Biology
• Data Science

Background:

• Diabetes mellitus is a metabolic disorder defined by elevated blood glucose levels.
• Early identification of diabetes is crucial for effective management and delaying disease progression.
• Machine learning (ML) offers advanced tools for identifying patterns in disease diagnosis and progression.

Purpose of the Study:

• To evaluate the efficacy of different ML algorithms for predicting diabetes in females.
• To compare the performance of Random Forest, Decision Tree, Naïve Bayes, and Logistic Regression models.

Main Methods:

• Utilized the PIMA dataset from Kaggle for analysis.
• Performed exploratory data analysis (EDA) using PCA, heatmap, and scatter plots.
• Implemented four ML algorithms (Random Forest, Decision Tree, Naïve Bayes, Logistic Regression) using Python and Rattle.

Main Results:

• Random Forest demonstrated superior performance with 80% accuracy, 82% precision, 20% error rate, and 88% sensitivity.
• The Random Forest model outperformed Decision Tree, Naïve Bayes, and Logistic Regression models.

Conclusions:

• ML-based predictive models show significant potential for early diabetes detection.
• Early prediction can facilitate timely intervention, potentially preventing disease worsening.