Battle of Machine Learning Titans for Diabetes in Females

Diabetes is a chronic illness and is very dangerous to other body organs. In females, the case of diabetes worsens due to female hormones. This study aims to apply various machine learning techniques for data analysis on the Pima Indian Heritage dataset, specifically focusing on predicting whether females have diabetes or not. The dataset contains various health-related features, including blood pressure, BMI, and age, among others. The key objective is to find a highly precise and effective machine learning algorithm for predicting diabetes in females. To achieve this goal, pre-processing of the dataset is done for feature selection, and six machine learning methods are utilized with regard to the pre-processed dataset, including support vector machine, logistic regression, random forest, KNN, naïve Bayes, and decision trees. Every algorithm's performance is assessed using criteria, namely F1-score, precision, accuracy, and recall. Preliminary outcomes disclose that the random forest technique rises above from the rest others with an accuracy of 85.0% and an AUC of 97.0% in predicting diabetes in females. Interestingly, results also show that age and BMI are the most significant features in predicting diabetes, while blood pressure and skin thickness have a lesser impact. Overall, this paper provides valuable insights into predicting diabetes in females using the techniques of machine learning. This analysis can be valuable for healthcare professionals in identifying individuals at high risk of developing diabetes and developing effective prevention and management strategies.