Enhancing Diabetes Diagnosis: A Comparative Study of Machine Learning Algorithms with Data Preprocessing Techniques

The primary aim of this study is to assess the performance of various ML algorithms in diagnosing diabetes depending on various data preprocessing methods. Utilizing the Pima Indian Diabetes Database, the study investigates how data preprocessing such as Min-Max scaling, Z-score normalization and SMOTE (Synthetic Minority Over-sampling Technique) influence model performance in contrast with models trained on raw database. Various Machine learning algorithms are employed and the performance are measured using metrics like accuracy, precision, recall and F1-score. This show that Data preparation process is very important in improving the performance of the model as the scaling and resampling enhance model performance as compared to using raw data. This paper, therefore, ensures the need and significance of proper handling of data while building predictive models, especially in the field of medical diagnosis of diseases such as diabetes.