ENHANCING EARLY-STAGE DIABETES PREDICTION USING DATA MINING ALGORITHMS AND NORMALISATION TECHNIQUES

The global rise in diabetes prevalence is a pressing health concern, emphasizing the critical need for timely interventions and effective diagnostic strategies. Early detection of diabetes is of paramount importance as it allows for timely interventions, potentially preventing or delaying the onset of complications associated with the disease. Early-stage diabetes detection can lead to better management and control of the condition, reducing the risk of severe complications and improving the overall quality of life for patients. In the realm of medical research, data mining algorithms have emerged as powerful tools for extracting meaningful patterns and insights from vast amounts of data. These algorithms, when applied to medical datasets, have the potential to revolutionize the way diseases like diabetes are detected and managed. The focus of this study is to explore the potential of detecting early-stage diabetes using attributes that are not strictly medical in nature. Such an approach can broaden the scope of diabetes detection, making it more accessible and potentially more efficient. To achieve this, the study delves into the evaluation of various data normalization techniques. Normalization is a crucial step in data preprocessing, ensuring that the features in a dataset are on a comparable scale. This is vital for the performance of many machine learning algorithms, as features on different scales can unduly influence the outcome. Several algorithms were employed in this study, including Naive Bayes, K-Nearest Neighbour (KNN), Support Vector Machines (SVM), Decision Tree, Random Forest, and Gradient Boosting Classifier (GBC). These algorithms were applied to the Early Stage Diabetes Risk Prediction Dataset, which was preprocessed using normalization methods like Decimal Point Scaling, Z-Score Normalisation, and Pareto Scaling, to name a few. The results of this study are promising. Notably, even without relying on traditional medical diagnostic data, early-stage diabetes prediction was achievable. The Gradient Boosting Classifier (GBC), when combined with the right data normalization technique, stood out among the algorithms, achieving a prediction accuracy rate of 99.038%. This accuracy is a testament to the potential of data-driven approaches in medical research and their ability to provide valuable insights into disease detection and management.