A Hybrid Big Data Analytics Approach for Predicting Type II Diabetes Using H-SMOTE Tree

Abstract

Effective prediction models are required as the global prevalence of Type II Diabetes rises. This study proposes a hybrid Big Data analytics technique for predicting Diabetic Type II. The three key phases are data preparation using an Amalgam Multivariate Statistical Modeling Algorithm, feature extraction with Decision-Making Weighted Feature Selection, and D-H-SMOTE Tree classification. An Amalgam Multivariate Statistical Modeling Algorithm is used to preprocess the massive diabetic patient datasets. To handle data complexity and intricacies, this application employs a variety of statistical models. This research improves data quality and dependability to prepare it for analysis. The second step extracts features using Decision-Making Weighted Feature Selection. This research assesses characteristics based on their predictive power for Type II Diabetes using decision-making techniques. This phase reduces dimensionality and retains just the most relevant characteristics, enhancing prediction model efficiency and interpretability. Third, train the model using Artificial Neural Networks. ANNs can learn complicated data patterns and correlations. The trained model underpins categorization. In the last step, this research presents D-H-SMOTE Tree, a new categorization method. This approach addresses diabetes dataset class imbalance by combining SMOTE with decision trees. Oversampling and decision trees improve the model's generalization and classification, particularly with unbalanced class distributions.