Stacking Ensemble Methods to Predict Obesity Levels in Adults

One of the causes of chronic disease is obesity, where obesity is a condition of someone who is overweight. The beyond-weight problems simplest befell in high-earnings nations; however, weight problems are growing dramatically in low-and middle-earnings countries. Based on this, obesity needs further investigation to suppress the increasing number of obese people. To predict obesity levels, we use a machine learning approach in this study. The dataset is a dataset on obesity levels in adults, so the method used is a classification method including Random Forest, Gradient Boosting, KNN, SVM, and Stacking Ensemble. This study uses Cross Validation (CV) to reduce the possibility of bias in the accuracy of results. The Cross Validation used is k = 10, 20, and 30. The classification method is evaluated using Performance measures, including accuracy, precision, and recalls, to get the best approach. This research includes data exploration, pre-processing data, seeking accuracy and comparing each method, namely Random Forest, Gradient Boosting, KNN, and SVM with CV k = 10, 20, and 30, and looking for accuracy and comparing the results of the stacking ensemble method with its meta-models. The highest accuracy is the accuracy value of the ensemble stacking method with the Gradient Boosting meta-model. The accuracy value obtained is 97.87% with k= 20 and k =30. With the same data, this accuracy is higher than the random forest, which gets an accuracy value of 97.64% with k = 20.