Chronic non-communicable diseases such as cancer, stroke, diabetes mellitus (DM), hypertension (HT), chronic kidney failure (CKF), and cardiovascular disease (CVD) have become major health issues worldwide. Another challenge arises when predicting these diseases using datasets from general checkup (GCU) examinations. One of the problems is the imbalance in the number of positive and negative classes in the data. In addition, doctors need additional information from GCU data to provide preventive therapy to people at risk of developing chronic diseases in the future. This can be achieved by integrating expert knowledge with machine learning models. This research aims to predict chronic diseases using a single type of GCU data. Another objective is to modify the synthetic minority oversampling technique (SMOTE) to handle imbalanced data and implement voting ensemble learning based on expert judgment. The results show that the proposed model improves the prediction performance by 10\% to 47\% compared to traditional models. This system provides guidance to medical professionals to perform preventive interventions more accurately and efficiently, helping to improve the quality of life of patients.