In the realm of computational intelligence, the application of machine learning techniques to medical data presents a promising avenue for advancing disease pattern exploration and analysis. This paper presents a comprehensive approach to analyzing Parkinson’s disease data through advanced machine learning strategies. Initially, the study involved the integration of three distinct sub-dataset granules: train_proteins, train_peptides, and train_clinical. The datasets were subjected to a comprehensive preprocessing regimen, encompassing operations such as grouping, pivoting, and merging, in order to assemble a unified dataset. The resultant merged dataset comprised four target labels: updrs_1, updrs_2, updrs_3, and updrs_4. Subsequently, the dataset was divided into training and testing subsets to facilitate the model’s development and evaluation. The core of this study involved training two customized Artificial Neural Network (ANN) architectures on the training dataset. The performance of these trained models was then critically assessed on the testing dataset using symmetric mean absolute percentage error (sMAPE) and mean squared error (MSE) as key performance metrics. This approach offers novel insights into the complex mechanisms of Parkinson’s disease and demonstrates the potential of machine learning in medical data analysis.