Thyroid Disease Prediction Using Multi Layer Perceptron
S. Sachin Kumar, Pasuladi Narsimha Reddy, V Nivethitha, Suthir Sriram · 2025
Thyroid diseases are among the most common endocrine disorders and need precise and timely diagnosis for effective management and treatment. This project applies advanced machine learning techniques, focusing on the Multi-Layer Perceptron (MLP) model, to predict thyroid conditions with high accuracy and reliability. Extensive preprocessing involves handling null values, duplicates, outliers, feature selection with XGBoost, and class imbalance with SMOTE. The dataset utilized in this case is 2800 samples with 30 features. Scaling of data and 80/20 split for training and test were applied to further refine model performance. To make the system more user-friendly, an interface was designed using Flask which can take in input easily and provide real-time predictions. It targets healthcare providers and patients with a reliable and effective diagnostic tool. The findings indicate that the model performs well in achieving high diagnostic accuracy and, therefore, possesses a potential to revolutionize healthcare and improve decisionmaking processes by employing machine learning coupled with user-centered interfaces. Thyroid diseases are one of the most prevalent endocrine diseases and require accurate and timely diagnosis for proper management and treatment. This project utilizes state-of-the-art machine learning methods, with emphasis on the Multi-Layer Perceptron (MLP) model, to accurately predict thyroid conditions. Heavy preprocessing includes dealing with null values, duplicates, outliers, feature selection using XGBoost, and class imbalance using SMOTE. The dataset employed here is 2800 samples with 30 features. Training and test optimization through data scaling and the 80/20 split was employed to further improve model performance. An interface was developed through Flask for improved user-friendliness to take input easily and provide real-time predictions. It targets healthcare practitioners and patients with a reliable and effective diagnostic tool. The outcomes indicate that the model performs effectively in achieving high diagnostic precision and, therefore, possesses a capability of transforming healthcare and decision-making processes using machine learning together with user-centered interfaces.