Obesity, a chronic condition influenced by genetic, behavioral, and environmental factors, affects one in eight individuals globally, significantly increasing the risk of diabetes, heart disease, and various cancers. Early prediction and intervention are critical to mitigating related health issues, reducing healthcare costs, and improving public health outcomes. This study explores the effectiveness of several Machine Learning (ML) algorithms, including K-Nearest Neighbors, Support Vector Machine, Decision Tree, Random Forest, and Neural Network in forecasting obesity. Among these, the Neural Network demonstrated the peak accuracy at 95.5%, surpassing other models due to its capacity to handle complex, non-linear relationships and large datasets. The investigation focuses on further analysis of variants of Neural Networks, including Narrow, Bi-layered and Optimized Neural Networks. The Optimized Neural Network outperformed the others, although its recall rate remained suboptimal. To enhance recall accuracy, proposed research is dedicated to Feature Engineering, employing methods such as MRMR, ANOVA, Chi2, and Kruskal Wallis to refine the dataset. This approach seeks to enhance model performance, improving the accuracy in predicting the obesity and facilitating targeted interventions for individuals at risk. The study underscores the importance of continuous model optimization and the integration of advanced techniques to achieve higher prediction accuracy. Ultimately, these advancements can contribute significantly to public health strategies aimed at combating the obesity epidemic.