Beyond Parental Height: A Multi-Model Deep Learning Approach for Personalized Adult Height Prediction

A new multi-model deep learning approach is proposed for the prediction of adult height from the Galton historical dataset with advanced feature engineering. Traditional approaches to height prediction have relied on linear relationships between the heights of parents and their offspring, which cannot explain the more intricate interplay of genetic, environmental, and lifestyle factors. This study bridges the gap by integrating other influences, such as birth order and physical activity, in developing a more holistic model for height prediction. Experiments were performed on raw and processed data, mainly on the impact of removing outliers on the accuracy of the model. Results reflect the fact that a multi-modeling system would predict better than the single-model scheme because combining all the factors is thought to make it more flexible and reliable. Data pre-processing was a very important activity, particularly outlier handling since results indicated that the predictive accuracy significantly improved when outliers were removed. This underlines robust data cleaning in machine learning algorithms. In summary, this study furnishes the pediatrician and the parent with a useful tool in delivering more reliable growth forecasts: underlining the role advanced data science techniques can play within personalized healthcare. This advances height prediction but lays grounds for further studies in individually modeling growth.