A1391
Title: Systematic review: Comparison between generalized additive models and neural networks across application areas
Authors: Jessica Doohan - University of Limerick (Ireland) [presenting]
Kevin Burke - University of Limerick (Ireland)
Abstract: Over the past two decades, advancements in machine learning, particularly neural networks, have reshaped predictive modelling. While neural networks are often regarded as black-box algorithms, their universal approximation properties have made them highly effective predictors, albeit difficult to interpret. Much of this development has occurred outside statistics, creating a disconnect in terminology and methodology. Recent works have highlighted parallels with traditional statistical approaches, showing that multilayer perceptrons can be viewed as nonparametric generalisations of regression models. Despite these connections, comparative studies have largely focused on simple models such as linear regression, often within specific application domains. Such comparisons are yet to consider more flexible statistical models. Generalized additive models (GAMs), which extend linear regression by allowing non-linear predictor effects while preserving interpretability, provide a more appropriate benchmark. To date, no systematic review has examined how GAMs perform relative to neural networks. The focus is on that gap through a systematic review of over 140 papers and 400 datasets. Beyond summarizing comparisons, reported performance metrics are analyzed using mixed-effects modelling to investigate characteristics that can explain and quantify observed differences, including application area, publication year, sample size, number of predictors, and neural network complexity.
