TPOT-NN: augmenting tree-based automated machine learning with neural network estimators

Abstract Automated machine learning (AutoML) and artificial neural networks (ANNs) have revolutionized the field of intelligence by yielding incredibly high-performing models to solve a myriad inductive tasks. In spite their successes, little guidance exists on when use one versus other. Furthermore, relatively few tools exist that allow integration both AutoML ANNs in same analysis yield results combining strengths. Here, we present TPOT-NN—a new extension tree-based software TPOT—and it explore behavior automated augmented with network estimators (AutoML+NN), particularly compared non-NN context simple binary classification number public benchmark datasets. Our observations suggest TPOT-NN is an effective tool achieves greater accuracy than standard some datasets, no loss others. We also provide preliminary guidelines for performing AutoML+NN analyses, recommend possible future directions methods research, especially TPOT.