Linear iterative feature embedding: an ensemble framework for an interpretable model

Abstract A new ensemble framework for an interpretable model called linear iterative feature embedding (LIFE) has been developed to achieve high prediction accuracy, easy interpretation, and efficient computation simultaneously. The LIFE algorithm is able fit a wide single-hidden-layer neural network (NN) accurately with three steps: defining the subsets of dataset by projections nodes, creating features from multiple narrow NNs trained on different data, combining model. theoretical rationale behind also provided connection loss ambiguity decomposition stack methods. Both simulation empirical experiments confirm that consistently outperforms directly many other benchmark models, including multilayers feed forward (FFNN), Xgboost, random forest (RF) in experiments. As NN, intrinsically interpretable. Meanwhile, both variable importance global main interaction effects can be easily created visualized. In addition, parallel nature base learner building makes computationally leveraging computing.