Optimal blending of multiple independent prediction models

Abstract

We derive blending coefficients for the optimal blend of multiple independent prediction models with normal (Gaussian) distribution as well as the variance of the final blend. We also provide lower and upper bound estimation for the final variance and we compare these results with machine learning with counts, where only binary information (feature says yes or no only) is used for every feature and the majority of features agreeing together make the decision.

An intelligent feature recognition method of natural gas pipelines based on shapelet and blending fusion model

“…The core idea of blending is to use two layers of algorithms in tandem, as shown in figure 4. There are multiple base learners on the first layer and one and only one meta learner on the second layer, and the base learner on the first layer is responsible for fitting the relationship between the data and the true labels, outputting the prediction results, and composing a new feature matrix, and then allowing the meta learner on the second layer to learn and predict on the new feature matrix [49,50]. The specific steps are as follows.…”