Binary classifiers' outputs for ensemble creation

This dataset was created based on the paper 'Andras Hajdu, Gyorgy Terdik, Attila Tiba, and Henrietta Toman: A stochastic approach to handle knapsack problems in the creation of ensembles'.To summarize our experimental setup for UCI binary classification problems, we have considered base classifiers perceptron, decision tree, Levenberg-Marquardt feedforward neural network, random neural network, and discriminative restricted Boltzmann machine classifier for the 5 UCI datasets MAGIC Gamma Telescope, HIGGS, EEG EyeState, Musk (Version 2), and Spambase; datasets of large cardinalities were selected to be able to train synthetic variants of base classifiers on different subsets.To check our models for different numbers of possible ensemble members, the respective pool sizes were set to 30 and 100; the necessary number of classifiers has been reached via synthesizing the base classifiers with training them on different subsets of the training part of the given datasets.

本数据集基于Andras Hajdu、Gyorgy Terdik、Attila Tiba和Henrietta Toman发表的论文《一种处理背包问题的随机方法在集成学习中的应用》构建。为概括我们针对UCI二元分类问题的实验设置，我们选取了感知机、决策树、Levenberg-Marquardt前馈神经网络、随机神经网络以及判别性受限玻尔兹曼机分类器作为基础分类器，应用于五个UCI数据集：MAGIC伽玛望远镜数据集、HIGGS数据集、EEG EyeState数据集、Musk（版本2）数据集和Spambase数据集；选取大型基数的数据集以便在不同的子集上训练基础分类器的合成版本。为了验证不同可能集成成员数的模型，我们分别设定了30和100的池大小；通过在给定数据集的训练部分的不同子集上训练基础分类器，合成了所需数量的分类器。