@inproceedings{fddad9fa50f64d8184415a63d51db340,

title = "Weight Learning in Weighted ELM Classification Model Based on Genetic Algorithms",

abstract = "In cost sensitive classification problems we often suppose to have a known cost matrix in which each element represents the cost of mistakenly classifying an object from one class into another. Weighted least square, which does not equally consider individual classes and therefore assigns a different weight to each class of samples, is a typical approach to dealing with cost sensitive classification problems. Theoretically and experimentally it is confirmed that reasonable class weights will greatly improve classification ability of a learning model. Unfortunately we only know that these weights depend generally on cost matrix but very few methods can be used to specifically determine these weights according to cost matrix. This paper proposes a weighted least square (WLS) model of random weight network and then successfully uses the model in cost sensitive classification. A genetic algorithm to determine weights of different sample classes based on a cost matrix is given. Model analysis and experimental simulations are conducted. Considering the total misclassification cost as the evaluation index, a comparative study shows that our WLS model is far superior to the existing cost sensitive ELM and cost sensitive naive Bayes models.",

keywords = "Cost matrix, Cost sensitive, Genetic algorithm, Random weight network, Weighted least square method",

author = "Peng YAO and Xi-Zhao WANG",

note = "This work was supported in part by the National Natural Science Foundation of China (Grant 61772344 and Grant 61732011), in part by the Natural Science Foundation of SZU (Grant 827-000140, Grant 827-000230, and Grant 2017060), and in part by GuangDong Province 2014GKXM054.; 17th International Conference on Machine Learning and Cybernetics, ICMLC 2018 ; Conference date: 15-07-2018 Through 18-07-2018",

year = "2018",

doi = "10.1109/ICMLC.2018.8526986",

language = "English",

series = "Proceedings - International Conference on Machine Learning and Cybernetics",

publisher = "IEEE Computer Society",

pages = "370--377",

booktitle = "Proceedings of 2018 International Conference on Machine Learning and Cybernetics, ICMLC 2018",

address = "United States",

}