An efficient SVM-GA feature selection model for large healthcare databases

This paper presents an efficient hybrid feature selection model based on Support Vector Machine (SVM) and Genetic Algorithm (GA) for large healthcare databases. Even though SVM and GA are robust computational paradigms, the combined iterative nature of a SVM-GA hybrid system makes runtime costs infeasible when using large databases. This paper utilizes hierarchical clustering to reduce dataset size and SVM training time, multi-resolution parameter search for efficient SVM model selection, and chromosome caching to avoid redundant fitness evaluations. This approach significantly reduces runtime and improves classification performance. Categories and Subject Descriptors I.2.6 [Artificial Intelligence]: Learning – concept learning, induction, knowledge acquisition, parameter learning. General Terms Algorithms, performance, experimentation. Keywords Classifier systems, data mining, machine learning, optimization, parameter tuning, genetic algorithms, support vector machines.

• Type: Conference
• Year: 2008
• Where: GECCO
• Authors: Rick Chow, Wei Zhong, Michael Blackmon, Richard Stolz, Marsha Dowell
• Keywords: Efficient Svm Model, GECCO 2008, Genetic Algorithms, Optimization, Support Vector Machines