Abstract

Feature Selection Is A Process Commonly Used In Machine Learning. Based On Binary Cuckoo Optimization Algorithm (Bcoa) And Information Theory, This Paper Proposes A
New Filter Feature Selection Method For Classification Problems. The Proposed Algorithm Is Based On Bcoa And The Mutual Information (Mi) Of Each Pair Of Features, Which Determines The Relevance And Redundancy Of The Selected Feature Subset.
Different Weights For The Relevance And Redundancy In The Fitness Functions Of The Proposed Algorithm Are Used To Further Improve Their Performance In Terms Of The Number Of Features And The Classification Accuracy. In The Experiments, An Artificial
Neural Network (Ann) Is Employed To Evaluate The Classification Accuracy Of The Selected Feature Subset On The Test Sets Of Six Datasets. The Results Show That Proposed Algorithms Can Significantly Reduce The Number Of Features And Achieve High Classification Accuracy In Almost All Cases.