An Improved Binary Cuckoo Search Algorithm For Feature Selection Using Filter Method And Chaotic Map

Feature selection is the process of reducing the number of variables for improving the classification model. The problem of feature selection can be broadly defined as an optimization problem. That is, finding a subset of input Features that results in the best model performance. feature selection is considered as a discrete binary problem. To have such binary vectors for the CS, a limit to the value of eggs (dimensions) must be applied by setting an upper bound and a lower bound .Then, the nests (solutions) are generated and updated in such way that the eggs can only accept the values between boundaries, so Binary Cuckoo Search (BCS) is the most effective and promising metaheuristic approach for this purpose. This approach proposes an improving BCS using a hybrid Chi-square-filter method and chaotic map for feature selection problems. Chi-square is employed for generating an initial solution problem and subsequently, it contributes to enhancing the quality of the final solution. Also, using the chaotic map (sinusoidal) to determine variable values of the step size (alpha) parameter via local search area. The proposed Chi-BCS is validated on several real-world datasets. The results of the experiments show that Chi-BCS has improved dimensionality reduction (76.69%) and classification accuracy (58.84%) when compared with other available methods like EBCS,ACO and FSFOA.