Tensor Factorization with Modified Artificial Bee Colony (MABC) Algorithm for Missing Value Imputation in Breast Cancer Diagnosis

Abstract

Cell division and uncontrolled growth caused by changes in cell results in a disease called Cancer. In large data set, patterns are discovered by the process of data mining. Database systems, statistics and machine learning intersections are involved in this method. In clinical diagnosis, machine learning and data mining are commonly used.  Missing values are included in this field which reduces the accuracy of diagnosis. Missing data is estimated by an enhanced Reduced Adaptive Particle Swarm Optimization (RAPSO) which is a modified version of tensor factorization. With insufficiency of data and issues in local optima, data’s cannot be estimated properly by RAPSO algorithm tensor.  So, instead of RAPSO algorithm, Modified Artificial Bee Colony (MABC) is used in tensor factorization method to enhance the accuracy.

Chaotic search is enhanced by Modified Artificial Bee Colony (MABC) algorithm. Process of exploration and exploitation is enhanced by computing all phases of ABC.In employed bee phase, chaotic search based new search is used to enhances onlooker bees probability of finding best solutions. In phase of onlooker bees, new solutions are used to replace worst solutions. Adapt distinctive method for MABC initialization in random manner and it uses Bayesian network. The results of proposed MABC-Bayesian Network (MABC-BN) method are superior to other RAPSO algorithm and BN with respect to specificity Root Mean Square Error(RMSE), sensitivity and accuracy.