Feature selection of unbalanced breast cancer data using particle swarm optimization

<p>Breast cancer is one of the significant deaths causing diseases women around globe. Therefore, high accuracy in prediction models vital to improving patients’ treatment quality and survivability rate. In this work, we presented a new method namely improved balancing particle swarm optimization (IBPSO) algorithm predict stage breast using unbalanced surveillance epidemiology end result (USEER) data. The work contributes two directions. First, design implement an (IPSO) avoid local minima while reducing USEER data’s dimensionality. improvement comes primarily through employing cross-over ability genetic as fitness function correlation-based guide selection task minimal feature subset sufficiently describe universe. Second, develop synthetic minority over-sampling technique (ISMOTE) that overfitting problem efficiently balance USEER. ISMOTE generates objects based on average with smallest largest distance from centroid object class. experiments analysis show proposed IBPSO feasible effective, outperforms other state-of-the-art methods; minimizing features 98.45%.</p>