SBC: A New Strategy for Multiclass Lung Cancer Classification Based on Tumour Structural Information and Microarray Data

Lung cancer has different subtypes which are different in cell size and growth pattern. Correctly classifying subtypes of lung cancer can help design specific treatments to increase patient survival rate. In this work, we propose an innovative Structural Binary Classification (SBC) strategy for classifying lung cancer subtypes using microarray data. The strategy is based on Gene Expression Programming (GEP) algorithm. Classification performance evaluations and comparisons between our GEP based model and common binary decomposition strategies, as well as three representative machine learning methods, support vector machine, neural network and C4.5, were conducted thoroughly on real microarray lung cancer datasets. Reliability was assessed by the cross-data set validation. The experimental results showed that GEP model with our strategy outperformed other models in terms of accuracy, standard deviation and area under the receiver operating characteristic curve. The work provides a useful tool for lung cancer classification based on tumour structural information.