• Dr Snehal Joshi

Support Vector Machine algorithm is very efficient particularly in case of non-linearly separable data as well as efficiently implemented where the data is labeled or non-labeled. However, at same time, the major constraint that affect performance of Support Vector Machine algorithm is choice of Kernel. Current study is pertaining to selection of appropriate Kernel in process to obtain optimal hyperplane. The study is based on a large dataset containing feature sets of human cancer cells characteristics. Dataset is consists of attributes related to human cells and their relevant properties. It contains records; which are classified as malignant or benign in scale of one to ten with various medical diagnostic procedures. Mapping of the data in higher order of polynomial is performed using four kernels, which include Linear, polynomial, Radial basis Function (RBF) and Sigmoid. Initially Linear kernel is used for mapping the data. The model is fit using kernel type linear using train set. This model is applied on cross-verification dataset to measure the precision, recall and f1-score. Accuracy is measured for cross-verification dataset, which is followed, by measuring accuracy for test-data-set. Similar process is repeated using other three types of kernels namely Polynomial, Radial bias Function and sigmoid. Confusion matrix is used for obtaining True Positive, True Negative, False Positive and False Negative results for cross-verification-data-set as well as for test-data-set. F1-score and jaccard-similarity-score are used to find the accuracy using all four kernels. Obtained results using all four kernels are compared and the variations are measured. This comparison provides the best-fit model, which is obtained using the specific kernel. On obtaining and comparing AUC (Area under Curve) and compare for performance evaluation, very interesting and unexpected result obtained. This is very significant in evaluation of performance measurement of the model. This study determines that selection of kernel is problem specific and accuracy or AUC cannot be sole criteria to decide the performance evaluation of the model.

Support vector machine, machine learning, Kernel, Linear, RBF, Polynomial, Sigmoid, AUC

"CANCER CELLS CLASSIFICATION AND MODEL PERFORMANCE EVALUATION USING SUPPORT VECTOR MACHINE BASED CLASSIFIER APPLYING DIVERSE KERNEL ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.5, Issue 12, page no.87-94, December-2018, Available :http://www.jetir.org/papers/JETIR1812E38.pdf

"CANCER CELLS CLASSIFICATION AND MODEL PERFORMANCE EVALUATION USING SUPPORT VECTOR MACHINE BASED CLASSIFIER APPLYING DIVERSE KERNEL ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.5, Issue 12, page no. pp87-94, December-2018, Available at : http://www.jetir.org/papers/JETIR1812E38.pdf