gridl / predictive-model-building-with-support-vector-machines Goto Github PK

This project forked from pranov1984/predictive-model-building-with-support-vector-machines

This dataset is from Wisconsin Breast Cancer Database with following attributes, • Clump_Thick • Uniformity_of_Cell_Size • Uniformity_of_Cell_Shape • Marginal_Adhesion • Single_Epithelial_Cell_Size • Bare_Nuclei • Bland_Chromatin • Normal_Nucleoli • Mitoses • Class Class is a classification type of two tumours with 2 for benign and 4 for malignant. The goal of the project are given below Examine if the dataset has any missing observation and use appropriate method to impute those missing values? Use linear support vector machine algorithm for classification of the above tumour types. First train the model with 70% of the data and validate your result with remaining 30% of the data. Provide the accuracy metrics in both train and test dataset obtained through model object. Investigate for a best decision boundaries using kernel type as polynomial instead of having it in linear type and give your conclusion.