Selection of attributes has been the current emerging research area for a long while and much work has been finished on. With the making of tremendous dataset and the resulting necessities for good machine learning systems, new issues emerge and ways to deal with feature selection are the area of research and in need. Preprocessed and generated datasets of multimedia event detection with numerical values is the input for our research. This paper recognizes the contribution of each attribute and combinational attributes towards the accuracy of the classifier model. The analysis is done using LIBSVM tool and different kernel modes using Support Vector Machine (SVM) classifier. After the analysis is done, we will get to know that which attribute yields the maximum accuracy in classifying the instances. © Published under licence by IOP Publishing Ltd.