Cyber bullying detection that are prevailing commonly in social networks like Twitter is one of the focussed research area. Text mining and detecting cyber bullying has several research challenges and lot of research scope to work with. This research work makes use of supervised feature selection by ranking method in order to choose the features from the tweets. After that extreme learning machine (ELM) classifier is employed in order to perform the detection of cyber bullying tweets. Performance metrics such as accuracy and time taken for classification are chosen in order to evaluate the efficiency of the classifiers namely ELM and the proposed SFS-ELM. Implementations are done in MATLAB tool. From the obtained results it is evident that the proposed SFS-ELM produces better results than that of ELM.