As data is growing rapidly day by day due to wide usage of social media, this social data will help business analyst as well as researchers to get the feedback about any service or product. Analysis of social media data can be used for many purposes like friend recommendation, product or service recommendation etc. There is need of such algorithms which will classify data accurately and will provide analysis results accurately. Also these algorithms need to scale rapidly with data sets. In this paper, Different phases of social media data mining are identified. We have studied algorithms whose training time remains same though the data size increases. We have implemented decision tree algorithm and compared it with other machine learning algorithms like NaiveBayes, Adaboost etc. It is found that decision tree algorithms gives accurate results as compared to other algorithms. In addition to that, out of different decision tree algorithms random decision tree (RDT) algorithm performs best than other decision tree algorithms like C4.5 or ID3. © 2015 IEEE.