Prediction of User’s Interest Based on Urdu Tweets

Recently the role of text analysis and classification in Data Mining has become significantly important. It is being used in Spam filtering, News classification, Noise reduction, and particularly Social Media Analysis. With the growing popularity and advancement in technology, sharing personal opinion and sentiments on social media is now the most convenient way deliver a message regarding a certain topic. Twitter is a very fast growing site like other social network sites because it provides tailored suggestions to its users where Promoted Accounts suggest accounts that people don’t currently follow and may find interesting. In this research work, we have filtered information on twitter present in Urdu text to improve the tailored suggestions based on the interests of a user. We have used supervised learning and natural language processing approach for Urdu short text. For this purpose, dataset consisting of 6000 tweets is used to compute the accuracy results over multinomial Naïve Bayes (NB), Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) classifier. Results shows that SVM performs better on all Unigram, Bigram and Trigram features as compared to NB and KNN.