Published Paper Details:

In recent years, the ever-increasing popularity of social networks like Twitter have become an important source for real-time information which offers unprecedented opportunities to aggregate people and news dissemination, since it is so rapidly updating which makes it easy to fall into the trap of believing everything as truth which opens new modalities for cyber-crime perpetrations and people become a prime target of spammers. The paper proposes the various classification learning techniques methods for the detection of the Twitter- spammers by using Support Vector Machine, Na�ve Bayes, extreme gradient boosting, Random Forest and neural network. The paper mainly focusses on the crispy set and not on the fuzzy set because the problem is to detect whether a tweet is a spam or not, so the annotation which is used to classify the tweets, �0� for the non-spam and �1� for spam thus it is a binary classification problem which is included in the crispy set. Initially, the tweets were first extracted manually and extracting features for classification. After selection of the features, five machine learning algorithms were cross-validated to determine the best base classifier for spam detection. Followed by handling class imbalance problem in which the proportion of the positive class is very small in comparison to the negative class which is very large. This problem is dealt with techniques which helps in improving the performance of the classifier even with the imbalanced dataset. Results showed that the proposed approach has Random Forest base classifier as the best traditional algorithm along with this data sampling technique Random undersampling showed the best sensitivity value of 93.4%.