1 results listed
Twitter is one of the most widely used social networks
today. Because of its wide usage, it is also the target of various
spam attacks. In recent years, Spam Detection on Twitter using
artificial intelligence methods became quite popular. Twitter
Spam Detection Approaches are generally categorized into
following types as as User Based, Content Based, Social Network
Based Spam Detection. In this paper, a user based features based
spam detection approach is proposed. Using a publicly available
recent baseline dataset, 11 lightweight user based features are
selected for model creation. These features selected for ease of
computing and rapid processing since they are numeric or
boolean. The advantage of user based spam detection approach is
that the results are obtained more rapidly since they do not
contain complex features. Selected Features are verified, default
profile, default profile image, favorites count, followers count,
friends count, statuses count, geo enabled, listed count, profile
background tile, profile use background image. Feature verified is
used as a class label to measure success of the model. After the
feature selection, the dataset is divided into test and training data.
Following 10 common supervised machine learning algorithms
are selected for the experiments: (1) Support Vector
Classification, (2) K Nearest Neighbor, (3) Naive Bayes, (4)
Decision Tree, (5) Bagging, (6) Random Forest, (7) Extra Trees,
(8) AdaBoost, (9) Multi Layer Perceptron, and (10) Logistic
Regression. Success of the algorithms are measured using
following 9 metrics: (1) Accuracy, (2) precision, (3) recall, (4)
True Positive, (5) True Negative, (6) False Positive, (7) False
Negative, (8) Training Time, (9) Testing Time. The results were
compared according to the metrics above.
International Conference on Cyber Security and Computer Science
ICONCS
Anıl Düzgün
Fecir Duran
Atilla Özgür