Spam Filtering in Social Networks Using Regularized Deep Neural Networks with Ensemble Learning

Abstract

Spam filtering in social networks is increasingly important owing to the rapid growth of social network user base. Sophisticated spam filters must be developed to deal with this complex problem. Traditional machine learning approaches such as neural networks, support vector machine and Naïve Bayes classifiers are not effective enough to process and utilize complex features present in high-dimensional data on social network spam. To overcome this problem, here we propose a novel approach to social network spam filtering. The approach uses ensemble learning techniques with regularized deep neural networks as base learners. We demonstrate that this approach is effective for social network spam filtering on a benchmark dataset in terms of accuracy and area under ROC. In addition, solid performance is achieved in terms of false negative and false positive rates. We also show that the proposed approach outperforms other popular algorithms used in spam filtering, such as decision trees, Naïve Bayes, artificial immune systems, support vector machines, etc.

Review Spam Detection Using Word Embeddings and Deep Neural Networks

“…It was also used as a baseline method in recent studies [20]. In this study, SVM was trained using the SMO algorithm with varying complexity C = {2 0 , 2 1 , … , 2 6 } and polynomial kernel function;  Random Forest (RF) is another well-performing benchmark method used in several comparative studies [10,11]. Here we used it with 100 random trees.…”

“…Note that only most relevant terms (attributes) were selected according to their weights vij. In agreement with previous studies [10,11], top 2000 terms were retained, including unigrams and bigrams as suggested in [26]. To obtain word embeddings, the skip-gram model was employed.…”

“…To further enhance the detection performance, we combine the generated word embeddings with bag-of-words in the second stage and train a deep feed-forward neural network (DNN) to classify spam / legitimate reviews. DNN is used to capture complex features hidden in high-dimensional data representations [9][10][11].…”

Email Spams via Text Mining using Machine Learning Techniques

“…A. Barushka and Petr Hajek (2018) in [27] depicted "ensemble learning algorithms with DNN" as the base learner is more accurate than "state-of-the-art spam filtering methods". From results we know that "bagging algorithm trained with DNNs" achieved high accuracy and best results on both classes.…”

Spam detection on social networks using cost-sensitive feature selection and ensemble-based regularized deep neural networks

“…This paper is a significantly extended version of [18]. The earlier version was limited to ensemble-based RDNNs without considering cost-sensitive learning and feature selection.…”