Eureka! Institutional Repository

Main target of this thesis is the implementation of machine learning algorithms in order to classify whether a web page has spam content or not. The algorithms we are going to use for this study are Support Vector Machines (SVM’s) in combination with linear, Gaussian and polynomial kernel and Naïve Bayes. Furthermore, we are trying a Convolutional Neural Network implementation (LeNet5) as it is proposed by Yann LeCun. The metric we use is Area Under the Curve with the best method in our case to score 78%. The code is written in Python and the environment we use is Eclipse Mars. SVM’s and Naïve Bayes functions can be found in Scikit library. Moreover, in order to achieve best results we wrote some methods for data preprocessing such as balance_dataset() and softmax(). Finally, in this thesis you can find a small review for collections about web spam.