NETWORKS ATTACK PREDICTION USING RANDOM FOREST ALGORITHM

The purpose of this article is to compare conventional machine learning with modern deep learning approaches for the task of detecting anomalies in self-organizing networks. While deep learning has gained a significant move, especially in application scenarios where large volumes of data can be collected and processed conventional methods can nevertheless offer strong statistical alternatives, especially when the right methods are used learning representations. For example, support vector machines have previously shown superior potential in many binary classification applications and can be further used with various representations such as single-class learning and data augmentation In four different application scenarios, we are presenting evidence for the first time that conventional machine learning can achieve superior performance compared to prior deep learning approaches by an average of 15%, as demonstrated on a publicly available dataset that has been previously published. Our results further indicate that the computational speed has improved by almost two orders of magnitude and an order of magnitude reduction in trainable parameters, conventional machine learning provides robust alternative for self-organizing 5G networks, especially when execution and detection times are critical.