A Dual-Stage Deep Learning Model for Cyber Data Classification and Analysis

Cyberbullying is on the rise, according to recent research, especially among adolescents. In this research, we provide an innovative approach to text-based automated categorization for the purpose of detecting cyberbullying. However, due to the problem of over fitting, a classifier cannot offer a locally converging solution. Given these constraints, we built a text categorization engine to do some preliminary work on the tweets, erasing noise and background details before extracting the required features and categorising without data overfitting. In this research, we create a new Dual Stage Deep Learning (DSDL) method by first using Deep Neural Networks for feature engineering and then employing Deep Q-Learning for classification. Results from validating the innovative Deep classifier show that it increases text categorization accuracy.