Mean black widow-based optimisation feature selection and enhanced kernel-based SVM classifier for cyberbullying twitter data

People are using online social networks (OSNs) to interact with others worldwide and discuss their preferences, which can lead to cyberbullying. A cyberbully sends threatening or damaging messages online. Social media cyberbullying must be found and stopped. For this, AI computations including pre-processing, highlight extraction, and grouping were built. Pre-processing Twitter data with tokenisation and stemming removes noise. Highlight extraction uses common data and SAE. Several AI classifiers detect cyberbullying. Due to the complexity of cyberbullying, identifying the best selection of features is difficult, affecting detection system scalability. Thus, this study selects and optimises parameters to improve cyberbullying categorisation. Twitter data is collected and pre-processed to remove stop words and other unwanted words. TF-IDF and SAE are used to extract features from pre-processed data. MBWO then selects the component subset that best reduces information dimensionality. To determine cyberbullying events, an enhanced bit support vector machine (EK-SVM) classifier uses these highlights. The study compares MBWO with EK-SVM against KNN, ANN, and RF in experimental findings. Classifier performance is measured by precision, recall, sensitivity, specificity, FPR, FDR, miss rate, and accuracy.