DeepGuard-HDL: A Hybrid Deep Learning Model for Uncovering Spatial Patterns of Online Harassment

Online harassment has developed into a major digital platform problem which affects millions of users throughout the globe. The research introduces DeepGuard-HDL which utilizes Hybrid Deep Learning Models and neural networks and recurrent units with attention mechanisms to detect spatial-temporal patterns in online harassment across integrated systems. DeepGuard-HDL uses textual parameter features with geographic and time-based data to achieve precise harassment detection within different internet platforms. The aggregator processed the dataset by performing text normalization while also applying geo-tagging to data collected from social media platforms and forums as well as complaint databases. DeepGuard-HDL produced outstanding performance according to experimental results by obtaining 98.44% accuracy on the test set surpassing both traditional models including Logistic Regression and Support Vector Machines (SVM) as well as isolated deep learning models CNN and RNN. Tests revealed the model exhibited precision at 0.95 and recall at 0.94 together with a 0.96 AUC score which indicates its stable performance. DeepGuard-HDL delivered both effective classification abilities and valuable hotspots information which enabled policy makers and platform moderators to create actionable insights. The obtained results demonstrate how hybrid deep learning methods effectively handle complex matters involving online harassment.