DAC-BiNet: Twitter crime detection using deep attention convolutional bi-directional aquila optimal network

Recently, the crime rates in social media are rising rapidly and it produces serious problems to the users. From the past few years, twitter platform has been gaining higher attention due to its effective services. But, the growth of crimes in twitter generates several issues and it redirect users to malicious or phishing websites. Thus, in order to identify various illegal activities on twitter, the proposed study introduces a robust Deep Attention Convolutional Bi-directional Aquila Optimal Network (DAC-BiNet) model. For effectively processing the proposed network, the raw inputs are pre-processed initially to reduce the noises through diverse steps like segmentation of sentences, punctuation removal, tokenization, removal of stop words, acronym and slang correction, lemmatization, lower casing, stemming, removal of hashtag and URLs. From the pre-processed data, the significant features are extracted using ITF-IDF (Improved Term Frequency-Improved Document Frequency), Feature hashing and Glove Modelling methods. Then, the feature dimensionality issue is solved by grouping the features into clusters by Possibilistic Fuzzy LDA (Latent Dirichlet Allocation) based clustering method. Finally, classification stage is initiated to identify the crime tweets in the given inputs through proposed DAC-BiNet model. The proposed study used python platform for simulation and the efficiency of proposed model is measured by evaluating various performance matrices and comparing with other existing techniques. The obtained simulation results proves that proposed model gains enhanced accuracy of 98.23%, precision of 83.86%, recall of 90.05%, specificity of 98.86%, and F1 score of 86.84%.