An optimized Clisteer Long-Range Kookaburra Attention Module Steerable Convolutional Neural Networks for Crime Predictions in Smart Cities

Crime prediction in smart cities is a complex yet promising field that leverages data and technology to enhance public safety. In the previous papers several deep learning algorithms are used to predict the crimes in smart cities but does not provide sufficient results that is accuracy is decreased, error rate, computational time, cost is increased. To overcome these limitations this work is proposed. In this manuscript, an optimized Clisteer Long-Range Kookaburra Attention Module Steerable Convolutional Neural Networks (2CLR-KoobAtMSNet) for Crime Predictions in Smart Cities is proposed to predict the crimes in smart cities. In this, to predict the crimes the input dataset is taken from crime data for city of Chicago city. Then these data are cleaned using the Adaptive Guided Side Window Box Filter with Multiple Layers (AGS-WBF-ML). Then the features are extracted using the Attention based Swin Transformer (AST). After that, the features are selected using the Dung Beetle Optimization Algorithm (DBOA) for crime prediction. Finally, the crimes are predicted using the 2CLR-KoobAtMSNet. The 2CLR-KoobAtMSNet model was implemented on the Python platform and evaluated using four datasets, achieving remarkable results with 99.9% accuracy and 99.8% recall, outperforming existing methods. These results demonstrate the model’s exceptional effectiveness and opportunities for advancement in customized stress-reduction strategies.