Video surveillance is increasingly being adopted for ensuring safety and security both in public and private places. Automated prediction of abnormal events like theft, robbery, murder etc from continuous observation of surveillance videos is a multidisciplinary study involving computer vision, deep learning and artificial intelligence. Deep learning-based video analysis and categorization is the most researched topic. Many deep learning models based on Long Short Term Memory are proposed for automated prediction of abnormal events. There are two contributions in this paper; the first contribution focuses on five models - Resnet, VGG16, VGG19, 3DCNN and Inception V3. The second contribution has proposed an approach called Recurrent-Residual-Inception V3 (RRIV3). Advantage of RRIV3 is performance will not get effected more by removal of any residual block. This work does a performance analysis of six LSTM based deep learning models for abnormal event prediction from surveillance videos before and after performing preprocessing. Deep learning models are combined with LSTM for the prediction of abnormal events from past observation of events in the video stream. These six models are executed against different benchmarked abnormal event detection datasets one among them is UCF-Crime dataset and efficiency is compared in terms of accuracy, precision, recall and execution time. It is observed that Recurrent-Residual-Inception V3 with LSTM performs better than other models with training accuracy of 90% and test accuracy of 85% compared to other models. The execution time is 20 milliseconds compared to other models.