Digital environments most often succumb to system faults, thereby mitigating their resilience and decelerate performance. While there are various methods to identify these system faults, the classification based on certain attributes using deep learning is less explored. The stratification of faults in systems is crucial in order to efficiently maintain the reliability of technological incorporations on a quotidian basis. The existing work relevant to this study delineates the identification of malicious interventions but lacks in scrutinizing the loopholes relevant to transactional process failure, and the usage of real-time data for explicit identification of system faults. This paper proposes to incorporate real-time data for efficient analysis of system fault by comprehending the feature statistics associated, to further delve into loading the dataset to the domain of deep learning networks. This study effectuates the use of pre-trained deep learning models such as the ResNet, DenseNet, and VGG-16 for classifying transaction statuses as either failure or success based on the defined parameters. Although the networks utilized have previously been explored for various research, the novelty of this research remains the leveraging of the real-time dataset into these sophisticated deep networks. The simulation is carried out in MATLAB, and the results indicate that pre-trained networks can significantly accelerate the process of fault identification stratification, thus providing robust solutions for real-time system monitoring and maintenance.