<p>Due to the evolution of the Internet of Things (IoT), a new paradigm of Cloud Computing has emerged as an extension known as fog computing. To lower latency, fog computing uses a variety of fog cells that are positioned at the network’s edge. Given that workload differences occur over time for IoT applications, it is necessary to automatically provide a sufficient quantity of fog resources. This will help avoid over and under-provisioning problems and simultaneously satisfy SLA and QoS requirements. To address such issues, a Fog-Enabled Auto-Scaling Technique (FEAST) has been developed in this study. This method predicts resource usage in relation to CPU, memory, disk, and network utilization using the LSTM model. A vital auto-scaling model has been developed to scale the host load dynamically based on prediction results. Comparing the suggested method to the current fog-based auto-scaling techniques, the experimental findings demonstrate that it improves cost, resource utilization, SLA violation, response time, and delay violation.</p>