Modern transportation’s contributions to air pollution and carbon emissions are integral to the escalation of planetary warming. Accurate traffic flow prediction, enabled by cutting-edge computing technologies such as Internet of Things (IoT), Cloud Computing and Artificial Intelligence (AI), may efficiently address issues like traffic congestion, pollution, and climate change. Monitoring traffic flows in real time is an essential part of ITS since it provides accurate information for traffic management and optimisation. In many cases, monitoring traffic flows requires collecting and analysing data manually, which is not only time-consuming but also taxing on available resources. Precise traffic forecasting is challenging due to the high geographical and temporal correlation of traffic data. In order to describe geographical dependencies, current approaches rely on graph neural networks that use static, predetermined spatial adjacency graphs of traffic networks. The hybridization of Long Short-Term Memory (LSTM) with the popular Support Vector Machine (SVM) classifier for traffic flow prediction is the core idea of the study to overcome these concerns. The given approach has two primary phases: counting vehicles and forecasting traffic flows. The method used to tally cars in this model is called Fully Convolutional Redundant Counting (FCRC). In addition to this the integration of Hybrid Cat Particle Swarm Optimisation (HCPSO) technique is deployed for weight update in the proposed LSTM. Extensive studies on four actual big traffic datasets demonstrate that our model regularly beats all baseline methods.