In recent years, the services offered by cloud computing gained prominence due to their scalability and cost-effectiveness. The increasing number of cloud data centers, driven by the demands of numerous users, has raised both economic and environmental concerns. However, cloud computing faces several challenges related to forecasting resource requirements for larger workloads. However, many recent approaches have introduced effective prediction techniques they faced issues related to an inability to handle multiple variates at prediction. Moreover, the prediction accuracy diminished when the number of tasks increased along with the enhanced time of execution. So, this research aims to put forward an effective approach using the hybrid approach of Functional Link Neural Network (FLNN) and Convolutional Neural Network (CNN). The output from a hidden layer of FLNN is fed into the input layer of CNN which acts as the point of integrating two types of architectures and helps in developing the proposed hybrid model for prediction. The data is obtained from Google cluster trace and the pre-processing takes place using one hot encoder and standard scalar. After this phase, feature selection is performed using Chaotic-PSO and the prediction of resource utilization is carried out using the proposed hybrid model. The efficiency of the suggested model is evaluated with a Functional Link Neural Network along with a hybrid approach of Genetic Algorithm and Particle Swarm Optimization (FLNNGAPSO), BG-LSTM and clustering-based stacked LSTM. The CPU usage based on RMSE of the suggested model for univariate is 0.001298 whereas the CPU usage of existing BG-LSTM and Clustering-based Stacked LSTM is 0.16 and 0.00558.