In contemporary network environment we are facing several cyber-attacks those are breaching out data on regular basis, So it our main responsibility to protect our network infrastructures. Several researchers are doing research in same era but still facing the problems. This research is focusing to analyze different machine learning models for better utilization of resources. Author combines supervised and unsupervised models for different types of network attacks and applying optimization algorithm to enhance the performance. UNSW-NB15 data set is used for model training, testing and validation of proposed hybrid models. Proposed model gain 99.45% of accuracy in detections of networks attacks including malware, phishing and DDOS. Further optimization enhance performance of model and showing effectiveness and utilization of models in real time intrusion detection