During the past few years, the virtual technology used in the cloud has become unsuitable for service delivery, with the emergence of containers technology and the spread of its use throughout a wide range of cloud service providers because of the ease of use and provision of the resources used. The institutions have become widely seeking to develop this technology to suit the different needs to provide a good service for the end-user. In this paper, we use machine learning to improve the container orchestration process. Our approach focuses on getting containers cluster resources idle as much as can by scanning and clearing malicious and unwanted fake load to enhance the workers load then using machine learning models to predict loads in advance. Hence, the Auto-Scaler module begins to auto-scale the number of resources to meet the cluster's workload, leading to efficient use of resources.