Numerous internet-based applications produce data streams. A data stream is a succession of available data that may shift over time. Data from the Internet of Things (IoT), social media, traffic lights, financial institutions, phone records, sensor data, banking and healthcare systems are examples of data streams. Obtaining knowledge from data streams presents defiances. The noise reduction process is one of them. Detecting and reducing noise is essential to improve the performance of any machine intelligence technique. In this paper, we create a performance evaluation of four different noise detection algorithms on data stream clustering that are implemented in MOA: Micro-cluster-based Continuous Outlier Detection (MCOD), AbstractC, SimpleCOD and AnyOut. We use each of these techniques to assess the quality of clustering produced by the clustering algorithm known as ClusCTA-MEWMA (Clustering based on Centroid Tracking and Exponentially Weighted Moving Average Chart Detection Method). We reference this algorithm as CM. We set up and monitor experiments using datasets created using a random data generator. The results evidence that CM gets better model quality with with Micro-cluster-based Continuous Outlier Detection algorithm.