A key component of network security is network anomaly prediction, which assists in spotting odd or perhaps harmful activity occurring within a computer network. There are a number of methods and strategies for predicting network anomalies, but machine learning (ML) is frequently utilized because of its capacity to identify trends and abnormalities in big datasets. In this research, hybrid Random Forest with Seagull Optimization algorithm for Network traffic Anomaly Detection (RFSO-NAD) is proposed. It has three stages for anomaly detection. In initial stage gathering data from the IEEE data port and preprocessing for data input to the ML models. In the second stage, important features and dimensional reduction is carried out by using PCA algorithm. In the third stage, ML models are trained and they are fine-tuned by Seagull Optimization algorithm. From the experimental results, the proposed RFSO-NAD give better performance (accuracy—97%) compared with other ML models.