This project was completed as a part of the Honors portion of the Unsupervised Learning, Recommenders, Reinforcement Learning Course on Coursera.
Credit to DeepLearning.AI, Stanford, and the Coursera platform for providing the course materials and guidance.
In this project, I will be implementing an anomaly detection algorithm to identify abnormal behavior in server computers. The dataset consists of two features: throughput (mb/s) and latency (ms) of each server's response. With 307 examples of server behavior, the dataset remains unlabeled ( {𝑥(1),…,𝑥(𝑚)} ).
The main objective is to distinguish between "normal" and "anomalous" examples within the dataset. To achieve this, I will employ a Gaussian model to detect anomalies effectively.
Initially, I will work with a 2D dataset to gain a visual understanding of the algorithm's functioning. By fitting a Gaussian distribution to this dataset, I will identify values with remarkably low probabilities, which can be considered anomalies.
Subsequently, I will apply the developed anomaly detection algorithm to a larger dataset containing multiple dimensions, simulating a real-world scenario with more complex data. Through this report, I aim to demonstrate the efficacy of the Gaussian-based anomaly detection approach and its potential application in various server monitoring contexts.
