Performance Analysis of K-Nearest Neighbor and Support Vector Machine in Anomaly Detection in Oil and Gas Pipelines

In the oil and gas industry, pipelines are used as a form of transportation for production. However, pipelines are often difficult to maintain and are prone to problems such as corrosion, leaks, and rust. Therefore, it is important to have a way or system that can automatically detect anomalies in order to maintain proper production processes. In this research, the focus is on the accuracy comparison between the K-Nearest Neighbor algorithm, and Support Vector Machine in detecting anomalies in gas transmission operational data. The goal of performing this comparison analysis is to offer trustworthy suggestions for detecting anomalies correctly so that it can lead to proactive maintenance strategies. Both models were able to detect almost all instances of extreme deviations as anomalies, with the one using all features having detected more instances that are way off the norm. A novel approach incorporating voting ensemble learning is introduced in which these algorithms are combined with a voting classifier to form a reinforcement learning framework. The resulting model was able to improve the precision of input model but made a decline to its recall. These findings contribute to the improvement of production processes in the oil and gas industry, providing valuable insight for decision-makers to implement more effective and robust anomaly detection systems.