Advances in Density-Based Outlier Detection Algorithms: Exploration of LOF with Experimental Analysis

One of the decisive steps for refining the competence of machine learning algorithms is outlier detection which has bought considerable attention across many research fields as well as various application areas. Algorithms that deal with outlier detection are usually either density based, or distance based. The outlier detection algorithms are being developed while addressing the essential characteristics such as effectiveness, distributed environments, accuracy in addition to high dimensional data. Notably, many algorithms have limits as well, such as sensitivity to parameter choices and difficulties in high-dimensional or complicated data environments, even while they excel in certain areas like local density detection or computing efficiency. This review presents an outline of density-based outlier detection algorithms for instance LOF and other prominent approaches that have been proposed over the past two decades. Every single algorithm is evaluated according to its methodology to detect outlier. The review illustrates how these algorithms have been developed over time, giving their strengths as well as restrictions. This paper also presents a case study concerted on the parameter analysis of LOF to emphasize upon their real-world relevance in addition to the identification of the possible areas for the improvement.