Machine Learning Algorithms for Abnormal Entity Detection

Abstract

Abnormal entity detection is the process of finding patterns in data that do not follow normal patterns of behavior. False data points or abnormal entities that significantly distort the sample or push the boundaries of the data set are rare. Aberrant values usually indicate incorrectly recorded or incorrectly entered data points. In some cases, abnormal entities suggest that the pattern is too narrow and that the boundaries need to be corrected. However, in any case, the deviation is a kind of anomaly that can distort the results of the analysis. There are many types of anomalies that, when not detected, create pauses or interruptions in several domains. newlineRecently, Machine Learning (ML) has been widely used for various tasks. ML algorithms have been proven successful in a variety of fields across all domains. The basic principle of any ML algorithm is to train the model to perform a specific task. The ability to access data will have a significant impact on the system s construction and the AI approaches that are employed. The quality of the finished result will depend on the amount and calibre of data that are available. In this view, it is possible to claim that the primary factors influencing the development of products utilising AI technology are data availability and accessibility. The data varying across the domains bring different set of challenges and constraints. newlineThis thesis explores different domains with varying data and provides different machine learning approaches to detect the abnormal entities in those respective domains. The domains explored are pharmacovigilance, facial images and medical images. newline