Divulgação - Defesa Nº 227

Aluno: Caio Augusto Sobreira Melquiades de Lima

Título: “Isolation Forest-based semi-supervised Anomaly Detection of multiple classes”.

Orientador: Prof. Fernando Buarque de Lima Neto

Data-hora: 29/Dezembro/2020 (16:00h)
Local: Escola Politécnica de Pernambuco – Formato Remoto ((http://meet.google.com/hve-mmku-die))


Resumo:

“"In the age of Big Data, previously defining rules for finding all the uncountable patterns of events of interest, in complex and critical applications, is infeasible, if not impossible. A big challenge in this kind of scenario is to find previously unknown data that do not conform to the expected behavior. Anomaly Detection methods are techniques well-suited for tackling this category of problem due to its ability of identifying data that significantly deviate from an expected pattern, many times without any training example. Isolation Forest is a state-of-the-art technique in the unsupervised Anomaly Detection area, this, in addition of having a low computational cost. However, unsupervised Anomaly Detection many times suffer from high false-positive rate and high false-negative rate. Semi-supervised techniques can significantly improve the unsupervised algorithms with low human effort, and can also aggregate semantic knowledge into the models. Hybrid Isolation Forest is an Isolation Forest semi-supervised variation, which aggregates known anomalies, but only in a single class. This research proposed a Hybrid Isolation Forest-based model capable of aggregating known anomalies in distinct classes. The research work included a comprehensive literature aiming at identifying possibilities of aggregating expert feedback into Anomaly Detection techniques. The proposed model was accompanied by a very large set of experiments presented. Our proposed multiple anomaly classes semi-supervised model showed better performance in some distinct datasets and scenarios and showed an ability to significantly improve the underlying unsupervised algorithm.”

Go to top Menu