Designing computer-assisted problem-based learning (CAPBL) environment for performance analysis of isolation forest algorithm

Abstract

Anomaly detection is important applied in various fields of application to determine errors in the system. By detecting it can minimize losses on the system. Anomaly detection can be done using several algorithms. One such algorithm is the Isolation Forest algorithm. Isolation Forest Algorithm is an efficient and effective algorithm in detecting anomalies. Isolation Forest has better performance than other algorithms in terms of execution time, especially in large datasets. Although the Isolation Forest algorithm has many advantages, there are still very few tools that provide this algorithm. One tool that provides the Isolation Forest algorithm is Scikit Learn. However, using special tools such as Scikit Learning requires sufficient time and experience to be able to manage the features of the tool. Thus, in this study the authors developed a web-based application that is used to assist users in assisting and improving the performance of Isolation Forests by using sample data sets, setting the parameters of Isolation Forests, visualization, and evaluation. This application was developed using the waterfall process model and CAPBL concept. In determining the features of the application, an analysis of features is based on the CAPBL concept with the business process of CRISP-DM. The software testing result shows that this application is fit to be used as a new solution to facilitate users who want to learn and analyze the performance of Isolation Forest.