| Summary: | This paper presents a visualization technique designed to simplify the process of comparing machine learning classification results and subsequently improving interpretability. The main goal of this study is to construct an end-user interface capable of visualizing the outcomes of classification results generated by multiple machine-learning algorithms. To accomplish this, we conduct an empirical study that assesses the performance of machine learning algorithms on both the standard Iris dataset and a real-world zakat financial aid dataset. Our research involves an examination of the behavior of machine learning algorithms on the Iris dataset, providing practical insights into their functionality. We specifically focused on J48, PARTS, and Random Tree, chosen for their ease of implementation as ruleset in various programming languages. Additionally, we incorporated Neural Network into our study, considering its growing popularity in deep learning and deep neural network research. The interface incorporates interactive charting functionality to visually represent the outcome of each machine learning models. Based on the implementation, the proposed methodology has shown potential in aiding visual comparisons of machine learning algorithms, showcasing both interactivity and usability. The methodology could also prove beneficial for research aiming to optimize user-interface elements that represent input features, addressing concerns about overcrowding users' screens. © 2023 IEEE.
|
|---|