Toward Identifying The Best Base Classifier in Multi Label Classification-an Investigative Study
Mazen S. Alzyoud, Raed Alazaidah, Haneen Alzoubi, Najah Al-shanableh, Mohammad Aljaidi, Sattam Almatarneh · 2023
Classification is a significant task in data mining, machine learning and data science. It aims to predict the class label for a new case accurately. Classification is of two types: Single Label Classification (SLC) and Multi-label Classification (MLC). In SLC, cases and instances are linked to one class label only, while in MLC, cases and instances could be linked to more than once class labels. Several SLC algorithms have been adapted and upgraded to handle MLC and showed different predictive performance. Hence, this paper attempts to identify and investigate the best SLC that could handle the problem of MLC with respect to three datasets and using three evaluation metrics. Moreover, the paper also aims to identify the best Problem Transformation Method (PTM) among five well-known methods. The results revealed that RandomForest and DecisionTable showed the best performance among the fifteen classifier. Also, Most Frequent Label (MFL) is the best transformation method among the five considered PTMs in term of Accuracy, Precision, and Recall.