Designing Multi-Class Classifier for Imbalanced Dataset Usingimbalanced Ratio

What is it about?

In this paper, a novel multi-class classification for imbalanced data using Cellular Automata(CA) will be proposed. This classifier overcomes skewed result due to class distributionalong with no information loss, unlike other sampling based techniques. To ensure it, weprioritized equal class distribution of imbalanced dataset using imbalanced ratio. The entiretask is accomplished in three stages. The Cellular Automata based classifier demands anspecific type of CA. Such is designed using the concept of PE bits, reported in Algorithm1. Algorithm 2, depending on the specific datasets, identifies CA and selects CA with thebest accuracy using imbalanced ratio after rounds of iterations for each dataset. Whereas,Algorithm 3 is used to test the prepared CA and find the class for the specific dataset.The performance of reported classifier is compared with other standard classifiers basedon Accuracy, Precision, Recall and f1 score. For all datasets, the Precision and Recall areincreased by 20% − 25%, f1 score is increased in maximum datasets by 30%. Whereas, theAccuracy is nearly the same for all datasets.

Featured Image

Why is it important?

This paper introduces a novel approach using Cellular Automata for multi-class classification of imbalanced data, prioritizing equal class distribution without information loss, yielding significant improvements in Precision, Recall, and f1 score compared to standard classifiers while maintaining similar levels of Accuracy.

Perspectives