An accuracy analysis of classical and quantum-enhanced K-nearest neighbor algorithm using Canberra distance metric

The k-nearest neighbor (kNN) algorithm is a widely used machine learning technique for classification tasks. Recent advancements in quantum computing have introduced quantum-enhanced kNN (QKNN) algorithms, which offer potential improvements in computational efficiency. This study aims to perform a comprehensive accuracy analysis of classical and quantum-enhanced kNN algorithms using the Canberra distance metric across a variety of datasets. We evaluated the performance of both classical and quantum kNN algorithms on eight diverse datasets: Wine, Breast Cancer, Diabetes, Seeds, Iris, Raisin, Obesity, and Liver. For each dataset, we compared the accuracy of classical kNN and QKNN models using the Canberra distance metric with k-values of 3 and 5. The Canberra distance metric proved to be effective, yielding high accuracy rates for both classical and quantum models. Notably, the classical kNN achieved higher accuracy rates compared to QKNN, with improvements in accuracy observed as the k-value increased from 3 to 5. Among the eight datasets, the breast cancer and Iris datasets perform with high accuracy of 93.85% and 93.33% for quantum-enhanced K-nearest neighbor with \(k = 5\) $k=5$ using the Canberra distance. At the end our work is able to figure out the important metric and its accuracy level as well. The detailed results and comparisons done are elaborated in our work.