Performance analysis of classification between a particular number and average using the same distance measurements.

Artificial intelligence techniques are used in many areas today to find solutions to different problems. Scientists are trying to solve some problems in people's daily lives using these techniques. To solve these problems, researchers often use some Machine Learning (ML) algorithms. It is important for researchers to have preliminary information about some metrics of machine learning algorithms in their scientific studies. In this study, k-Nearest Neighbors (k-NN) and Minimum Distance to Means (MDC) ML algorithm, which classifies according to distance measurement methods, were analyzed using the same distance measurement methods. k-NN, which measures distance according to a specified number of distance, and MDC algorithms, which measure distance according to the averages of the classes, were examined in terms of classification performance. The performance comparison of these two algorithms with 5 different distance measurements (Euclidean, Manhattan, Minkowski, Chebyshev, Hellinger) was made using 2 different datasets (Ecoli and Cardiotocography). For Ecoli dataset, the highest train score 100%, test score 84.85% from the k-NN algorithm, and the highest train score 73.78% and test score 76.12% from the MDC algorithm were obtained. For the cardiotocography dataset, the highest train score 99.94%, test score 84.52% from the k-NN algorithm, and the highest train score 69.50% and test score 61.57% from the MDC algorithm were obtained. According to the results of statistical experimental studies, the k-NN algorithm, which classifies according to a certain number, gave better results in both datasets. [ABSTRACT FROM AUTHOR]