engleski

Performance of machine learning methods in classification models with high-dimensional data

The paper investigates the performance of four machine learning methods: artificial neural networks, classification trees, support vector machines, and k-nearest neighbour in classification type of problem by using a real dataset on entrepreneurial intentions of students. The aim is to find out which of the machine learning methods is more efficient in modelling high-dimensional data in the sense of the average classification rate obtained in a 10-fold cross-validation procedure. In addition, sensitivity and specificity is also observed. The results show that the accuracy of artificial neural networks is significantly higher than the accuracy of k-nearest neighbour, but the difference among other methods is not statistically significant.

machine learning; support vector machines; artificial neural networks; CART classification trees; k-nearest neighbour; large-dimensional data; cross-validation

nije evidentirano