Investigation of effect of reducing dataset´s size on classification algorithms

Data mining is now one of the most active field of research. Extracting those nuggets of information is becoming crucial and one of its important technique is classification. It helps to group the data in some predefined classes. Various techniques for classification exists which classifies the data using different algorithms. Each algorithm has its own area of best and worst performance. This paper concentrates on the four most famous algorithms, i.e., Decision Tree, Naïve Bayes, K Nearest Neighbour and Genetic Programming and the effect on their performance of time and accuracy when the number of instances are incrementally decreased. This paper will also investigate the difference in result when working with binary class or multiclass datasets and suggest the algorithms to follow when using certain kind of dataset.