Decision Tree based Classifiers for Large Datasets

In this paper several algorithms were developed for building decision trees from large datasets which overcome some restrictions of the most recent algorithms in the state of the art. Three of these algorithms were designed to process datasets described exclusively by numeric attributes, and one more for processing mixed datasets. The proposed algorithms process all the training instances without storing the whole dataset in main memory. Besides, the developed algorithms are faster than the most recent algorithms for building decision trees from large datasets, obtaining competitive accuracy rates.