Improving learning step in artificial intelligence with categorization

Abstract

The aim of this paper is to improve learning stage of machine learning. Input data are the main engine and first step in learning. Every small quality defect in the beginning can cause huge difference in result. There are two possible strategies to correct or filter. To correct data need to have some more knowledge about relations. For filtering we can calculate threshold and eliminate anomalies. It reduce amount of input data, what is also unwanted phenomenon. So we will mainly focus on correcting them and filling missing features in records. We expect more effective learning step which should result in better final (predicting) step of machine learning.