Özet:

The processing of big data and the removal of meaningful information from these data has become a major problem. Unknown and irrelevant data in real-life problems increases the complexity of the model that will be created whether the data in hand is problematic or not. Data size reduction approaches are applied to create simpler and lower-cost models. Information theory-based mutual information approaches are widely used in size reduction. The aim of these approaches is to ensure that the input in the subset that will be obtained from the data set has a minimum surplus with each other and maximum dependence on the output. However, the intuitive functions used in the recommended approaches show a stable approach to the changing problems while addressing the relationship between minimum excess and maximum addiction. In this study, a new mutual knowledge approach was proposed. The intuitive function used in the proposed approach calculates the effectiveness of the minimum surplus on choice, the mutual knowledge of the subjects with the class and the relationship between their mutual knowledge with each other. Similarly, for maximum dependence, both conditional mutual information and mutual information are calculated. Thus, the recommended intuitive function shows a dynamic approach to changing problems. The results of the test show the success of the proposed approach.

Anahtar Kelimeler: