Data mining methods with trees

Present world is characterized by ever growing volume of data collected and saved into data- bases. Data often can‘t be analysed by using standard statistical methods because they contain many missing figures or are in qualitative units, and because some databases are in very wide usa- ge. Each organization must be able to extract important information from an extensive database. These were the main reasons why data mining was initiated. Tree structures are used in many diverse areas. Tree structures are frequently used in statistical data analysis, particularly in data mining. This paper describes decision trees, their data structure and their implementation in statistical data analysis. Decision trees offer a non-algebraic method for partitioning data. Using decision trees is attractive because they offer visualization, simplicity of interpretation and high accuracy. We can utilize them to solve various classificatory and predictive exercises. They are a perfect instrument to help managers in the decision-making processes. The decision trees are also used to form different groups of clients in order to prepare special offers and campaigns. Their potential lies in the ability to predict potential debtors on which may be decided whether to give or reject a loan or insurance to a particular costumer. The decision trees are also used to predict the potency for a new product designed for targeted customer, detect an insurance fraud, or foretell the number of people, who want to attend the competition and so on. There are quite a few algorithms, which have been described and are being used to form deci- sion trees. The following two are among the basic ones: algorithm ID3 and its improved version C4.5. The author is J. R. Quinlan. The first one is very illustrative and it is really important in order to acquire the basic understanding in decision trees problematic. The article contains an example of this ID3 algorithm application