Abstract

This paper proposes an alternative to Quinlan's algorithm for forming classification trees from large sets of examples. My algorithm is guaranteed to terminate. Quinlan's algorithm is usually faster. I. The Nature of the Problem. We have a population of objects which we want to classify into two* * groups. We have a set of attributes, each with a snail finite number of distinct values, and a set of examples whose attributes have been measured and which have already been classified. Our goal is to find a rule, based on these examples, which we can use to classify other members of the population. In general this will lead us to statistical methods such as cluster analysis (Everitt, 1974, Kendall, 1975, Sturt, 1981a, Sturt, 1981b). The larger our collection of examples, the more likely it is that some of them are misclassified. (Sturt, 1981a) provides an excellent illustration of how improving the fit of a rule to the training set (beyond a certain point) can make It perform worse on the rest of the population. Even so, there are interesting tasks where the domain is formal, and we can be sure that we have all the information we need and that our classifications are correct. There are, however, interesting problems where the domain is formal rather than real, and we can be sure that all the relevant information is available and our classifications are correct. Chess positions and algebraic equations are two such domains. *This work was supported by a Commonwealth Scholarship. Computing was done on the SERC Dec-10 under grant CR/C/20826 **ID3 and ray algorithm are both presented in terms of two categories. The information heuristic generalises to any fixed number of categories, and so do the algorithms.

Citations

88	Discovering rules by induction from large collections of examples - Quinlan - 1979
1	Multivariate Analysis. : Charles Criffin &'co - Kendall - 1975