Reducing the size of databases for multirelational classification: a subgraph-based approach
Journal of Intelligent Information Systems
| Hi-index | 0.00 |
This paper describes different methods for exact and approximate implementation of the MIN-FEATURES bias, which prefers consistent hypotheses definable over as few features as possible. This bias is useful for learning domains where many irrelevant features are present in the training data. We first introduce FOCUS-2, a new algorithm that exactly implements the MIN-FEATURES bias. This algorithm is empirically shown to be substantially faster than the FOCUS algorithm previously given in [Almuallim and Dietterich 91]. We then introduce the Mutual-Information-Greedy, Simple-Greedy and Weighted-Greedy Algorithms, which apply efficient heuristics for approximating the MIN-Features bias. These algorithms employ greedy heuristics that trade optimality for computational efficiency. Experimental studies show that the ;earning performance of ID3 is greatly improved when these algorithms are used to process the training data by eliminating the irrelevant features from ID3''s consideration. In particular, the Weighted-Greedy algorithm provides an excellent and efficient approximation of the MIN-Features bias.