Supporting Slab Formwork Selection with Different Types of Classifier Ensembles

Formwork is the largest single cost item in cast-in place reinforced concrete structure. A properly selected formwork system, in particular slab formwork, affects pace of concrete works and therefore has an influence on the efficiency of the building performance as a whole. When the formwork items are leased to perform building works, choosing the right system gains in importance. This is why the problem of formwork selection has been discussed widely since early 90's until now by various researchers all over the world. The present paper introduces different classifier ensembles as a tool that can be applied to solve formwork selection problem and presents the mathematical model of the decision problem. In literature, boosted classifiers were already used to solve formwork selection problem by Korean researchers. However, boosting (AdaBoost) algorithm is not the only one that can be applied in the subject matter. In this paper different classifier ensembles are investigated. The comparison includes both different aggregation methods and different types of classifiers that are being combined. In order to compare ensembles' accuracy, the data collected on the national construction sites are used. The paper also includes remark and indications that improve classification precision. (paper)