engleski

Neural Networks as a Method of Evaluating Residential Houses in Tvrdja, Eastern Croatia

Tvrdja is the baroque core of the city of Osijek, constructed as a fortress at the beginning of the 18th century. Its 106 edifices - originally built as residential houses and army barracks - form a unique example of fortification architecture in Croatia. As a result of a century long economic decline, several major population exoduses and serious war damages (1991-1992), a once dominant and prosperous part of the city became a decaying and run-down neighbourhood. Around 675 people live in Tvrdja today, most of them in extremely bad housing conditions. In this article, a neural network (an area in the field of artificial intelligence) is created in order to asses the present state of every building in Tvrdja. There are 22 input criteria (most important elements of the building as walls, windows, ceilings, floors, heating, sanitary conditions, installation) and one output criterium (general condition of the building) used to form the network, their marks ranging from 1 (very bad condition) to 5 (excellent condition). The result of the process, developed in three phases, is a neural network called NM 1 that defines the specific weight of every input criterium and its influence on the final house evaluation. The NM 1 network is also able to asses (predict) the general condition of any house given the values of its 22 input characteristics.

baroque architecture; residential houses; present state evaluation; neural networks

nije evidentirano