Architecture of ensemble neural networks for risk analysis

Abstract

Assembling of neural networks referred to as “Ensemble neural networks” consist with many small “expert networks” that learn small parts of the complex problem, which are established by decomposing it into its sub levels. Ensemble neural network architecture has been proposed to solve complex problems with large numbers of variables. In this paper, this architecture is used to analyze maintainability risks of high-rise buildings. An ensemble neural network that consists with four expert networks to represent four building elements namely roof, façade, basement and internal areas is developed to forecast the maintenance efficiency (ME) of buildings. The model is tested and the results showed good performance. The model is further validated using a real case study.