Abstract

Fuzzy-based models and Artificial Neural Network (ANN) based systems have provided effective tools for addressing uncertainties in decision-making. Uncertainty, as an ineradicable part of construction projects, justifies the utilization of such intelligent systems in the construction industry. In the past few years, these systems have been widely applied to develop forecasting models in the construction management area. The estimation of productivity of construction operations, as a basic element of project planning and control, has become a remarkable target for forecasting models. A glimpse into this interdisciplinary field of research exposes the need for a system, which 1) studies the effect of qualitative and quantitative variables on construction productivity, 2) improves the previous models in terms of accuracy of estimation, 3) is able to clearly illustrate the reasoning process, 4) considers the interdependence of input variables; and 5) has the capability of dealing with both crisp and linguistic input variables.
The main objective of this research is to develop a hybrid intelligent system for estimating productivity of construction operations based on several qualitative and quantitative factors. Among all models developed for productivity estimation, those established based on the functional relations and controlled by a specific number of control rules are more compatible with the human reasoning and logic. Neural-Network-Driven Fuzzy Reasoning (NNDFR) structure, as one of such models, displays a great potential for modeling datasets among which clear clusters are recognizable. The lack of compatibility between conventional NNDFR and fuzzy clustering algorithms together with the insufficient attention paid to the optimization of number of clusters in this model, created a potential area for further research. Thus, the main contribution of the proposed model is to develop a modified NNDFR system for modeling construction data. It forms a nonlinear multi-dimensional membership function, which internally combines all fuzzy variables via Fuzzy C-Means (FCM) clustering. An ANN is then trained based on the clustering process to automate this step. While the clustering step constitutes “IF” parts of the rules, “THEN” parts are built by another set of ANNs. In addition, the parameters of the proposed system are optimized by Genetic Algorithm (GA) to fine-tune the system for the highest possible level of accuracy. The model is also capable of dealing with a combination of crisp and linguistic input variables through the use of a Hybrid Modeling Approach, which is based upon the application of alpha-cut technique.
The proposed model is further verified through simulating a construction operation considering qualitative and quantitative factors where a considerable improvement in the estimation accuracy is witnessed. Several models are developed using ANN, Adaptive Neuro-Fuzzy Inference System (ANFIS), conventional three-cluster NNDFR and the Genetically Optimized NNDFR. The proposed model showed 83%, 72% and 69% improvement over ANN, ANFIS and conventional NNDFR, respectively, in terms of Mean Squared Error (MSE). The developed model helps researchers and practitioners use historical data to forecast productivity of construction operations with a level of accuracy greater than what could be offered by traditional techniques.