Neural network approach for early cost estimation in construction projects

Abstract

Cost estimation of construction projects is a very complex process containing many variable factors that affects to the total cost of the project. Because there are many factors affecting the cost, which are used as inputs of a model to predict the cost of construction project. The objectitive of this study is to develop artificial neural network model for cost estimation problems in construction projects that will able to predict construction cost by considering different parameters. In this study, we used two problems in construction projects for that we have collected the dataset from Richa Yadav, Monica Vyas. To predict the construction cost in the early stage using supervised algorithm Multilayer Perceptron is used. We have performed Multilayer Perceptron (MLP) using different parameters such as hidden layer size and activation function with respect to epochs. The results are compared with the traditional model such as Regression and Artificial Neural Network in which MLP gives the better results as compared to both models. The result shows that the Multi-Layer Perceptron model with ‘elu' activation function gives the better result than other activation functions.