Automated cost estimating system using neural networks
This paper presents an automated cost estimating (ACE) system for low-rise structural steel buildings. ACE is developed and coded using Borland C++ 5.0 in a Windows '95 environment. An efficient user interface allows for cost adjustments and markup allocation to individual cost items. Moreover, a number of project reports can be displayed upon request. ACE integrates artificial intelligence technologies and traditional spreadsheet applications. Neural networks are used in the system development in view of their learning and generalization capabilities. These networks utilize the knowledge obtained from previously constructed building projects to generate conceptual cost estimates in a timely manner. As such, estimates can be generated for different project scopes and characteristics in order to meet the user's needs. ACE uses NeuroShell 2, a commercial software, in the design and training of the neural network models. The capabilities of ACE are demonstrated through an example application.