Intelligent Estimator

Purpose

This research explores the development of a cost estimating system which integrates artificial intelligent (AI) learning algorithm, probability estimation and knowledge-based system as a user-friendly and reliable tool for forecasting the cost of services elements at the early stage of design when information about project scope is limited. The research project combines knowledge from the field of construction economics, quantity surveying, cost modelling, statistics, construction technology, and information systems to develop an easy to use tool that can assist in decision-making with regards to the costs of services elements during the planning and design of building projects. This first stage of the research is focused on the cost of ‘Electrical Services’, and in particular power wiring, light wiring and cable pathways.

The specific objectives are:

1. to propose an integrated cost estimating system for forecasting cost during design stage.
2. to explore the distribution pattern of the cost of various services elements, and in that regards, identify the significant cost components using the pareto principle.
3. to identify the variables that influence the cost of services elements.
4. to assess the impact and the significance of each individual variable on cost.
5. to develop, test and validate a learning model that can be used to reliably forecast the cost of services elements during the early stage of design when limited  information is available about project scope.
6. to integrate risk analysis, probability estimating and knowledge-base with the learning model thereby providing practitioners with a more robust, reliable and effective tool for estimating cost during design.

In the first phase, the study objectives were addressed using electrical services for building, and in particular, power wiring, light wiring and cable pathways.

Analytical approach

The research is based on data mining of over 200 building projects in the office of a medium size electrical contractor. The objectives were achieved in twelve stages:

• Define the output variable;
• Identify the input variables;
• Data mining from project records;
• Evaluate the distribution pattern of the cost of components;
• Data processing;
• Train the learning model;
• Test the learning model;
• Evaluate the performance of the learning model;
• Conduct sensitivity analysis;
• Develop, test, and validate regression model;
• Compare the performance of learning model with regression model;
• Develop a user interface for the learning models.

Findings

The findings show that cost forecasting models based on artificial neural network algorithm are more viable alternative to regression models for predicting the costs of light wiring, power wiring, and cable pathways. The prediction errors achieved are 6.4%, 4.5%, and 4.5% for the three learning models developed whereas the regression models were insignificant. The data did not fit any of the known regression distributions. A user interface was established to talk to the validated learning models.  This was performed by consolidating incoming and outgoing weights of the nodes in the network. These weights were bundled in a form of a software code (WSC) in Neurosolutions package. The UI has been arranged to capture inputs from the user. The application - Intelligent Estimator, is an easy to use tool that estimators can use when estimating cost at the conceptual design stage of building projects.

Practical implications

Intelligent Estimator software application is important because it can be used by construction professionals to reliably and quickly forecast the costs of power wiring, light wiring and cable pathways using building characteristics that are readily available or measurable during early stage of design i.e. fully enclosed covered area, unenclosed covered area, internal perimeter length and number of floors.

Future research

Future research will build on the findings of the first phase. The learning model developed will be integrated with risk analysis, probability estimating and knowledge-base thereby leading to a more robust, reliable and effective tool for making design decisions on services elements. The system will be developed for other building elements.  When completed, the system could greatly reduce the time and resources spent on cost estimation at the early stage of design as well as provide a benchmark to compare detailed cost estimates. The potential is that it could allow the design team to quickly and efficiently conduct economic evaluation of the costs of many alternative design solutions using ‘what if’ analysis. The risk analysis module will help practitioners evaluate the uncertainties associated with cost estimates. This should provide useful information for setting contingency allowance on projects. In addition, future research would link the learning models to 3D building information model (BIM) so that the input of the learning model can be automatically extracted and estimates can be quickly and reliably prepared during design.