ABSTRACT

In this research stud) , the investigation of building details on the construction project cost (Naira) and duration (days) using artificial neural network (ANN) which possesses the ability to generalize complex input-output relationships between given datasets was carried out. From relevant literature review, expertjudgment and extensive Held survey, system database were generated with six input factors namely; number ofactivities (Act.), building area (BA), tvpe offoundation (FT), number of floors (storey), class of clients and contractors, and twooutput parameters (duration and cost). The results obtained indicated higher cost and duration variations for the projects given to sole and mini contractors. This is due to inadequate modernization, technical advancements and quality of resource personnel provided by the company to coordinate and manage the construction project activities. The medium and multi companies possess sophisticated tools and equipment which they utilize to achieve optimal results in terms of desired quality within planned time and resources to enable efficient execution ofthe project. This helps to prevent cost overrun and enable proper cost estimation. The bidding cost and negotiation fees were also observed to effect the choice of class of contractors recruited for the construction job as the clients with higher financial capacity such as government and cooperate organizations negotiated and hired the multi and medium companies. Feed-Forward Back propagation network was used in the smart intelligent modeling development in MATLAB using Lcvenbcrg-Marquardt training algorithm and mean squared error (MSE) performance criteria to achieve (6-22-2) optimized network architecture. Using loss function parameters (mean absolute error (MAE) and root mean squared error (RMSE) and multiple linear regression (MLR) statistical method, the developed ANN-model prediction performance was evaluated. The computed results indicate a good correlation between ANN-model and actual results with average R2 of 0.99995 better than MLR result of 0.6986. Also, MAE of 0.2952 and RMSE of 0.5638 were calculated which indicate a robust model.