THE UTILIZATION OF ROBUST INTELLIGENT MODEL FOR PROJECT DURATION PREDICTION

Department of Civil Engineering, University of Anbar, Al Anbar, Iraq

E-Mail: eng.saadish@uoanbar.edu.iq

INTRODUCTION

Construction project time completion is highly important for contractors as well as the clients. Incomplete projects with its stipulated duration can affects ?these parties and cause a lot of loses due to the increment in the project cost and other consequences on the clients’ objectives. Hence, in order to keep any project activities within prestigious management, budget and quality, an accurate and reliable prediction for the project duration is highly essential for successful completion. In accordance to effective accomplishment criterion, project duration prediction is vital for both contractors and customers prospective. Here, the customer can initiate finance, materials, and cash procedure plan in preset of time and establish an optimal for their targeted project. On the other hand, contractors can predict the progress of the construction accurately with performing works actions appropriately which lead to get control in the business development and give good decision precautions against postponements.

METHODOLOGY

The ELM learning algorithm tool for the SLFN framework which randomly chooses the input weights for the analytical determination of the output weight of the SLFN. The ELM has a faster learning speed and favorable characteristics, requiring less running time (by minimizing the manual interventions through analytically determining the network parameters) when compared to the traditional algorithms. Some of the benefits of using this algorithm are the ease of using it, the fast learning speed, the compatibility with several non-linear activation and kernel functions, as well as its superior performance when compared to the other algorithms. Over the past five years, the ELM has been successfully implemented in various applications in the literature and approved its proficiency such as learning capacity for clustering, regression, feature learning, and classification application. Up to the author knowledge, there is no single research has been undertaken the skill of ELM approach for modeling project duration, here where the novelty lies. In more detailed graphical presentation, Figure 1 shows the structure of the generalized ELM model. In accordance to the figure 1 conception, the input variables were the (i.e., total area of construction, total volume value, cost value, and facade area value) while the output variable was the construction project duration.

magnitude values but instead evaluates all deviations of predicted from the actual values in an equal manner irrespective to the value indication (negative or positive) On the other hand, NS is crucial as it objects to assess theerror in a normalized manner . Fundamentally, this metric measures the ability to forecast data that deviate from the mean. As a ratio, this assesses the agreement between actual and predicted data, and is sensitive to the differences in actual and predicted means and variances. Finally, an employment for the WI was established. This metric can provide more advanced information than NS, since this metric does not square the differences between actual and predicted data. Specifically, WI (namely the ratio of the mean square error and potential error multiplied by the number of observations and then subtracted from one) aims to assess the differences based on squared differences.

CONCLUSIONS

In the current research, an intelligence model namely extreme learning machine is proposed to predict construction project duration. The significant of conducting this research owing to the necessity of establishing a reliable and robust predictive for the applied application and can be implemented practically. Five construction project features data set were collected from the department of construction technical works, Ankara, Turkey. To achieve this feat, several construction variables used to determine the project duration “the target variable”. For the purpose of validation, classical intelligent model called artificial neural network was developed. Several performance metrics calculated including absolute error and fitness measurements. The implementation of the ELM model for the applied application was successfully established. The level of similarity between the predicted and actual values was enhanced significantly with the ELM model when compared to the ANN model. The utilization of the proposed non-tuned model can be used by the contractors to estimate the duration of construction and compare it with the duration earlier supplied by the client to check for the possibility of realizing the with the given budget ad at the specified time. Such modeling requires their own databases. This historical data based modeling approach of the contractor will be more concrete, practical, and reliable when compared to the subjective methods based on intuitive estimations that are currently been used by planners.