Utilizing Artificial Neural Networks to Compare Construction Cost Estimates

Utilizing Artificial Neural Networks to Compare Construction Cost Estimates
Student's Name
Institution/Affiliation
Course
Professor
Date
Abstract
Construction cost forecasting is crucial for construction enterprises to remain competitive and grow. For initial feasibility studies and a successful conclusion, reliable building cost estimates must be made early. A vast array of variables influences the cost estimate. This investigation compares regression analysis, support vector machines, and artificial neural networks (RA). Estimating building expenses and forecasting price increases are critical for builders, estimators, and project owners. The pricing technique is challenging since building prices have always been susceptible to fluctuations that tend to increase over time. The construction cost index has frequently been utilized in project cost estimates. The problem is that few organizations provide estimates for it. The most significant addition to this work is providing a reliable method for calculating project costs, especially considering the present inflation rate. Construct cost forecasts are created using an Artificial Neutral Network (ANN) approach. The choice of transfer function may have a substantial effect on the neural networks' performance. This Study aims to create a model for predicting building expenses and compare it to regression analysis and support vector machines while considering the cost of materials.
Table of Contents Abstract 2 1.0 Introduction. 4 1.1 Background. 4 1.2 Motivation of the Study. 5 1.3 Different Techniques for Project Cost Estimation. 6 1.4 Statement of the Problem.. 8 1.5 Aim of the paper 8 2.0 Predicting construction cost using ANN Model 9 3.0 Research Methodology. 10 3.1 Collection of Data for the Research. 10 3.2 Building the ANN Models. 11 3.3 Trial Models. 12 3.4 Analysis of Results. 12 4.0 Conclusion. 13 Appendices. 15 References. 17 1.0 Introduction
According to some, the accomplishment of a construction project is measured by how well it adheres to the owner's expectations for quality, timing, and budget. Because of this, a construction manager or contractor needs efficient tools for job scheduling and budget or cost estimation. Any construction project needs to have an early budget or cost prediction stage. An expected profit can quickly become lost due to inaccuracy in the budget or cost prediction (Matel et al., 2022). It is crucial to remember that estimating the costs of building projects is challenging because numerous unpredictable elements impact them. According to Matel et al. (2022), a variety of categories can have a significant impact on project costs. These variables include, among others, the cost of materials, transportation costs, site conditions, project scale, and schedule. The cost of the materials, which affects the overall cost of building, is one of those elements.
1.1 Background
Fan and Sharma (2021) state that the problem of predicting project costs is classified as a multifaceted issue. It has been evaluated utilizing approaches such as support vector machines, artificial neural networks, and regression analysis for the project. These prediction strategies utilize historical cost data to establish a functional relationship between cost changes and cost-affecting variables. The primary obstacles to accurate construction cost estimations are specific project information, alterations to the design specifications, and project development uncertainties. Analysis based on linear regression is unsuccessful. In the literature, regression analysis and statistical approaches are commonly used for cost estimation. Due to the vast number of important variables and their interconnections, all conventional methodologies have limits regarding accurate project cost forecasting (Fan & Sharma, 2021). Artificial intelligence techniques like the ones mentioned above can resolve the problem of cost prediction for building projects.
In the 1990s, neural network (NN) technology emerged as a practical substitute for calculating construction costs. The use of NNs as an alternative to finding a solid cost-predicting relationship that mathematically expresses the overall project cost as a function of the variables for the project that has the greatest impact on that building's cost is advantageous since it avoids the requirement to create such a relationship. An artificial neural network (ANN) technique is applied to predict construction costs. The nonlinear and complex factors' interplay between the project input and the desired result can be handled by ANN. Previous studies have demonstrated that the neural network model for cost estimate is superior to conventional regression techniques (Fan & Sharma, 2021). The ANN model demonstrated how neural networks might lessen the uncertainty surrounding construction project costs. The additional research aims to create an appropriate and practical model for precisely calculating building expenses. The model's architecture affects the accuracy of NN models.
According to Perez et al. (2019), the two key factors in NN architecture are as follows:
1 The number of nodes in the unseen layer Poor training is caused by fewer hidden neurons, while overfitting is caused by too many hidden neurons in the hidden layer.
2 Transfer Function: Input patterns are subjected to the transfer function at every output and hidden node. The transfer function choice may significantly impact the performance of neural networks.
1.2 Motivation for the Study
Budgeting, planning, and monitoring for compliance with the client's available money, schedule, and unfinished work are also crucial in school building construction projects (Imam & Zaheer, 2021). Additionally, the success of a building project is significantly influenced by how accurately construction expenses are estimated, which also impacts the owners' decision-making. But because the plans and documentation are sometimes incomplete, it can be challenging to swiftly and effectively estimate construction costs during the planning stage. Because of this, some methods have been developed to predict construction costs using the scant early-stage project data precisely. Neural networks (NN), support vector machines (SVM), case-based reasoning (CBR), and regression analysis (RA), among others, are common methods for cost estimation.
Examples of models created for forecasting or estimating building costs include the RA, NN, SVM, and CBR models. Imam and Zaheer (2021) note that statistical and linear regression-based approaches to cost estimate have been developed since the 1970s. Artificial intelligence techniques, including expert systems, NN, and CBR, have been used since the late 1980s. The cost-predicting model has also been researched during the 2000s. According to earlier studies, an NN model for cost estimation is preferable to a RA model.
Additionally, the SVM technique's accuracy for cost estimation is comparable to that of the RA technique for cost estimation. As a result, it is required to evaluate RA, NN, and SVM to choose the best method for estimating building costs. Construction costs were estimated using historical cost data in this Study to assess the accuracy of the three estimating methodologies. It allowed regression analysis to examine a cost estimation model that essentially combined neural networks with support vector machines.
1.3 Different Techniques for Project Cost Estimation
Musarat et al. (2021) claim that there are two primary cost-estimating techniques: traditional and non-traditional. The estimators frequently employ traditional approaches for estimating the costs of construction projects. The Quantity Rate Analysis is the most conventional technique for estimating construction costs. Various cost estimation models have been created and studied in recent years, in addition to standard cost approximating practices, to advance the accuracy of cost estimates in predicting the final prices. Many of the developed methods use artificial intelligence or statistics. The following are some non-traditional techniques: Forecasting for Reference Classes, Neuronal systems Support vector machine, Regression Analysis, Fuzzy Inference System, Monte Carlo simulation, and Case-Based Argumentation.
Figure 1. Mode of operation of Artificial Neural Network
The Study's conceptual structure relies on a neural network-based method for estimating the initial construction cost for apartment developments. To accomplish the Study's goals, a systematic approach was used. There are four main steps in it:
1 Choosing an appropriate neural architecture and identifying the input variables
2 Training and testing the neural network model concerning the project
3 Choosing an appropriate neural architecture and identifying the input variables
4 Gathering data for training and testing 
To predict building costs, the Artificial Neutrals Network (ANN) technique is utilized (Musarat et al., 2021). The transfer function choice may significantly impact the performance of neural networks. The Study objects to develop a model for predicting building costs and compare it to regression analysis and support vector machines while considering the cost of materials.
1.4 Statement of the Problem
To achieve the goal of a quality cost estimate, the cost estimation process must be applied effectively across the whole life cycle of the building project. Project cost estimating has numerous challenges throughout the process, although crucial and sometimes disregarded (Leśniak et al., 2021). Common obstacles include Cost Overruns: A cost overrun is the discrepancy between a project's completed final...