PRINCIPAL COMPONENT ANALYSIS FOR PUBLIC BUILDING CONSTRUCTION PROJECTS COST ESTIMATION in ADDIS ABABA

Abstract

Cost estimate, in general, offers stakeholders in all types of construction projects with a valuable tool for forecasting the construction costs required to execute the building projects. Without an initial cost baseline provided by a cost analyst, it would be difficult to assign the budget that substantially determines the performance of construction projects. The independent variables for cost estimation are specific to each construction project thus posing challenges for researchers to apply multiple regression model for analyzing the cost variables. Multiple regressions are frequently used to estimate the cost of a particular project, however the variables considered independent sometimes demonstrate a high degree of connection, therefore they are likely to be excluded from such a model, resulting in inaccurate forecasts. This initiated a search for alternative tool for selecting cost influencing factors to develop a better predictive cost model. As a result, this research tries to solve these issues that typically arise in predictive cost estimating models, and it has proposed a hybrid technique based on the Akaike information criterion (AIC) and principal component regression (PCR). The methodology pursued to this goal is undertaking extensive literature review to identify most important factors that are related to public building project cost estimation. These data established the basis for the analysis and consequently four prediction models were developed with step wise multiple linear regression using MATLAB program. According to the findings of the study, principal component analysis efficiently solved the problem of multicollinearity with variance inflation factor less than two, whilst stepwise cross validation solved the overfitting problem at the lowest AIC at 1% level of significance. Seven stepwise regression models were developed to assess the predictive capability of linear regression models. Six models were fitted to the linear model, and four of them were determined to be valid based on their prediction capabilities. The cost prediction model sorted out five factors: design completion by the public body when bids are invited, Completion of project scope definition when bids are invited, Level of construction complexity, importance of project completion within budget and subcontractor experience and capability have all been identified as the main cost determining factors for cost Estimation of public building construction projects in Addis Ababa at 5 % level of significance. In the research region, four cost estimating comparisons for projecting public building costs for distinct project categories were found. The final model revealed 14 factors that explained 55% of the total variation, with R2 values of 59%, 61%, 62%, and 62%, respectively. The study also concluded that AIC based model selection is the best fit model as compared to SSE. The study also recommended to use detailed evaluation of location, procurement method, project cost to arrive at other promising alternatives in the future researches.