Predictive Modeling for Cost and Duration Estimation in Residential Construction Projects Using Machine Learning Algorithms

This study explores the application of machine learning algorithms in predicting cost and duration ofstandard residential projects. The researchers considered eight (8) parameters, and these are gathered fromdifferent construction companies in Pampanga. A total of 60 projects were gathered. Using the mostoptimal data splitting method, 42 of these projects were used for training the predictive model, and 18were used for testing. Python and visual studio code were used in creating the predictive model. Pearson ofcorrelation coefficient (r) was used to measure the correlation of the variables. Six (6) resulted to havestrong positive correlation. The other two (2) were categorized as moderate positive correlation. The Rsquared values of the regression models were tested to acquire the most optimal model. The Ridgeregression model resulted in the highest r squared value of 0.81. To test the accuracy of the forecastedresults, mean absolute percentage error (MAPE) was used. A MAPE value of 7.20 was obtained from thecost model, and 6.21 for duration model. These MAPE values indicate that the predictive model isconsidered “highly accurate forecasting”.