Comparison of linear regression and neural network models to estimate the actual duration of Greek highway projects
Résumé
The aim of this study is to highlight and evaluate parameters that correlate with the actual project duration and compare performance of linear regression with neural network models to predict a reliable final duration for highway construction projects based on data available at the bidding stage. The thirty-seven highway projects that are examined, constructed in Greece, are similar in available data like the extent, the type of work packages and the significance. Considering each project's characteristics and the actual construction duration, correlation analysis is implemented, with the aid of SPSS 25. Correlation analysis identified the most significant project variables. These include archeological findings, type of terrain, land expropriation, the existence of bridge, tunnel and embankment. Furthermore, the WEKA application, through its attribute selection function, highlighted the most efficient subset of variables, both within the initial available attributes but also within the set of attributes identified through correlation analysis. These selected variables through correlation analysis and / or WEKA and appropriate combinations of these are used as input data for linear regression models and neural network models. The corresponding models (linear regression and neural networks) are presented along with their performance. Finally, this paper presents the comparison of linear regression and neural network models to estimate the actual duration of Greek highway projects. Results' discussion and conclusions along with limitations and further research are appropriately analyzed.
Origine | Fichiers produits par l'(les) auteur(s) |
---|