Aristotle University Of Thessaloniki (AUTH)
Ανάπτυξη υβριδικού πλαισίου πρόβλεψης κόστους κτιριακών έργων ανακαίνισης με συνδυασμό θεωρίας χρησιμότητας, πολυκριτηριακής ανάλυσης και αλγορίθμων μηχανικής μάθησης
Abstract
dc:descriptionConsidering the inherent uncertainty, incomplete initial data and unpredictable factors that characterise renovations in existing structures, accurate total cost estimation in engineering projects—and renovations in particular—represents one of the construction industry's most significant challenges. Conventional methods of estimation sometimes produce significant cost variances since they are unable to accurately model these non-linear features. The present doctoral dissertation synergistically incorporates Utility Theory, Multi-Criteria Decision Analysis (MCDA), and Hybrid Artificial Neural Networks with Multiple Linear Regression (MLR) to propose an innovative, integrated methodology for total cost estimation in technical projects, specifically applied to renovation projects. The develop-ment and analysis of a significant database of completed renovation projects provided the foundation for the research. Following the organisation and statistical evaluation of historical data, the proposed approach developed in three distinct yet integrated phases. Initially, the degree of uncertainty for each completed project is quantified per group work utilising utility theory. After a multi-parametric process which consisted of (a) a thorough literature review, (b) expert judgement through structured questionnaires, (c) statistical data processing, and (d) data significance evaluation, the Utility Indices UIj, which indicate the uncertainty weight of each project, were ultimately developed. The best possible historical projects are then assessed and chosen utilising the multi-criteria analysis method referred to as TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution). Through employing this filtering procedure, 40 projects are selected as the ideal homogenous training sample, demonstrating the highest degree of similarity to the project's attributes. Ultimately, a Hybrid Combinatorial Model that combines a Multiple Linear Regres-sion (MLR) model with a Radial Basis Function Artificial Neural Network (ANN-RBF) is provided with the processed data. Both linear patterns and intricate non-linear correlations between cost elements can be captured concurrently with this particular structure. The model was assessed utilising three distinct case studies and trained on actual data. The findings demonstrated the methodology's outstanding accuracy in forecasting, as the mean prediction error for typical projects was significantly low with a value of 1.70%, while even in projects with extensive technical particularities, it remained within acceptable ranges (~20%). The doctoral dissertation intends to provide a reliable decision support system for the building sector and the scientific community. Additionally, the architecture of the suggested model ensures its dynamic adaptability to market conditions by permitting continuous updates utilising additional information. The suggested methodology aims to reduce risk to the company while strengthening construction organisations' ability to withstand the unpredictability associated with intricate renovation projects.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Papadimitriou, Vaso
- Παπαδημητρίου, Βάσω
Subjects
dc:subject × 23- Πρόβλεψη κόστους
- Εκτίμηση κόστους κτιρίων
- Εκτίμηση κόστους ανακαίνισης κτιρίων
- Θεωρία χρησιμότητας
- Μέθοδος topsis
- Σύστημα λήψης αποφάσεων
- Τεχνητά νευρωνικά δίκτυα συναρτήσεων ακτινικής βάσης
- Υβριδικά μοντέλα
- Υβριδικά συνδυαστικά μοντέλα τεχνητών νευρωνικών δικτύων - πολλαπλής γραμμικής παλινδρόμησης
- Cost estimation
- Buildings cost estimation
- Renovation buildings cost estimation
- Utility theory
- Topsis method
- Decision support systems
- Αnn rbf RBF neural networks
- Hybrid ann models
- Ensemble models ann rbf-mlr
- Επιστήμη Πολιτικού Μηχανικού
- Επιστήμες Μηχανικού και Τεχνολογία
- Πολιτική μηχανική
- Civil Engineering
- Engineering and Technology
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/62410