Loading...

Buildings Construction Cost Prediction Using Hybrid Machine Learning Models

Daneshgar, Arshia | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: English
  3. Document No: 57006 (53)
  4. University: Sharif University of Technology, International Campus, Kish Island
  5. Department: Science and Engineering
  6. Advisor(s): Haj Kazem, Kashani; Ghahramani, Amir Ali
  7. Abstract:
  8. Since the awareness of project costs is a prerequisite for resource allocation and budget planning in construction projects, it is crucial to forecast project costs. In addition, as a part of the decision-making process, managers use predicted construction costs to assess and reduce project time risks. This study aims to estimate and predict residential building construction costs using hybrid machine learning models. In this research, to estimate the construction costs, in addition to considering the general characteristics of buildings, economic parameters are also considered. These physical and economic features are determined using experts' opinions and the results of previous studies. The physical parameters include the total floor area, lot area, project locality, preliminary estimated construction cost at the beginning of the project, and duration of construction. Moreover, the economic features contain the building services index (BSI), cumulative liquidity, private sector investment in new buildings, land price index, the nonofficial exchange rate with respect to dollars, notes and coins in circulation, rental housing index of Tehran, consumer price index (CPI), and building part of gross domestic product (GDP). These economic time series are collected seasonally from the time series database of the Central Bank of Iran. Contrary to the general characteristics of the building that can be determined for each project at any time, the economic parameters, when there is no data, must be estimated. Therefore, a Gated Recurrent Unit (GRU) model was developed for each economic parameter to estimate and forecast them. To accurately represent the changes in economic parameters during a construction project, a combination of each factor had to be chosen to reflect their fluctuations over the project's duration, like the average and standard deviation of each factor over the project construction period. After determining each project's physical and economic parameters, they were entered into cost estimation models. Linear regression, XGBoost, and MLP neural networks were implemented to predict building construction cost. The XGBoost method, with a 94.1% r-squared score, reached the highest accuracy among other models. After determining the output by evaluating the best model and analyzing it, the most critical features affecting the building construction cost were the consumer price index (CPI), preliminary estimated construction cost at the beginning of the project, notes and coins in circulation, building services index (BSI), and land price index. This research worked on building projects in Tehran, and, therefore, the implemented model is useable in this region
  9. Keywords:
  10. Machine Learning ; Construction Projects ; Project Management ; Manufacturing Cost ; Construction Industry Projects ; Building Feasibility Ptojects ; Building Construction Cost Estimation

 Digital Object List

 Bookmark

No TOC