Probabilistic Cost Estimation of Highway Projects Considering Terrain Type

نوع: Type: thesis

مقطع: Segment: masters

عنوان: Title: Probabilistic Cost Estimation of Highway Projects Considering Terrain Type

ارائه دهنده: Provider: Mohammad Reza Shiri

اساتید راهنما: Supervisors: Dr.Mohsen Babaei

اساتید مشاور: Advisory Professors:

اساتید ممتحن یا داور: Examining professors or referees: Dr. Javad Taherinejad - Dr. Morteza Heydari Mozafar

زمان و تاریخ ارائه: Time and date of presentation: 20244

مکان ارائه: Place of presentation: 43

چکیده: Abstract: Abstract: Accurate cost estimation in the early stages of construction plays an important and vital role in the success of any project. Initial wrong estimates or the absence of an accurate cost estimate can cause the loss of the project or even the closure of the project. One of the main applications of road construction cost forecasting is in the problem of road network development, where a number of projects are selected according to budget constraints and the efficiency of projects in providing optimal routes (in terms of time or distance) for construction in a long-term horizon. . The aim of this study is to provide a model for predicting the cost of construction of road construction projects in the early stages by considering the topographical conditions. In designing and planning, it is very important to choose the right route considering the slope of the land. By determining optimal routes that limit the slope of the land to a minimum, road construction costs can be reduced. Therefore, in this research, by reviewing and studying the information of road construction projects used in the design of the country's road network, road characteristics and land toll conditions were extracted using GIS software. First, an initial cost estimate was made using the software output information, road characteristics and land tolls. Then, using the Python programming language and the libraries in it, models for cost estimation with a comparative approach of models, artificial intelligence and cumulative learning models based on Boosting including XGBoost, CatBoost, LightGBM, deep neural network (DNN) and linear regression multiple (MLR) was used. In this research, the information of 539 road construction projects was examined and the examination of the architectural types of the models, by examining the output of the results and the accuracy of the introduced models, showed that the CatBoost model compared to other models with MAE = 0.013, RMSE = 0.048 and coefficient Determination of 0.997 has better accuracy and performance in cost estimation. Also, one of the big challenges in complex machine learning models is their transparency and interpretability, SHAP interpretation allows to analyze complex models in an understandable way. Also, in this way, the impact of the features used in cost estimation can be checked and show how each feature helps to increase or decrease costs.

فایل: ّFile: Download فایل