Classification of Reinforcement Costs of Masonry Walls Using Hybrid Extreme Gradient Boosting and Softmax
It has been built for centuries as housing and animal shelters, especially in rural areas, due to the advantages of masonry buildings being economical, being built with local materials, and not requiring skilled labor. The walls, which are the bearing elements of masonry structures, are formed by placing stones, bricks, or blocks on top of each other with a binding mortar. In this study, a model with the XGBoost algorithm, which is a tree-based classification algorithm, is proposed to scale cost of the samples reinforced with welded wire reinforcement/polypropylene fiber added dry mix shotcrete. The model executes cost classification based on concrete, steel mesh, steel, epoxy, fiber and workmanship independent parameters. A softmax function was incorporated into the model for classification. A complexity matrix was produced to evaluate classification performance of model. Also, it was compared to other machine learning algorithms. The model yielded higher accuracy and lower false-positive rates. As a result, the proposed model can make better estimates in cost classification compared to other machine learning methods. In conclusion, using the classification ability of the model, it is aimed to measure the cost effect in the construction process that calls for high labor force, time and cost.