Estimation of approximate cost uncertainties in flood protection projects using artificial intelligence techniques

In this thesis; It is aimed to create the necessary mathematical models with artificial intelligence techniques that will allow the pre-tender approximate costs of flood protection projects tendered by the public sector to be estimated quickly and accurately. In accordance with the Public Procurement Law No. 4734, 88 flood protection projects tendered between 2012 and 2018 in Turkey were examined and all factors affecting the approximate cost were determined. The relations of the approximate costs of the projects with each other were evaluated and converted into a normal distribution. The independent variable inputs that make up the project input approximate cost (dependent variable) are shown with 9 different mathematical estimation models. These models are linear,power, exponential, quadratic and semi-quadratic estimation models developed by using the optimization technique known as the response surface method in the literature. Mathematical equation models are optimized using the training data set. At the optimization stage, the Flower Pollination Algorithm was used. In the optimization phase, 60% of the data set was used for educational purposes, while the remaining 40% was used for testing purposes. In the optimization studies, it was seen that the estimation model with the highest success rate was the P8 model. Among the mathematical optimization models of the P8 model, the highest success rate was achieved in the semi-quadratic model with approximately 94,72%. The results of this study, made through the public or private sector, especially where insufficient data exists prior to the procurement of infrastructure projects such as flood protection, and if it is necessary to identify the approximate cost in a short time; it was found that in high accuracy by using artificial intelligence techniques could be detected. Keywords: Flood protection project, Approximate cost, Response surface method, Flower pollination algorithm, Artificial intelligence techniques