Prediction of compressive strength experiment results of concrete used in buildings using machine learning techniques

Measuring concrete compressive strength takes place in the construction industry as a very important process for giving information about the durability of the structure. Performing concrete compressive strength tests requires a series of processes and appropriate steps. The most important element in the test stages is the addition of the necessary components in the concrete samples taken at the required rate. In addition, in order to obtain the best concrete compressive strength, the tests must be carried out within a certain period of time. In addition, a serious cost and experts in the field are needed to carry out these experiments. However, inaccuracies in obtaining and recording pressure test results may occur due to the workload intensity. Since the early 1990s, artificial intelligence algorithms have been actively used in automatic classification and regression problems. As time progresses, artificial intelligence algorithms with data types and sizes have been developed and these algorithms have achieved great success as decision supporters in object recognition, natural language processing, audio, image and video classification tasks. Artificial intelligence algorithms, which are used in many fields such as security, energy, metallurgy and medicine, have also been frequently used in the field of civil engineering.. In this thesis study, compressive strength estimations were made with modern artificial intelligence algorithms (machine learning). An extensive experimental dataset of 1030 times was used to implement these algorithms. This test includes concrete compressive strength results according to the mixing ratios of 8 different concrete components. Before applying the machine learning algorithms, exploratory data analysis was made and the effect of the components on the concrete compressive strength was examined. Then, concrete compressive strength estimates were made with 9 different machine learning algorithms. A best and reliable estimation result was obtained with the Gradient Boosting algorithm.