Enhancing pile-bearing capacity estimation through hybrid artificial intelligence models using K-nearest neighbors approach augmented with northern goshawk and beluga whale optimization techniques

A crucial challenge in pile analysis and design is the use of Artificial Intelligence (AI) and Machine Learning approaches to estimate the final load-bearing capacity of piles based on field testing data. The rmain objective is to develop sophisticated AI prediction models that are especially made for estimating pile-bearing capacity. The estimative model utilized in this study is founded on the K-nearest neighbors method. The study has implemented an innovative hybrid approach that combines Northern Goshawk Optimization and Beluga whale optimization techniques to achieve precise and optimal predictive outcomes. A comprehensive dataset has been carefully curated, encompassing various attributes related to piles and soil properties from various literature references. This compilation includes data collected from Cone Penetration Tests and the results of pile load tests. These datasets are utilized for the developed models' train, validation, and test stages. The efficacy of the suggested models has been validated by the accurate findings obtained from the methods used in this study. As a result of this integration, three hybrid models have been created: KNNG, KNBW, and KNN. Of particular note is the performance of the KNNG model, which has exhibited exceptional accuracy, with a minimal RMSE value of 118.29 and an impressive R2 value of 0.996.