Machine Learning-Based Prediction of Static Pile Load Capacity and Development of Equations for BNBC 2020 Refinements. (Working Paper)

By

The specific objectives of this study are: 

  1. To develop new equations for pile capacity prediction under varying soil and pile type conditions using Genetic Programming. 
  2. To apply, compare, and rank multiple ML algorithms (MLR, SVR, RF, ANN, XGBoost, GP) for static pile load capacity prediction using evaluation metrics R², RMSE, MAE. 
  3. To optimize the models using Bayesian Optimisation for hyperparameter tuning strategies and identify the globally best-performing model. 
  4. To identify the influence of input features over model predictions and check physical consistency. 
Scroll to Top