Akurasi Penggunaan Metode Support Vector Machine (Svm) Dalam Prediksi Kapasitas Dukung Fondasi Tiang

Abstract

Bearing capacity represents the strength of soil subjected to loading. Many methods can be used to estimate the bearing capacity of a foundation, such as artificial intelligence (AI). Support vector machine (SVM), which belongs to AI, is a popular method to predict the bearing capacity of a foundation. The advantage of using the SVM method is that it reduces assumptions in model building. This study is aimed at investigating the accuracy of SVM to predict the bearing capacity of a pile foundation using cone penetration test (CPT) data. The features (independent variables) used are pile diameter (D), pile length (L), pile material, pile type, installation method, conus resistance (qc), and frictional resistance (fs). For validation and calibration purposes, loading test data is used. The 10-fold cross-validation method is employed in the validation process. The result shows that the SVM model reaches its highest accuracy when the kernels C and g are set to a polynomial kernel of 0.5 and 0.1, respectively. The ratio of 90:10 is the best ratio of training to testing data, which has an R2 value of 0.9287 and an RMSE value of 637.109 kN in the testing process.