Table 4 Database for prediction of the ultimate bearing capacity.
Algorithm | Type | Strengths | Weaknesses | Best use cases | Performance | Computational cost |
|---|---|---|---|---|---|---|
Probabilistic | Probabilistic predictions with uncertainty estimates. Handles non-linear relationships. Strong theoretical foundation | Computationally expensive. Sensitive to kernel choice. Poor scalability for large datasets | Small datasets with uncertainty quantification. Scientific modeling, Bayesian optimization | High accuracy for small datasets, but slow for large datasets | High | |
Ensemble (Boosting) | High accuracy and scalability. Handles missing values. Regularization to prevent overfitting | Requires careful hyperparameter tuning. Computationally expensive for very large datasets | Large datasets, complex prediction tasks | State-of-the-art performance for structured data | High | |
Ensemble (Boosting) | Extremely fast and memory efficient. Handles large datasets. Supports categorical features | May overfit on small datasets. Requires tuning for optimal performance | Large datasets, high-dimensional data, ranking tasks | Faster than XGBoost, with comparable accuracy | Moderate | |
Ensemble (Boosting) | High accuracy. Handles non-linear relationships. Robust to outliers | Computationally expensive. Prone to overfitting without tuning | Diverse datasets, complex prediction tasks | High accuracy but slower than XGBoost and LGBM | High | |
Ensemble (Bagging) | Robust to overfitting. Handles high-dimensional data. Provides feature importance | Less interpretable than single trees. Computationally expensive for large datasets | Diverse datasets, feature selection | High accuracy, but slower than boosting methods | Moderate to High | |
Ensemble (Boosting) | Handles categorical data natively. Robust to overfitting. GPU support for fast training | Requires tuning. Less flexible for custom loss functions | Tabular data with categorical features, click-through rate prediction | High accuracy, especially for categorical data | High | |
Ensemble (Boosting) | Improves weak learners. Simple to implement. Robust to overfitting | Sensitive to noisy data. Computationally expensive for large datasets | Binary classification, face detection | Good accuracy for small datasets but struggles with noise | Moderate | |
KNN65 | Instance-based | Simple and intuitive. No training phase. Handles non-linear data | Computationally expensive for large datasets. Sensitive to feature scaling. Struggles with high-dimensional data | Image recognition, recommendation systems | Good for small datasets, but slow for large datasets | Moderate to High |
Ensemble (Bagging) | Reduces variance and overfitting. Combines multiple models for robustness. variance, improves generalization, robust to noise | Computationally expensive. Limited bias reduction | Regression tasks, noise reduction | Improves stability and accuracy of base models | Moderate to High | |
Non-parametric | Easy to interpret and visualize. Handles mixed data types. No need for feature scaling | Prone to overfitting. High variance. Struggles with extrapolation | Classification and regression, interpretable models | Good for small datasets, but prone to overfitting | Low to Moderate | |
Kernel-based | Handles non-linear relationships using kernels. Robust to outliers. Provides probabilistic predictions | Computationally expensive. Requires careful hyperparameter tuning. Poor scalability for large datasets | Regression tasks with non-linear relationships, outlier-resistant modeling | High accuracy for small datasets, but slow for large datasets | High |
