Machine Learning
Asymptotic behaviors of support vector machines with Gaussian kernel
Neural Computation
A study of the behavior of several methods for balancing machine learning training data
ACM SIGKDD Explorations Newsletter - Special issue on learning from imbalanced datasets
SMOTE: synthetic minority over-sampling technique
Journal of Artificial Intelligence Research
| Hi-index | 0.00 |
Rheumatoid arthritis (RA) is a chronic inflammatory joint disease that leads to irreversible joint destruction. To prevent this, new biological therapies, such as infliximab, have been successfully developed. The present analysis is based on an expanded access program in which 511 RA patients with chronic refractory disease were treated with infliximab. They received a standard dose of 3mg/kg on weeks 0, 2, 6, 14 and every 8 weeks thereafter. On week 22, the treating rheumatologist evaluated the situation of every patient and decided whether the current dose should be increased or not. This decision can be considered as a measure of insufficient response. In the present analysis, 3 machine-learning classification techniques-the self-organizing map (SOM), multilayered perceptron (MLP) and support vector machine (SVM)-are implemented to model the decision to give a dose increase. Their performance on increasingly multivariate real-life data will be studied and compared to classical statistics-linear discriminant analysis (LDA) and logistic regression (LR). Results show that the SOM is an excellent tool for data visualization but not for classification. All the remaining methods show good classification performance, if configured well. However, as the number of features increases, the performance decreases. The SVM suffers to a lesser degree from this curse of dimensionality. Expectation maximization (EM) comes out as a good method to cope with missing values in such real-life data.