TD_OneClassSVM is a linear support vector machine (SVM) that performs classification analysis on data sets to identify outliers or novelty in the data.
- Classification (loss: hinge). During the training, all the data is assumed to belong to a single class (value 1), therefore ResponseColumn is not needed by the model. For TD_OneClassSVMPredict, output values are 0 or 1. A value of 0 corresponds to an outlier, and 1 to a normal observation/instance.
TD_OneClassSVM is implemented using Minibatch Stochastic Gradient Descent (SGD) algorithm, which is highly scalable for large datasets. See TD_GLM for details on SGD.
The function output is a trained one-class SVM model, which can be input to TD_OneClassSVMPredict for prediction. The model also contains model statistics of MSE, Loglikelihood, AIC, and BIC.
One-class SVM is a machine learning algorithm used for anomaly detection, which means identifying data points that deviate significantly from the norm. It is a variant of Support Vector Machines (SVM) that is designed to work with only one class of data, hence its name.
One-class SVM works by first mapping the input data to a high-dimensional feature space and then finding the hyperplane that separates the data from the origin with the largest margin. The hyperplane is then used to classify new data points as either belonging to the same class as the training data or not.
One of the main advantages of One-class SVM is that it can be used with unlabeled data, meaning it can detect anomalies without requiring labeled data for training. This makes it particularly useful for detecting fraud, intrusion detection, and other types of security applications where anomalous behavior is often the most important signal.
However, One-class SVM has some limitations, such as being sensitive to the choice of kernel function, and not being suitable for datasets with high levels of noise or a significant overlap between classes. One-class SVM remains a powerful tool for anomaly detection and is widely used in industry and research.