Classification modes

In general, there are two different classification modes: supervised and unsupervised learning modes. It is important to understand their differences so that the potential techniques you want to use can be put into perspective. The supervised and unsupervised learning modes refer to the training procedure and how their requirements can affect the construction of the model.

Supervised Learning

Typically, supervised learning requires the training data to be fully labelled, for example in the field of fault detection and diagnosis, each data instance is assigned with either a normal or abnormal class. Any unseen data is compared again the trained model to determined which class it belongs to. Example supervised learning algorithms are:

• Naive Bayes
• Artifical Neural Network (ANN)
• Support Vector Machine (SVM)
• Adaptive Neuro-Fuzzy Inference System (ANFIS)

Unsupervised Learning

By contrast, unsupervised learning doesn't have any classes assigned to training data. The algorithm itself needs to determime what those classes are and how to separate them. The most well-known unsupervised learning algorithms are:

• k-Means Clustering
• Fuzzy c-Means (FCM)
• Self-organising Feature Mapping (SOM)

Semi-supervised Learning

In addition, semi-supervised learning is a class of supervised learning tasks. It uses training data that consist of labelled as well as unlabelled samples. The challenge in semi-supervised learning is that the labelled samples should be fairly accurate. Its applicability is dependent on whether the accurate labelling can be made regarding the training data.