| SV Learning algorithms |
A model is trained on a range of input data that are associated with a known outcome (but there is no knowledge on predictors). |
| USV Learning algorithms |
Does not involve the knowledge of the outcome; they are usually used to find undefined patterns or clusters in datasets or to reduce the number of features. |
| Reinforcement Learning |
The algorithms do not need to know the outcome; they use the estimated errors as rewards or penalties. |
| Association Rule Mining (ARM) |
Techniques aim to observe frequently occurring patterns, correlations, or associations in the data; how items are associated to each other. |
| Classification techniques |
The objects are assigned to one of a pre-specified set of classes. Some classification techniques are:
|
| Clustering techniques |
The objects are grouped without any pre-specified knowledge on the rule of their grouping (based on using the distance metrics). Some clustering techniques are:
K-means K-Nearest Neighbor (KNN) Principal-component (PC) based clustering Self-organizing maps (SOMs) Latent Class Analysis (LCA)
|
| Deep Learning |
More recent concept of ML; has much better ability of feature representation in the abstract level; has an ability to translate the information from the high level of an abstraction to the level that is more understandable for human reasoning; uses complex algorithms, such as ANN. |
| Advanced computer-based methods |
Techniques that can be used to organize highly complex or unstructured data or to find temporal trends in data:
Graph-based DM Data Visualization and Visual Analytics Topological DM Fuzzy set theory-based algorithms Natural Language Processing (NLP) methods Dynamic BNs Temporal Association Rules (TARs) Non-negative Matrix factorization (NMF) and tensor factorization (NTF) approaches and their developments in DL
|
