Bayesian networking involves identifying and analyzing a relationship between symptoms through a particular method of graphical modeling. In Bayesian networks, a probabilistic graph model is devised to represent both random variables and any dependencies these variables have. Bayesian networks can be used in a variety of different scenarios. One example of a possible use of this type of probabilistic model is determining the probability that an individual has a given set of diseases by inputing that patient’s symptoms.

Bayesian networks are in effect directed acyclic graphs, or DAGs, that include nodes representing the various quantities, variables, or parameters that serve as input information. While an edge in such a setup represents a conditional dependency, a node that is not connected shows a variable that is independent by other variables represented by nodes in the graph. The various nodes in the Bayesian network are identified with a particular probability function.

Algorithms can be created that allow for learning and inference through the use of a Bayesian network. Also, a dynamic Bayesian network can be produced that is capable of modeling variable sequence. Influence diagrams ware another possible product of the use of Bayesian networks. These are generalization of Bayesian networks that can demonstrate and find solutions to decision problems.

### The history of Bayesian networks

The term “Bayesian networking” itself was not used until 1985. This term was chosen to stress three aspects of the concept: the sometimes subjectivity of information that is put into the graph; the pertinence of Bayes’ conditioning to the concept; and the difference in reasoning using evidential or causal analysis.

### Inference tasks

There are three major inference tasks for which Bayesian networks are commonly used. These include:

- Unobserved variables- A Bayesian network will represent an all-inclusive model of both a group of variables and the relationship between those variables. As such, it can answer certain queries regarding the variables and relationships. Most often, Bayesian networks do this through the following three inference methods: variable elimination, clique tree propagation, and recursive conditioning/search.
- Parameter learning- Devising a complete Bayesian network requires that each node be specified regarding its probably distribution in relation to conditional variables. Often, discrete or Gaussian distributions are used to complete this type of inference tasks using Bayesian networks. If only a distribution’s constraints are available, a single distribution can still be determined through using the principle of maximum entropy.
- Structure learning- Machine learning is usually used so that a Bayesian network’s graph structure can be learned using automation. Alternatively, structure learning may be possible with Bayesian networks if an optimization based search is used. This involves the use of not only search strategy, but also the development of a scoring function.

### Applications of Bayesian networks

Bayesian networks can be used in many areas where modeling knowledge is necessary. However, they are especially important in applications relating to biology. For example, they are commonly applied in computational biology, medicine, bioinformatics, and biomonitoring. Non-biology areas where their use is commonly applied include sports betting, document classification, image processing, engineering, gaming, risk analysis, and law. A computer science degree is excellent preparation for a career working with Bayesian networks.