Overview over classes of CyDASMining dll

The CyDASMining class library is realized with Visual Basic .Net as a .Net assembly.

The class library contains a set of classes useful for ...

Data Mining is intended to find patterns in large sets of data. For this purpose, the original data have to be transformed into an easily accessible flat structure. This structure is defined by the MiningInputData class. It mainly contains three items: a list of the case names, a list of the event names, and a boolean matrix indicating which case comes with which events.

The structure does not ask for the quantity of each event in a case, only if it was encountered or not. If quantity is important, it is the events that have to be redefined: one times event A is an event distinct from two times event A, and so on, such leading to a greater number of events which in turn become useful for the above boolean matrix.

The MiningInputData are then passed into a Miner object which performs the actual work: counting of events, counting of combination of events, the dependence of events from each other and their reliabilty, etc.

For visualisation purposes, a few controls were built:

  • The Dependence Viewer visualises the dependence between distinct events (positive or negative correlation, and its strength)
  • The Evolution Tree shows a hypothesis on the series of events.

The mining library is intended for general use. There is no specialisation for cytogenetic purposes. Hence, apart from aberrations in the chromosomes of cancer patients, also e.g. items in a shopping basket can be analysed.