Читать книгу Data Analytics in Bioinformatics - Группа авторов - Страница 50

The DBSCAN algorithm proposed by Ref. [27] is based on the following criterion:

Eps(€)—this depicts the space around the data point, if the difference of distance between 2 data points is greater than or equal to € then they are declared as neighbors, if the (€) value is very small then they are termed as outliers, if the (€) value is high then that data point is merged to form cluster.

Minpts—this represents the minimum number of data points within the radius (€) for the larger datasets like microarrays Minpts are obtained from the dimensions of the dataset.

From the following definition of data points clusters and outliers are identified

Core point—presence of higher Minpts within radius of (€)

Border point—presence of less Minpts within radius of (€) but close to core point

Outlier—point which is neither core nor border point.