2.2.2 Similarity Measure :
Measure the basis of: the direction of two vectors as a basis for similar consideration whether the vector length is not important:
1. The angle similarity coefficient (cosine)
Note: Scaling rotation and scale invariant coordinate system is but in general linear transformation and coordinate translation is not invariant.
2. The correlation coefficient
it is actually a vector of the cosine of the angle of the data center.
Note: correlation coefficients is (-1, 1) - 1 is substantially related, related to almost 1.
2.2.3 match measure
as measure only two states (0,1), matching measure used. 0 indicates no such feature, this feature 1 expressed. So called binary feature.
For binary n-dimensional feature vector similarity measure can be defined as follows:
(. 1) Measure the Tanimoto
Example:
(2) Measure Rao:
Note: the number of feature number (1-1) and the selected matching feature ratio.
For example:
(3) the simple matching coefficient
Note: The molecule of formula (1-1) wherein the number of matching the number of (0-0) and matching features, the denominator is the number of features considered.
Example:
(. 4) DIEC coefficient
(5) Kulzinsky coefficient
2.3 class definitions:
For a set S to be classified, the classification requirements of various types S1, S2, S3, SN satisfied:
Definition 1:
Definition 2:
Definition 3: