Page 255 - Data Science Algorithms in a Week

The Utilization of Case-Based Reasoning 237

Figure 3. Different paths in the EDs.

The Nearest Neighbor Approach

Algorithms such as K nearest neighbor or R nearest neighbor are deployed to
determine the similarities between the attributes of both the new case we are seeking a
solution for, and the cases stored in the case-base. Similarities are then normalized to fall
between zero “0” and one “1” or as percentages. These functions use various similarity
metrics such as Euclidean distance, city block distance, probabilistic similarity measures,
and geometric similarity metrics. Similarity percentages are retrieved using a predefined
parameter value “K”. However, in the R nearest neighbors, cases with similarities
percentages (see equation below) that are greater than or equal to a predefined value “R”
are retrieved.

where,
NC represents the new case
SCs are stored cases in the case-base.
n is the number of attributes in each case
w is weight, and
f is the similarity function.