Page 258 - Data Science Algorithms in a Week

240 Khaled Alshareef, Ahmad Rahal and Mohammed Basingab

1} to retrieve similar cases from the case-base by matching the elements of this target set
to those of the same level in the case-base. This comparison guides the search as it
traverses through the decision tree. The approach starts at the root node (l = 0) where the
first step in the retrieval process is to match g1 to an element in V1 (all children of the root
node) such as:

1∈ 1→ ℎ= ℎ
Else 1∉ 1→ ℎ= ℎ.

If a match does not exist, the retrieval process will terminate. If on the other hand the
new case finds a match in the base-case, the decision tree will then choose the edge that is
connected to the node (at l = 1) with the same category as the target case. The step to
follow, is to match all the remaining attributes of set 〈 〉= {g2, … gl-1}by comparing the
second attribute, g2 to a subset of 〈 2〉; where V2 is the set that contains all the possible
paths taken by patients, and 〈 2〉 contains all the paths under the matched category g1.
Due to the nature of this attribute, four different paths might be possible in the case-base.
The attribute match function yields three possible results as follows:

2 = 2 → ℎ = ℎ
2 ≠ 2 → { ℎ △ ℎ ±1 ℎ △ ℎ ±3 → ℎ =
ℎ ℎ △ ℎ ± 2 → ℎ = ℎ ℎ

Based on the value of the attribute match, the approach will choose the edge that is
connected to the node (at l = 2). This choice will yield the same path number when
perfect matching is achieved. However, if perfect matching is not achieved, then a partial
match or somewhat match will be chosen. The next step includes the matching of the
remaining attributes of set 〈 〉= {g3, … gl-1} to a subset of 〈 3〉; where V3 is the set
containing the possible number of doctors in the ED, and 〈 3〉 contains all number of
doctors matched under path g2. The remaining attributes are numerical, and will have
similar matching functions. For g3, the attribute matching function will use the absolute
difference between g3 and each of the elements in 〈 3〉
Based on the difference value zi, the approach will choose the node (at l = 3)
corresponding to the minimum difference value. The attribute match value indicates the
degree of similarity between the target case’s attribute g3 and each one of the elements in
the subset 〈 3〉. Similarly, the same matching process is also used to match the remaining
attributes of the target case such as g4, g5, and g6. Finally, the subset 〈 7〉 containing the
children of the node is matched with g6 to return the result of this retrieval engine. This
result will define the case(s) Ax from the case-base that are similar to the target case G.