This paper applied an a priori algorithm to the self-care service
log data to reveal actionable association rules. The condition support
>5% and confidence >90% revealed 63 association rules. For
brevity, Table 1 lists only three interesting association rules revealed
by the a priori algorithm. The rules are sorted by ‘‘rule support’’.
Consider the first rule in the list. The antecedent is Exercise
management function, and the consequent is Diet management
function. The form of the rule is therefore Exercise management
(1376))Diet management (1265)conf:(0.92). The support of the
rule is 37.46%, meaning that the rule applies to almost 37% of the
3391 total sessions in the data, which is a very high support level.
A 92% confidence indicated that, of these 3391sessions, 1376 met
the antecedent condition; that is, 1376 requested Exercise management
at some point. Of these 1376 sessions, the Diet management
function was also requested in 92% (1265 sessions).
According to rule 1, elders who pay attention to the Exercise management
also emphasize Diet management in self-care behavior.
According to rule 2, elders who pay attention to the BMI management
also emphasize Blood management in self-care behavior.
Rule 3 is that elders who pay attention to the BMI management
& Exercise management also emphasize Diet management & Blood
management. Surprisingly, in the support >5% and confidence >90%
condition, the analysis showed that, of the functions in the four different
topic, only the Query weather forecast function in the lifeinformation-related topic was associated with Diet management
and Exercise management in the health management-related
topic.