This is a very good question.

Here is an approach to setting up the desired model:

1. Let Char be a set of features or characteristics of the elderly. For example, let

Char = {Age, Health, Hobbies, Origin}. 

2. Let Mem be a set of members of the elderly living in WSN smart homes.

3. Sample rules:

if x,y in Mem are the same age, then x belongs to the same group G1 as y.

if x,y in Mem are the same age and health, then x belongs to the same group G2 as y.

if x,y in Mem are the same age with the same hobbies, then x belongs to the same group G3 as y.

if x,y in Mem are the same age with the same origin, then x belongs to the same group G4 as y.

if x,y in Mem are the same age and health with the same hobbies, then x belongs to the same group G5 as y.

And so on. 

A more refined model would take into account the proximities of the classified members, i.e.,sort the members of group G2 so that members x and  y with closeness in age (refine the age characteristic) and x and y with closeness in health (refine this characteristics to reflect dietary and exercise habits, for example), then add a meta-classification of x and y in terms of their proximity in age and health.

More to the point and for specific approaches to establishing frequency patterns, see, e.g.:

https://www.researchgate.net/publication/247830588_Normative_Data_for_Behavioural_Tests_of_Auditory_Processing_for_New_Zealand_School_Children_Aged_7_to_12_Years

Article Normative Data for Behavioural Tests of Auditory Processing ...

The simplest algorithm is AprioriAll, which is very similar to the Apriori algorithm for association rule mining. It might be a good way to start.

The cSPADE algorithm is more used in practice. It is implemented in a R package called "arulesSequence".

There are also methods based on Markov Chains, which are frequently used in studies in the field of bioinformatics.

hi?

A more refined model would take into account the proximities of the classified members, i.e.,sort the members of group G2 so that members x and  y with closeness in age (refine the age characteristic) and x and y with closeness in health (refine this characteristics to reflect dietary and exercise habits, for example), then add a meta-classification of x and y in terms of their proximity in age and health.

More to the point and for specific approaches to establishing frequency patterns, see, e.g.:

https://www.researchgate.net/publication/247830588_Normative_Data_for_Behavioural_Tests_of_Auditory_Processing_for_New_Zealand_School_Children_Aged_7_to_12_Years