Page 237 - Big Data Analytics for Intelligent Healthcare Management

230 CHAPTER 9 INTELLIGENCE-BASED HEALTH RECOMMENDATION SYSTEM

Recommender system

Content-based filtering Collaborative filtering Hybrid filtering
technique technique technique

Model-based Model-based
filtering technique filtering technique

Clustering techniques
Association techniques
Bayesian networks User-based Item-based
Neural networks

FIG. 9.2
Hierarchy of a recommendation system based on filtering.

filtering technique, user content determines recommendations. The user content deals with different
attributes of items along with the user’s previous buying history. Users give their preferences in terms
of ratings, which are positive, negative, or neutral in nature. In this technique, the system recommends
top-rated items to the user.
B. Collaborative-based filtering technique

Instead of considering features and attributes of items to determine their similarity, this approach uses
user-based ratings to find similarity between items. After collecting all user ratings, the system com-
pares these ratings with other users with the help of a utility matrix and recommends top-rated items to
the user. We use various distance measures such as Jaccard’s distance, cosine distance, and Pearson’s
coefficient, etc. to find a user’s degree of similarity. This filtering method is usually used in
e-commerce websites to recommend items based on users’ ratings.
C. Hybrid filtering technique

This technique combines the above two methods to strengthen the performance and accuracy of the
recommendation system. Hybrid filtering technique can be achieved by using any of the following