Explaining Hybrid Filtering in recommendation system. Types of Hybrid recommendation system

The major problem of many on-line websites is the presentation of many choices to the client at a time; this usually results to a strenuous and time-consuming task in finding the right product or information on the site.

With heuristic analysis and evaluation, researchers have illustrated that a hybrid recommendation algorithm can effectively increase the accuracy and stability of recommendations to achieve better recommendation results. [Written in Research PaperResearch of Hybrid Collaborative Filtering Algorithm Based on News Recommendation]

Combinatorial types in hybrid recommendation system

Weighted: The weighted approach incorporates the results of various recommendation systems by combining them. The weightage of the contributing systems may be fine-tuned for more accurate predictions.

Switching: This category chooses content–based RS or collaborative filtering RS or any other RS such as utility based, demographic–based etc. depending on the requirement of the user.

Mixed: In mixed approach, the results of different recommendation systems are combined to produce the recommendation.

Feature combination: Preferential unique features of the recommendation systems are considered for evaluation and combined.

Cascade: It is similar to waterfall model, where the result of one system advances to the other, which further fine–tunes the result. The final results given by the concluding system would be highly refined.

Feature Augmentation: It uses specified results or definite characteristics of other systems, applies its distinctive methodology and gives the final recommendation.

Meta–level: A kind of divergent approach that gets the model generated by the previous RS rather than the results, it also develops a robust model along with providing recommendations.

Knowledge–based recommendation system: In this system, knowledge about the user’s requirement is stored and analysed for further recommendation. Knowledge–based and collaborative filtering approaches are similar to an extent. However, the collaborative filtering system uses ratings, whereas the knowledge–based system stores the user’s information and the products’ information as metadata, which are suitable to handle complex user problems in difficult domains.

Constraint–based recommendation system: A conditional statement that specifies user requirements added to knowledge–based system is called constraint–based recommendation system. The research work in uses taxonomy, based on which it frames conditional rule satisfying test cases. 

Ontology–based recommendation system: Ontology represents the formal, explicit specification of shared conceptualization. Ontology is used in applications to store data in a hierarchical format. As ontology acts as a repository of knowledge bases, it can be used in collaboration with knowledge–based RS. By encapsulating data hierarchically and reducing the time taken for searching an item, ontology overcomes the problem of information overload. Also, the ontology can be reused for other suitable applications in the same domain.


The following are good research papers related to recommender system:

  • Classification of Recommendation System for E-commerce Application.
  • Classification of Recommendation System for E-commerce Application.
  • Building a Book Recommender system using time based content filtering.
  • Recommendation System Based on Combine Features of Content Based Filtering, Collaborative Filtering and Association Rule Mining.
  • Library Book Recommendations Based on Latent Topic Aggregation.
  • Hybrid Book Recommender System for an E-Commerce Application.
  • Book Recommendation System Using Opinion Mining Technique.
  • Combining Content –Based and Collaborative Filters in an online newspaper.



Leave a Comment


Enjoy this blog? Please spread the word :)