Recommender System for Franchise Business
Skills:Research, System Design, User Experience
Aim: Recommending Franchises to the client on the basis of his available resources.
Duration: 2 Months (Winter 2010 | Michigan, US)
Mentor: Asso. Prof. Rahul Sami
Download: Complete system design report (pdf)
Epitome: As the human life is quickly shifting towards the virtual side of the world, it has become important to assess and utilize the reputation system in virtual world. The reputation could be of any specific user on any social network site or on an ecommerce website. A detailed report describing the design, outlining the strengths and weaknesses of the system (including potential attacks), and explaining the design choices that were made was submitted.
Project Brief: This paper addresses the system design of a recommender system that will recommend Franchise to the Franchisees. There are about 3500 different Franchises available to choose from and all the Franchisors have their own expectations and requirements from the Franchises owner. As Franchisees have to totally rely on the knowledge of the Consultant and so there’s lots of room for manipulate the Franchise recommended by them.
This recommender system uses hybrid collaborative filtering algorithm to recommend list of Franchises. It matches both the interests of the Franchisees and their available resources, and what Franchises were opted by other clients with similar requirements. It also takes, error in recommendation, into account and updates the next recommendation accordingly.
Design and Methodology: Present cycle followed in the franchise business is shown in [Fig. 1].
A recommender system using two-tier hybrid algorithm was designed. This is an interesting situation and no single algorithm will give us the most appropriate results here. We can’t apply only User-User algorithm to find all the users who have similar values for the resources and then recommend the same list of Franchises (that were recommended to similar user) to the new user. They might not have the same level of matching in values for the resources and interests at the same time.
The graphical user Interface for the recommender system was developed and is shown in [Fig. 2]
Deliverables: More details could be read in the complete paper (Click here to download)