Recommend a Business to Me!

About

One way to create a network between the businesses in the Yelp dataset was to look at the sequences of businesses that users reviewed. To create our weighted network, we organized each user's reviews by their timestamps. Each business became a node, and each pair of businesses that a user reviewed in sequence became an edge.

To illustrate this process, imagine that businesses A, B, and C exist. A first user reviews A, then B, then C. As a result, an edge is drawn between (A, B) and between (B, C). A second users reviews B, then C, then A. Now, the edges (B, C) and (C, A) are added to the weighted network.

This kind of network is helpful because, if a user rates business A well, the business with the highest weight could be recommended to the user. For example, the edge between business B and business A might have a weight of 120, meaning that 120 users rated business A and then business B, or vice versa. Therefore, business B would be recommended to a user who liked business A.

This kind of network could be helpful in a cold start situation where very little is known about the user. It would be most helpful in a situation where there is an abundance of data, as networks with low edge weights might not be as effective in producing quality recommendations.

BUSINESS SEQUENCE NETWORK CREATED BY: John Anthony Bowllan, Kailey Nobuko Cozart, and Mohammad Mehdi Seyednejad.

WEBSITE CREATED BY: Kailey Nobuko Cozart, drawing partly from the design of Will Robots Take My Job?