Talk Of the Town


Project Association Of Facebook Likes:

Also called Market Basket Analysis
So Whats The Story:

We have data on the things that were "LIKED" by a large collection of users of Facebook (Koisinski, Stillwell, & Graepel, 2013). By analogy to market basket data, we can consider each of these users to have a "basket" of Likes, by aggregating all the likes of each user.

Now we can ask, do certain Likes tend to co-occur more frequently than we would expect by chances? We will use this simply as an interesting example to illustrate association finding, but the process could actually have an important business application. If you are a marketer looking to understand the consumers in a particular market, you might be interested in finding patterns of the things people like.

If you are thinking data-analytically, you will apply the sort of thinking I would apply: you will want to know what things co-occur more frequently than you would expect by  chance.

Since we are using market basket as an analogy at this point, we should consider broadening our thinking of what might be an item.

Why cant we simply put anything in the basket for association, lets see why?

For example, we might put a user's location into the basket, and then we could see association between likes and location. For our Facebook data, we might obtain psychometric data on many of the customers, such as their degree of extroversion or agreeableness, or their score on a IQ test. It may be interesting to allow the association search to find associations with these psychometric characteristics as well.

OK, so let's see what association we get among Facebook Likes. These associations were found using the popular association mining system  OAuth v2,

Now we are working with core API system, you dont have to do everything from scratch here python will take care of most of it.

OAuth v2 allows searching for association that give the highest lift or highest leverage, while we filter out association that are less interesting. Some of the interesting association are:

Pink Floyd & Slipknot & Sytems of a Down -> Korn

Support=0.010; Strength=0.675; Lift= 23.99; Leverage= 0.0097


I am still working on the project as I am a beginner at this and I need a lot of more support, So any of you guys are free to post comments and mails.

Thank you, keep you posted


Leave a Reply

Your email address will not be published. Required fields are marked *