CHAPTER 6Learn, Map, and Recommend

User interaction is one of the most commonly collected type of data in social media platforms. People use online platforms to make decisions about what goods to purchase, what movies to watch, when to interact with friends, and so on. All these decisions constitute valuable information to help you understand users’ behavioral patterns. Social media platforms also consist of several different components where people make choices and their choices can easily be logged for further processing. This chapter elaborates on how to model this type of data and extract meaningful, valuable information, and use this for decision making and recommendations.

Recommending items to users in social media has different problem stages depending on the age/state of the social media service we're working with. For example, suppose you have a new platform in which you jump start the product by providing new content to new users. This is a unique problem—the so-called cold-start problem. In another case, suppose there is some level of maturity in the product with some history of user engagement. This problem—the mainstream problem—is significantly different than the cold-start problem. This chapter starts by assuming some level of history and data liquidity. This can help you understand the fundamentals of making recommendations. The end of the chapter talks about edge cases such as the cold-start problem and incorporating covariates.

To discuss the mainstream ...

Get Social Media Data Mining and Analytics now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.