Wanna become a data scientist? Checkout Beyond Machine!
This year I had the chance to teach Social Media Analytics at the Cyprus International Institute of Management. Social media analytics is not really recognised as an independent field. It is more of a collection of techniques applied to data gathered from social media. However, I believe that we should start teaching social media analytics as a separate course. There are two reasons for that.
First of all, social media have a huge impact on our lives. With more than 2.5 billion users worldwide, this is not really up for debate. Data extracted from social media is used for all sorts of purposes from building recommender systems for products, to building statistical models that affect voters and elections.
Secondly, every field related to data science has its own domain-related peculiarities. For example, econometrics deals a lot with panel data, and transparent statistical model. Bioinformaticians deal with big datasets with multiple interacting variables. Natural language processing requires powerful methods for feature selection and cleaning. Social media is no exception, with specific algorithms and tools being better suited to this domain than others.
I have more than 5 years of experience working with social media data. In my work with Brandtix I had the opportunity to work with data coming from Facebook, Instagram and Twitter and do all sorts of things: topic modelling, sentiment analysis, forecasting engagement and many more. Before that I had also explored the potential of social media for predictive analytics, and more specifically I used Twitter data to predict football outcomes.
Based on my experience, I designed the curriculum you can see below. The course is also offered through Experfy.
Natural language processing
Most data from social media comes in the form of text. And not just text, but very noisy text in short chunks. In social media we get emoticons, poor syntax and grammar, many different languages. Being good in NLP is an important prerequisite for analysing most of the data that is coming from social media. Plus, you can do all sorts of useful stuff, such as sentiment analysis and topic modelling.
Statistics for engagement metrics
How can you measure popularity? How to compare the engagement rates of two pages? How can you forecast an official page’s likes and comments count? Answering all these questions
Social network analysis
Many things in social media are related to networks. Facebook’s API is called GraphAPI, specifically because everything is set up as a network. Your friends on any social media platform form a network. Therefore, it makes sense to learn the techniques to analyse and understand networks. From the centrality of a node to the clustering coefficient.
Recommender systems is a field on its own. However, I believe that since recommender systems are so closely linked to social media these days, it also makes sense to have at least one session dedicated to them. Recommendation engines are related to social media in two ways: they can use social media data, and they are used within social media networks. Your facebook wall, the friend recommendations, all these things are run by recommenders.