ASSOCIATION RULE MINING WITH TWEETS: THINKING OUTSIDE THE BASKET

Abstract

In the past few years, the usage of social media around the world has impacted people’s lifestyles because of the widespread use of recommender systems. As Twitter being of the main social media platforms, a high interest on text mining for this social media platform has been observed. Because of this, the need of exploring relationships among words in Twitter data is of a high importance. The main purpose of this thesis is to observe existing association rules mining techniques and use those techniques to build correlations between words in tweets. Market Basket example is the most common example used in association rule mining. By using the logic of market basket practice, this study aims to use a similar approach to build a logical file composed of tweets to understand the associations between tweets. This thesis allowed me to understand better on the steps needed to take, in order to have a well formatted file containing the tweets by using R programming. Results show us that different words between different tweets have a high measure of Lift, making the association rules between words meaningful.