Maskininlärning som instrument för att analysera Twitter : En studie kring datahantering på Twitter för tillämpning på riktad marknadsföring

Sammanfattning: Marketing in social media becomes increasingly more common in today's society thanks to the rapidly expanding digitalization, while the interest and integration of machine learning in everyday products and services has increased. The purpose of this paper is to investigate the possibility of determining the interests of Twitter users from their everyday tweets with the help of machine learning and thus target relevant advertising. To achieve this goal a Bag of Words approach, a supervised machine learning method, was used to collect data from Twitter users using the Python based library Tweepy. The collected data, consisting of the user's most common words was compared to predetermined interest-classified glossaries. All of the selected Twitter users were also given a survey where they would rate their interest areas, which would later be used train the program. The program's and the surveys' results were compared to determine the deviation whereas the program was further enhanced with the training data to achieve an improved precision. The program was enhanced to a sense, with the cosine similarity angles small enough, that it could be considered effective. An analysis of targeted marketing was made to investigate the possibilities of applying machine learning on collected data. This study shows that it is possible, with a cosine similarity less than 45o, to determine the interests of a user using data retrieved from Twitter and suggest how the applied method can be further improved.