Predicting Influencers From Tweets Data

Norezzati, Md. Nor Hayati (2015) Predicting Influencers From Tweets Data. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Predicting Influencers From Tweets Data 24 Pages.pdf - Submitted Version Download (286kB)

Abstract

Twitter generates such overwhelming information every day and allows developers to get the public stream for analytic purposes. Influencer analysis is thus important to provide insight on market segmentation. However, there are not many researches done on identifying influencer because there exist no objective quantification to determine influencing factor. This project aims to find the influential factors and to cluster into the influencer and non-influencer clusters. The experimental was collected from Twitter on #NepalQuake topic. The attribute proposed for influencer prediction includes retweets, favorites, followers and friends. Simple scoring algorithm and Fuzzy C-Means cluster were use in the project. Data crawling was done through Twitter application in Python. Comparison was made to compare the performance of the simple scoring algorithm versus the Fuzzy C-Means cluster in generating influencer list. Both proposed algorithm is able to list and produce the most influential user. The result of influencer was obtained after the weightage is set to each attribute to represent the importance of attributes. The scoring algorithm will list the influential user with their rank. There is only one user for the influencer cluster. This is because only one user has obvious number of retweets, followers, favorites and friends. In future work, the algorithm can be improved to be commercialized. The measuring on how important the attribute need to be made and also the algorithm will be apply to the unstructured data.

Actions (login required)

View Item