PERANCANGAN MODEL PREDIKSI PERTUMBUHAN FOLLOWER MEDIA SOSIAL INSTAGRAM (KASUS AKUN @INFIA_FACT)

• Main Author: Pramudita Larisa Kira, Elisabet
• Format: Final Project
• Language: Indonesia
• Subjects: Media berita, jurnalisme, penerbitan
• Online Access: https://digilib.itb.ac.id/gdl/view/54230
• Institution: Institut Teknologi Bandung

Infia is a media and advertising-based company. One of Infia’s core businesses is providing news in various categories through several accounts on Instagram. The largest news account held by Infia is @infia_fact with the total number of followers of 3,1 million Instagram users. Unfortunately in the past three years, @infia_fact’s follower growth experiences high level of fluctuations and tend to have a declining trend. In order to maintain its position in regards to other competitors, Infia attempts to perform empirical analysis by comparing follower growth periodically and analyze root causes of the fluctuations qualitatively and speculatively. The purpose of this study is to design a quantitative model which is able to predict follower growth. This model will assist Infia in identifying variables influencing follower growth of @infia_fact. This model is created based on 10.297 Instagram posts from 983 days. Follower growth prediction model is developed by creating 2 separate models: models to predict follower gain and loss, then combining these models to come up with the final model. From 5 modeling methods used, multiple linear regression and random forest (regressor) produced the best performance and ease of interpretation. The values of R2 for training and testing data set of the follower gain prediction model based on multiple linear regression and random forest method in sequence are as follows: 53.7%, 57.7%, 95.5%, and 74.5%, whereas for the follower loss model are 34.8%, 35.8%, 90.5%, and 52.3%. Variables chosen in this model are time period t, number of characters in a post’s caption, number of hashtags used in a post, number of contents posted per day, percentage of contents in a form of photo, percentage of contents in a form of carousel, and interaction between percentage of contents in a form of video and percentage of contents posted at 00.00-12.00 GMT+7. This research concludes advices for Infia regarding content creation based on variables influencing follower growth and strategies to improve operational business processes to increase the performance of @infia_fact in terms of follower growth.