Machine learning models to predict the engagement level of Twitter posts: Indonesian e-commerce case study

• DOI: 10.1016/j.procs.2023.10.588
• Publication Date: 2023-11-25T12:03:25Z
• Authors: Dinda Thalia Anda...; Meditya Wasesa
• Keywords: Interactivity

The growing utilization of social media platforms enables direct interaction between companies and consumers. However, the expanding range interactions real-world data complexities necessitate development more sophisticated decision models. To address this, current research focuses on constructing machine learning models, namely multinomial logistic regression, tree, k-nearest neighbor, random forest, to forecast engagement level Twitter posts from three prominent e-commerce in Indonesia: Bukalapak, Blibli, Tokopedia. analysis comprises a dataset 12,786 unique tweets, accumulating 11,870,254 favorites 2,735,886 retweets over seven-month period February 1 August 31, 2021. prediction models are built upon theoretical constructs with seven features, encompassing interactivity (e.g., links, hashtags), vividness images, short videos, long videos), temporal factors day post, last post time). Factors such as frequency, interactive posting elements, static visual elements emerge significant features for predicting posts. Results demonstrate that forest model outperforms singular classifier including neighbor terms precision, recall, F1 score.