Business competitive analysis using promoted post detection on social media. (May 2020)

• Record Type: Journal Article
• Authors: Arora, Anuja; Srivastava, Aman; Bansal, Shivam
• Is Part Of: Journal of retailing and consumer services. Volume 54(2020)
• Publication Date: 2020-05
• Subjects: Brand promotion -- Ensemble model -- Post promotion -- Random forest -- XGBoost -- Bagging -- Boosting -- Logistic regression; Retail trade -- Periodicals; Service industries -- Periodicals; Customer services -- Periodicals; Commerce de détail -- Périodiques; Service à la clientèle -- Périodiques; Customer services; Retail trade; Periodicals; 658.87
• DOI: 10.1016/j.jretconser.2019.101941
• Languages: English
• ISSNs: 0969-6989

Abstract

Abstract: Social media has leveraged many brands and companies started interaction with customers through it. As the upsurge of brand popularity and their social growth on social media channels, the role to analyze and investigate the competitor's behaviours on social media becomes a crucial part. By utilizing these behaviours a brand can investigate and optimize their competitive strategies in order to enhance the audience reach, improved customer acquisition and increase in overall profits. Therefore, monitoring and analyzing the social media behaviour of competitors is a cutting edge direction towards the enhancement of competitive benefits and efficiently decipher the competition. This research paper is an effort to help brands to know the insights such as the posts which are promoted compared to those which are organic. This article presents the detailed description of novel approach applied for promoted post detection problem. Promoted post detection makes use of social media data to provide insights related to brand's own and competitors posting behaviours and promotional strategies. We make use of ensemble machine learning techniques-bagging and boosting to train different weak learners and combine the results by majority voting and weighted average. We make use of a logistic regression model in stacking while combining used generalized linear model and ensemble models. Results show the 95% accuracy, 97% precision, 96.5% Recall and 97% F1 score of classification results of brand promoted posts as boosted or organic. The result unveils that ensemble technique stands out as an effective approach for social media promoted post detection.