A Hybrid RFR–BiLSTM Framework for Social Media Engagement and Web Traffic Prediction

Abstract

The amount of user interaction data generated per second by social media platforms and other websites is staggering. Understanding the variations of social media engagement and its impact on website traffic is an important line of inquiry for organizations of all types and sizes, including businesses, public policy bodies, and researchers. This study proposed a hybrid models framework that combine machine learning and deep learning approaches to examine the extent of correlations among social media interactions (likes, shares, comments, and impressions) and web traffic access methods (page views, bounce rate, and session duration). A hybrid approach to data analysis was used employing real-world datasets from multiple data sources - Facebook, Twitter, and Google Analytics. The model used Random Forest Regression for feature importance selection, then used BiLSTM for sequential traffic forecasting. The outcome showed relationships existed between social media engagements and website traffic resulted in findings with correlation coefficients over 0.7 for shares and impressions, showing social media engagement positively influenced web traffic. The hybrid model had a better prognosis than classical regression, and offered improvement perspectives than standalone neural networks. The research also confirmed hybrid and integrated models can enhance digital marketers and customer analytics research.