Quantifying Influencer Performance through a Data Mining Analysis of Engagement and Sentiment

In the context of digital commerce, particularly within niche markets like Muslimah fashion, quantitatively evaluating influencer marketing effectiveness remains a significant challenge. This study proposes and applies a framework to quantify influencer performance by conducting a data mining analysis of audience engagement and sentiment. The analysis focuses on 19 Instagram accounts, categorized into micro-and mega-influencer tiers, within the Indonesian Muslimah fashion sector. Utilizing the Cross-Industry Standard Process for Data Mining (CRISP-DM), user comments and interaction data were systematically collected and processed. A Multinomial Naive Bayes (MNB) classifier was implemented to perform sentiment analysis on this dataset. The findings indicate that micro-influencers achieve a higher average engagement rate (166.5%) than their mega-influencer counterparts (93.7%). However, the results also highlight that mega-influencers can generate exceptional engagement, demonstrating that content strategy is a critical performance factor alongside audience size. Sentiment analysis revealed an overwhelmingly positive audience response across all influencer tiers, with a notable absence of negative comments. Ultimately, this work provides a robust, data-driven methodology for influencer evaluation, offering empirical evidence that engagement metrics and sentiment polarity are more indicative of performance than follower count alone. These insights enable stakeholders to optimize digital marketing strategies through more informed, evidence-based influencer selection.