Classifying Professional Photographers on Instagram: Data Collection and Processing for Computational Learning

Abstract

Nowadays, the surge in open data on the internet allows researchers to investigate and broaden the understanding of numerous significant disciplines. However, there remains a notable deficiency in the advancement of methodologies for identifying artistic skills, particularly in the field of expertise finding, due to their subjectivity and the shortage of available datasets. Thus, we saw an opportunity in the popularity of photo sharing platforms to create a dataset for the identification of professional photographers’ profiles. Our first contribution is a comprehensive, multimodal dataset that encompasses a wide array of attributes from 29 679 Instagram posts, originating from 1042 corresponding user profiles labelled as professional or not professional photographers. Employing this extensive dataset, we explored different machine learning (ML) models to assess their efficacy in classifying these profiles into their respective categories. The Random Forest (RF) model showed the best performance, being able to understand the common structure for professional photographers Instagram profiles. Further statistical analysis revealed significant distinctions between both types of profiles. The most important features for identifying a professional photographer are the number of users tagged, the technical score in their posts, and the height variance of the pictures made. The results obtained in this work hold the potential to significantly inform future research and offer practical applications across multiple real-world scenarios.