Multimodal Instagram User Dataset for Global and Local Interest Modeling

This dataset consists of Instagram user-generated content collected from 884 users over a ten-year period (2013–2023). It was created to support research on user interest detection, with a particular focus on identifying both global interests (long-term user preferences) and local interests (short-term or context-dependent preferences). The dataset contains a total of 86,996 Instagram posts, including both images and videos, providing a rich multimodal source of information for machine learning, deep learning, recommendation systems, user profiling, and social media analytics research. Dataset Statistics Statistic Value Number of users 884 Total number of posts 86,996 Number of image posts 47,144 Number of video posts 39,842 Coverage period 2013–2023 User Activity Distribution Metric Value Mean posts per user 98.4 Median posts per user 100 Minimum posts per user 0 Maximum posts per user 100 Standard deviation 11.3 Potential Research Applications The dataset can be used for: Global user interest modeling. Local and temporal interest detection. Multimodal user profiling. Personalized recommendation systems. Social media behavior analysis. Computer vision and multimedia understanding. Representation learning from user-generated content. Data Composition The dataset contains approximately 54.2% image posts and 45.8% video posts, ensuring a balanced multimodal representation of user activity. Most users contribute close to 100 posts, providing sufficient historical information for studying both persistent and evolving interests. Ethical and Usage Considerations The dataset contains content collected from publicly accessible Instagram profiles between 2013 and 2023. It is provided exclusively for scientific research purposes, particularly in the areas of user interest modeling, multimodal learning, and social media analytics. Users of this dataset must comply with applicable privacy laws, ethical guidelines, and data protection regulations, and must refrain from any attempt to misuse the data or identify individuals beyond the scope of legitimate research.