DataScientia at the ECAI 2024

Fausto Guinchiglia and Xi Xiaoyue Li from Data Scientia introduced the innovative “Big-Thick Data Generation via Reference and Personal Context Unification” at the conference.

Smart devices generate vast sensor data that predict human behavior but lack contextual depth. To address this, they introduced Big-Thick Data, which combines users’ subjective personal contexts with an objective reference context for richer, self-explanatory insights.

The Big-Thick Data model merges large-scale big data with detailed thick data, combining objective Reference Context and subjective Personal Context to create a context unification. It includes Etype and Property Unification (EPU), which matches entity types and properties; Spatio-Temporal Unification (STU), which links entities using location and time; and Entity Unification (EU), which matches entities by identifiers, names, and attributes.

This approach enables the collection of diverse questions and answers about user behavior and surroundings. A case study with university students’ data and OpenStreetMap demonstrates that combining these data types leads to better predictions, showing how adding context improves data quality and relevance.

Fausto and XI validated this approach with a four-week study of 158 students, integrating personal data with OpenStreetMap’s reference context. The study achieved high predictive accuracy, like 94.42%.

Here bellow is the visual representation of the study: