The authors participated in the HuMob Challenge, a competition for predicting human mobility. They developed a personalized model that predicts an individual’s movement trajectory based on their unique data. This model considers features like date, time, activity duration, weekdays, time of day, and frequency of visits to Points of Interest (POI). They also used clustering to include the movements of other individuals with similar patterns. This model, which uses Support Vector Regression, was tested for accuracy and underwent feature selection and parameter tuning. Despite using a traditional feature engineering approach, it achieved good accuracy with low computational cost.
Publication date: 20 Oct 2023
Project Page: Unavailable
Paper: https://arxiv.org/pdf/2310.12900
