Collective Prediction of Individual Mobility Traces for Users with Short Data History

Area of Science:

• Computational Social Science
• Data Science
• Human Mobility Modeling

Background:

• Predicting human mobility is crucial for various applications.
• Existing methods struggle with users exhibiting short or non-repetitive movement patterns, like tourists.
• Limited individual historical data hinders accurate next-location prediction.

Purpose of the Study:

• To develop and evaluate a novel sequential learning algorithm for enhanced human mobility prediction.
• To improve prediction accuracy for individuals with sparse behavioral data.
• To leverage collective user data for individual mobility forecasting.

Main Methods:

• A sequential learning algorithm was developed, utilizing large datasets of user movement sequences.
• The algorithm incorporates data from multiple users to infer typical behavioral patterns.
• The method was tested on a substantial dataset comprising 10 million mobile phone users.

Main Results:

• The proposed algorithm demonstrated significantly higher prediction accuracy compared to traditional methods, such as Markov models.
• Performance gains were particularly notable for users with limited historical mobility data.
• The algorithm effectively utilized shared behavioral patterns from the larger user base.

Conclusions:

• The sequential learning algorithm offers a robust approach to enhance human mobility prediction accuracy.
• Leveraging collective sequence data is an effective strategy to overcome limitations of individual data scarcity.
• The algorithm's applicability extends to various sequential prediction tasks with rich, diverse datasets.