Detecting and Reducing Biases in Cellular-Based Mobility Data Sets.

Area of Science:

• Mobile computing
• Network performance analysis
• Urban planning and simulation

Background:

• Accurate human mobility estimation is crucial for mobile network optimization, mobility modeling, and urban planning.
• Existing research often relies on Call Detail Records (CDRs), which are derived from cell tower connections.
• The representativeness of CDRs for detailed user movement analysis remains a key question.

Purpose of the Study:

• To assess the sufficiency of Call Detail Records (CDRs) for accurately estimating key human mobility features.
• To compare the accuracy of mobility feature estimation using CDRs versus an alternative data collection method.
• To develop and evaluate filtering techniques for mitigating biases in mobility data derived from CDRs.

Main Methods:

• Analysis of two distinct mobile phone trace datasets, including CDRs and an alternative data source.
• Evaluation of mobility features such as fraction of visits per cell, entropy, entropy rate, and predictability.
• Development and application of three novel filtering techniques to address data biases.

Main Results:

• Call Detail Records (CDRs) may introduce biases in the estimation of mobility features, affecting accuracy.
• The study identified specific biases in distributions of visits per cell, entropy, and predictability.
• Proposed filtering techniques demonstrated potential in reducing detected biases.

Conclusions:

• The findings underscore the critical importance of considering the data collection methodology when interpreting human mobility patterns derived from mobile phone traces.
• Conclusions drawn from mobility studies are significantly influenced by the type of mobile phone data used.
• Contextualizing results with respect to data source is essential for reliable mobile network design and urban planning.