[Paper Review] Scalable Mining of Daily Behavioral Patterns in Context Sensing Life-Log Data

Reza Rawassizadeh, Momeni, Elaheh+1 more|arXiv (Cornell University)|Nov 18, 2014

Human Mobility and Location-Based Analysis43 references3 citations

TL;DR

This paper proposes a scalable, sensor-agnostic framework for mining daily behavioral motifs from real-world life-log data using multiple smartphone sensors, applying human-like temporal granularity to improve pattern detection accuracy despite data uncertainty. The method achieves 55% confidence in motif identification and improves scalability via sliding windows, with one-hour temporal granularity yielding the highest accuracy.

ABSTRACT

Despite the advent of wearable devices and the proliferation of smartphones, there still is no ideal platform that can continuously sense and precisely collect all available contextual information. Ideally, mobile sensing data collection approaches should deal with uncertainty and data loss originating from software and hardware restrictions. We have conducted life logging data collection experiments from 35 users and created a rich dataset (9.26 million records) to represent the real-world deployment issues of mobile sensing systems. We create a novel set of algorithms to identify human behavioral motifs while considering the uncertainty of collected data objects. Our work benefits from combinations of sensors available on a device and identifies behavioral patterns with a temporal granularity similar to human time perception. Employing a combination of sensors rather than focusing on only one sensor can handle uncertainty by neglecting sensor data that is not available and focusing instead on available data. Moreover, by experimenting on two real, large datasets, we demonstrate that using a sliding window significantly improves the scalability of our algorithms, which can be used by applications for small devices, such as smartphones and wearables.

Motivation & Objective

• Address the challenge of data uncertainty and loss in real-world mobile sensing due to hardware, software, and battery limitations.
• Develop a scalable, generic approach to detect human behavioral motifs using diverse sensor data from real smartphones, not just idealized hardware.
• Improve behavioral pattern detection by aligning temporal resolution with human perception of time, using a novel temporal granularity transformation.
• Enable accurate behavioral profiling for applications on resource-constrained devices like smartphones and wearables.

Proposed method

• Utilize a combination of smartphone sensors (GPS, WiFi, Bluetooth, call logs, etc.) to infer user behavior, focusing on available data and ignoring missing sensor readings.
• Apply a temporal granularity transformation that maps raw timestamps into human-perceptible time segments (e.g., 15:00–16:00) to align with natural time perception.
• Implement a sliding window mechanism to enhance algorithm scalability and reduce execution time for large-scale life-log data.
• Define a behavioral motif as a time-stamped, confidence-scored combination of sensor data (e.g., stationary state + call), enabling pattern mining across diverse data sources.
• Use a multi-source fusion strategy to reduce uncertainty by relying on sensor redundancy rather than single-sensor dependency.
• Apply a damped window approach in future work to weight motif relevance based on average motif lifetime.

Experimental results

Research questions

• RQ1 How can behavioral motifs be reliably detected in real-world mobile sensing data with inherent uncertainty and data loss?
• RQ2 What temporal granularity best aligns with human perception and maximizes motif identification accuracy?
• RQ3 To what extent does using a sliding window improve the scalability of motif detection algorithms on large-scale life-log data?
• RQ4 Can a multi-sensor fusion approach reduce uncertainty and improve robustness in behavioral pattern mining compared to single-sensor methods?
• RQ5 How do users' behavioral motif distributions vary, and can they be grouped meaningfully for adaptive system behavior?

Key findings

• The proposed method achieved a motif identification accuracy of 55% confidence in real-world life-log data, demonstrating robustness to sensor uncertainty.
• Using a one-hour temporal granularity interval yielded the highest accuracy in motif detection, outperforming coarser or finer granularities.
• Applying a sliding window mechanism significantly reduced execution time, enhancing scalability for deployment on mobile devices.
• The system successfully grouped users into three distinct behavioral clusters based on their temporal motif distributions, enabling personalized adaptation.
• The dataset of 9.26 million records, collected from 35 real users with diverse smartphone brands, is the most realistic life-log dataset to date, reflecting real-world deployment challenges.
• The multi-sensor fusion approach reduced dependency on any single sensor, improving resilience to data loss and hardware variability.