Accommodating user diversity for in-store shopping behavior recognition

Abstract

This paper explores the possibility of using mobile sensing data to detect certain in-store shopping intentions or behaviours of shoppers. We propose a person-independent activity recognition technique called CROSDAC 1 , which captures the diversity in the manifestation of such intentions or behaviours in a heterogeneous set of users in a data-driven manner via a 2-stage clustering-cum-classification technique. Using smartphone based sensor data (accelerometer, compass and Wi-Fi) from a directed, but real-life study involving 86 shopping episodes from 30 users in a mall's food court, we show that CROSDAC's mobile sensing-based approach can offer reasonably high accuracy (77.6% for a 2-class identification problem) and outperforms the traditional communitydriven approaches that unquestioningly segment users on the basis of underlying demographic or lifestyle attributes.

Analyzing Shopper's Behavior through WiFi Signals

“…To address this, CSI-based sensing can be combined with activity recognition through shopper's wearable/mobile devices [6,7]. Such hybrid approach can overcome some of the limitations of CSI-based analytics by improving classification with multiple shoppers and fine-grained activity recognition.…”

“…Simply relying on video surveillance to understand shopper's behavior is not scalable, given that deployment of video cameras and mining the video stream to extract information can have very high cost along with some serious privacy implications. Recent works [6,7] have proposed to use a user-driven approach where inertial sensors and camera available on shopper's wearable devices are used for physical analytics. However, this approach requires the shoppers to carry such devices and also e↵ectively communicate the acquired information to the business owner.…”

Opportunities and challenges in multi-modal sensing for regular lifestyle tracking

“…However a major drawback of personalized models is that they do not scale easily. In [10], we showed that with minimized personalized training, we could build a system which could recognize shopping behavior moderately well. Data from users who were similar to the individual was used to identify the behavior.…”

“…Shopping -The first ADL that we wanted to monitor was the individual's shopping behavior [10] -specifically, to distinguish between users with no buying intentions and shoppers (with buying intent) who were confused or focused. We performed a small scale user study with participating shoppers, where we collected sensor data (accelerometer, gyroscope and Wi-Fi) from their smartphones.…”

“…All findings and recommendations are those of the authors and do not necessarily reflect the views of the granting agency, or Singapore Management University. combination includes shopping behavior analysis [10], where sensor data from the phone is augmented with infrastructure based location data and group-level interaction monitoring [9] where peer smartphones can help capture correlations of sensor readings among individuals in a crowd.…”