Back To Schedule
Saturday, October 20 • 3:30pm - 4:00pm
Labeling Foot Traffic in Dense Locations

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.
Mobile sensor data is enabling us to better understand user behavior. It is now possible to accurately and persistently model foot traffic to brick-and-mortar retail stores and other points of interest in near real-time. However, this capability comes with increasing state-of-the-art data processing and machine learning challenges.
Locations of stores in dense areas, such as shopping malls, are often indistinguishable using lat/long or street address data because of their close geographical proximity. This creates a complex problem for accurate foot traffic estimation to these stores, especially in the absence of accurate, large-scale ground truth data. The problem is further aggravated by the diverse nature of user behavior and visit frequency to stores of different categories.
In this talk, I describe our approach at Sense360 to solve this. In particular, I will talk about framing this issue as a probabilistic learning problem, engineering features from point-of-interest data, and using regularization. The developed approach is currently being used at Sense360 to further boost the accuracy of our market research insights

avatar for Om Patri

Om Patri

Data Scientist, Sense360
Om is a Data Scientist at Sense360, a market research and insights firm in Culver City, CA, which enables some of the world’s largest restaurant and retail companies to continuously measure and optimize their business in real-time. He focuses on using AI and machine learning approaches... Read More →

Saturday October 20, 2018 3:30pm - 4:00pm
Ballroom # 403B