Mall Crawler: Social Network Analysis for Reality Mining of Human Shopping Behavior
Over 50% of American’s are owners of smart phones and more than 70% of these have their devices for comparison shopping. Every time a user uses their mobile device they leave behind a few bits of information. Reality mining capitalized on the usage of wireless devices such as mobile phones and GPS systems providing a more accurate picture of what people do, where they go, and with whom they communicate with. The use of cell phones have become an important platform for the understanding of social dynamics and influence, through their pervasiveness, computational power, and sensing capabilities. The use of mobile influence is expanding and is leading to more sufficient productivity sales, cost savings and overall optimized statistical analysis. Currently, mobile influence is creating a sustainable advantage within the industry. Today’s retailers are shifting in an alarming agile rate, as mobile is having major impact on store purchases and product consumption. Once consumers start using their smartphones for shopping, they tend to use them again. The user establishes a more habitual behavior within smart phone use. These attributes clearly are being considered within organizations and marketing for future gain.
Researchers at The Johns Hopkins University Applied Physics Laboratory (APL) have invented common social network analysis algorithms to mall shopping behavior with the intent of shopping behavior through incentives.
This invention describes the procedure used to collect mall shopping behavior, converting shopping behavior into a social network, processing social network data, and visualize the results. Using multiple algorithms the researchers were able to represent influence and attractiveness of stores within a mall. It describes ideas for making this information financially valuable to store retailers.
When linked to the concept of “reality data mining” the invention represents financial value, where the information is generated from apps on a smartphone. If the smartphone is used as a “virtual coupon,” stores could incent behavior as a shopper visits stores in the mall, increasing the likeliness of making purchases at specific stores. Reality mining is an aspect of data mining and is a simple collection and analysis of machined sensed environmental data, which pertains to human social behavior and can further predict patterns of behavior. The idea is to measure people’s minute-to-minute behaviors, while recording important and often unnoticed details of our interactions with other people, habits and or interests. This data is enormous in viable untapped information. The data is captured and geo-tagged from the mass collective of unstructured data and then processed.
A survey was distributed through physical copies and attachments in e-mail requests. Participants were asked to visit the shopping mall and shop as they naturally would. While shopping they were to complete the survey describing the order in which they visited each store. For many stores, their prominence makes intuitive sense given the geographic distribution of stores. However, other stores with similar geographic advantages do not have nearly the same degree centrality scores. Also, it is difficult to determine without network analysis how much effect the geographic advantage plays within the mall. The degree centrality graph and in/out-degree scores give stores a measure of how connected they are. For example, investigating the Ego Network (nodes with in-links or out-links) of Bass Pro Shops gives additional information about the stores that are influenced by Bass Pro Shops. The in-links are the stores where a shopper visited before visiting Bass Pro Shops. With this information, Bass Pro Shops might investigate why a shopper would visit their store after the others; were the looking for a higher quality product, a better price, or were they simply looking for the exit? Even if the shopper was looking for the exit and had no intention of buying a product, Bass Pro Shops ownership might capitalize on this traffic by offering products that might lead to an impulse buy. The HITS algorithm consists of running a query against the product content of stores to determine a set of relevant stores. The differences between HITS compared to Degree Centrality and PageRank are even more subtle, but have the added benefit of representing both an authority (in-links) and hub (out-links) ranking reinforced by the hub and authority score of adjacent nodes. A high hub rating means the store is an influencer for many highly authoritative (attractive) stores. If the store is a profiteer of selling its influence to other stores, then it would benefit to have a high hub score.CONTACT:
Dr. G. R. Jacobovitz
Phone: (443) 778-9899