customer network visualization erik feng dawei shen kwan hong lee dr. marko popovic
DESCRIPTION
My Demo: Basics Retrieves data from a data file stored online Creates a graph based on the data Nodes represent customers. Links indicate that two customers visit the same restaurant. node linkTRANSCRIPT
Customer Network Customer Network VisualizationVisualization
Erik FengDawei Shen
Kwan Hong LeeDr. Marko Popovic
ConceptConceptUse a dynamic graph to identify
groups of consumers based on spending pattern
A graph is a network consisting of points or nodes connected by links.
Our project analyzes customer purchases at their top five restaurants.
Of course, the scope of the research is not limited to just restaurants, but can be applied to any industry. Food is an essential item that everybody will purchase, which would make data more readily available.
My Demo: BasicsMy Demo: BasicsRetrieves data from a data file
stored onlineCreates a graph based on the
dataNodes represent customers. Links indicate that two customers
visit the same restaurant. node
nodelink
Existing Graph Existing Graph VisualizationsVisualizationsGraph structure has to be
manually constructed to look nice.
Many intersecting links and nodes look messy.
Our Demo: FeaturesOur Demo: FeaturesUses physics concepts to
dynamically rearrange structure.Links act like springs that stretch
and shrink. Nodes act like charged particles
that repel each other to avoid overlapping and obscuring data.
Ability to see only the links of one node at a time
ScreenshotsScreenshotsInitial arrangement – nodes are placed randomly in a circle.
Node size indicates how much each customer spends in total.
Node color indicates how well connected it is. The more eye-catching yellow is used for nodes with the most connections, followed by red, magenta, and blue.
ScreenshotsScreenshotsAfter calculating the equilibrium forces, the graph settles down. Node repulsion follows the (r -2) law while the springs follow Hooke’s Law.
The length of each link is determined by the percentage of the total spending that the two corresponding customers spend at their shared restaurants.
ScreenshotsScreenshots
Mousing over a node highlights its links and makes extraneous links disappear. The restaurants of that node appear in the textbox above.
ScreenshotsScreenshots
Graph is interactive: users can physically drag nodes around if they wish to customize the arrangement.
Try It Out:Try It Out:http://mit.edu/erikfeng/www/Restaura
nt.swf
How Banks Can Use ThisHow Banks Can Use ThisBehavioral analysis
◦Customers A smaller node linked closely with several
large nodes can be a warning that a relatively poorer person is spending a significant proportion of his money at fancy stores
◦Businesses We can redefine nodes as businesses and
links as shared patrons. Businesses with declining number of patrons may present a greater risk if they wish to borrow money to make investments.
How Banks Can Use ThisHow Banks Can Use ThisNetwork of Financial Services
◦A bank can construct its own graph with different financial services (such as demand deposit account, savings account, credit card, etc.) as nodes and links if there are customers that use both services. The idea of a physical equilibrium can help a bank determine which services could use innovative development and which (if any) are obsolete.
Future ResearchFuture ResearchWith more data collected for a
period of time, we can construct a graph that changes over time. We would be able to observe changes in behavior, spending, or tastes.
We could develop a social network similar to Facebook that connects people based on similar restaurant preferences, musical tastes, etc.