IBM Almaden Research Center

Quest

Past Almaden Project The QUEST group is working on a variety of projects including research on Data Mining (especially classification and clustering) information management for e-commerce including privacy issues, P2P applications (specifically personal utilities) and local area distributed computation structures. Associations Given a database of transactions, where each transaction consists of a set of items, discover all associations such that the presence of one set of items in a transaction implies the presence of another set of items. ``30% of people who buy diapers also buy beer.'' Synthetic Data Generation Code Classification Given examples of objects belonging to different groups, develop profile of each group in terms of attributes of the objects. This profile is then used to predict the group of a new object. ``Buyers of expensive sport cars are typically young urban professionals whereas luxury sedans are bought by elderly wealthy persons.'' Synthetic Data Generation Code Sequential Patterns Given a database of transactions over a period of time, find inter-transaction patterns such that the presence of a set of items is followed by another set of items. ``10% of people with diabetes develop a treatable loss in eyesight.'' Similar Time Sequences Given a database of time sequences, find sequences similar to a given one, or find all occurrences of similar sequences. ``The closing net asset value of the Harbor International mutual fund has been similar to that of Ivy International and Scudder Global Fund.'' Application Examples A direct mailer wants to maximize cross-selling opportunites. By applying the Associations and Sequential Patterns technique to historical order data, the direct mailer can find out what articles sell together and what articles are bought in a sequence over time. The mailer uses this information to decide on placements of articles in the catalog and for deciding what flyers to attach with a bill. A retailer wants to optimize purchasing and store-keeping. By applying the Similar Time Sequences technique, the retailer can find groups of products that have similar forecasted seasonal sales for next year and use this information for combining purchases and inventory replenishment. A bank wants to assess the credit-worthiness of its customers. By analyzing the loan-history records with the classification technique, the bank gets a precise profile of high, medium, and low-risk customers. An auto insurer wants to study lapsing and retention among their customers. By applying the Sequential Patterns technique, the insurer can understand what events lead to lapses. A medical insurer is interested in detecting insurance fraud. By applying the associations technique, the insurer can determine if there is a ring of providers indulging in ping-ponging of patients between them. The above is only a sampling of the many cross-industry tasks that can be enhanced by using these technologies. More than one technique can be applied to an industry and more than one industry can benefit from a technique.

Second Column