Dr. Vandana Janeja and PhD Student Lei Shi to present their paper on Mining anomalous windows along Intersecting Paths at this year’s KDD Conference in Paris, France

Date Posted: - 4/29/2009

It is said that when a vehicle accident occurs at an intersection many factors are involved to determine how and why it happened. When this occurrence happens more frequently in the same location over time, the ability to mine rich spatial and temporal data becomes key to not only identifying existing “accident hubs” but it also may be beneficial in projecting where new intersections may emerge as trouble spots in the future.

As the population grows, the increase in people naturally adds to the number of vehicles traveling our seemingly endless chain of arterials—especially during peak commute periods. As a result, older road networks (intersecting boulevards, off-and on ramps to highways and interstates, “jug handles,” and roundabouts) may require more attention, assessment, and possibly updating or re-routing altogether. A component to overall assessments of this sort may lie in the research being done in the Department of Information Systems (IS) at UMBC.

Dr. Vandana Janeja, Assistant Professor in IS at UMBC, along with PhD student Lei Shi, will present their paper entitled "Anomalous Window Discovery through Scan Statistics for Linear Intersecting Paths (SSLIP)" at the 15th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, to be held June 28 through July 1, 2009, in Paris, France.

The paper proposes a method to find accident hubs along intersecting paths such as road networks. The student, Lei Shi, is in his first year of the PhD program and began working with Dr. Janeja in September and October of 2008. Dr. Janeja remarked about PhD student Shi saying, “He is a very hard working student and in a short time developed several algorithms for this paper and also a demo that is now available for review online.”

What makes Dr. Janeja’s and Lei Shi’s paper acceptance additionally noteworthy is being selected among the largest number of paper submissions in the conference’s history. According to the conference’s website, KDD'09 received a record-breaking number of submissions: 561 for the Research Track (up 10% from 510 last year), and 125 for the Industrial/Government Applications Track (up 50% from 83 last year), with a 15% increase in overall submission numbers from last year. The ACM SIGKDD conference upholds high standards for papers that are submitted, expecting paper submissions to “describe innovative ideas and solutions that are rigorously evaluated and well-presented.” Papers are judged based on their technical merit, rigor, significance, originality, repeatability, relevance, and clarity.

Dr. Janeja says, “Generally the trend is that the conference accepts about 10-15% of the papers submitted as full research papers, so this acceptance happens to be associated with the most selective of the Data Mining conferences in our discipline.” She further commented, “Coincidentally, the other top conference of Data Mining is SIAM DM where Dr. Aryya Gangopadhyay, Associate Chair for Academic Affairs for UMBC’s Department of Information Systems, I, and another student, also have a paper accepted this year.” Dr. Janeja will also present the paper at this year’s SIAM DM Conference.

The annual ACM SIGKDD conference is the premier international forum for data mining researchers and practitioners from academia, industry, and government to share their ideas, research results and experiences. The KDD conference falls under the umbrella of ACM-SIGKDD and will feature keynote presentations, oral paper presentations, poster sessions, workshops, tutorials, panels, exhibits, demonstrations, and the KDD Cup competition.