Lindsay Berry, PhD Student at Duke University.
The 2018 ISBIS World Meeting was recently held in Piraeus, Greece from July 4-6, 2018. Along with ISBA, held the week before in Edinburgh, Scotland, this was my first time attending a conference outside of the United States.
As a third year PhD student in Statistical Science at Duke University, this trip was an opportunity to present my work at two statistical conferences and travel to two interesting European cities. In the days before the conference, I visited the Acropolis and other historical sites, explored the National Archaeological Museum and Acropolis Museum, shopped at the flea market in Monastiraki, and ate lots of Greek food.
At the ISBIS meeting, I presented my research developed with Mike West in a collaboration with 84.51°, a subsidiary of Kroger (the largest supermarket company in the USA). A key motivation of our work is product demand forecasting – forecasting daily sales of products based on historical sales data. This is a priority for large retailers for many reasons including inventory control, production planning, and marketing decisions. The main challenges are the need for on-line forecasts updated in real time, and the high dimensionality of the time series forecasting problem that involves many stores, thousands of individual products, and multiple forecast horizons. In my talk, I presented the Dynamic Count Mixture Model (DCMM), a novel, univariate model I developed to flexibly model time series of non-negative counts. I then discussed a new multi-scale framework that effectively and efficiently incorporates cross-series linkages to borrow information across similar products. More details about this work can be found in our paper.
Coming from a large conference like ISBA, the ISBIS World Meeting was a somewhat more intimate setting to meet and engage with other statisticians. Given the emphasis on business and industrial statistics, there was a mix of statisticians from both academia and industry. After my talk, I met statisticians working in industry that shared their real world experiences of working on similar problems at other companies. Additionally, I attended a range of interesting sessions and learned about statistics and data science techniques being used in industry. A number of these sessions were relevant to my interests in Bayesian modeling, time series modeling and forecasting of counts and non-Gaussian data, and the challenges and techniques for handling big data.
Overall, attending the ISBIS World Meeting was a valuable experience with opportunities to meet and interact with statisticians from a wide variety of backgrounds. I would recommend the ISBIS conference to other young statisticians interested in the field of business and industrial statistics.