Essays on Decision Support Systems (Topic C.5)

– Develop new approaches to machine learning and use them to reduce product returns –

The motivation for my dissertation is to make both theoretical and practical contributions to the field of machine learning and predictive analytics. In addition to developing new machine learning algorithms I also develop methods that can be used to apply machine learning algorithms to hypothesis testing. In my applied research, I want to focus on product returns in online retail.

In the current decision support systems literature, the development of the prediction method and a decision support system is seen as the research question itself. These methods are hardly ever used to gain a deeper understanding of the underlying social phenomena. By contrast, one major goal of my dissertation is to explore ways in which these techniques can be applied to evaluating hypotheses.

Shmueli and Koppius (2011) discuss the possibilities of applying predictive analytics to the information systems literature. They also point out that such approaches, promising as they may be, have been given little attention in the information systems literature. They also define criteria for a rigorous application of predictive analytics and particularly emphasize the need for an out-of-sample examination of the decision support systems considered.

However, they do not emphasize the importance of testing the statistical significance of the out-of-sample predictions or whether the differences in predictions between different decision support systems are statistically significant. I believe that this is important for the traditional approach to decision support systems, but even more so when decision support systems are used to test hypotheses. Developing a suitable statistical framework is one of the contributions of my dissertation.

Product returns are not only an important cost factor for businesses, but also have significant environmental impact: About 10% of all products returned can no longer be sold and have to be discarded (Pur et al. 2013), leading to a waste of resources. In addition, transporting returned product leads to the emission of greenhouse gases: In Germany alone, product returns cause 143.000 tons of additional CO2 emissions annually (Asdecker 2014).

Product returns have been analyzed in the marketing literature, but little research on the issue has been conducted using predictive analytics. Moreover, academic research on the issue of product returns has so far ignored the issue of sustainability and their environmental impact.

I want to use predictive analytics to not only observe behavioral patterns, but to also develop a deeper understanding of factors influencing these patterns. Based on these findings, I then want to develop a decision support system that can be used to reduce product return rates.

Asdecker, B. 2014. Statistiken Retouren Deutschland - Definition, http://www.retourenforschung.de/definition_statistiken-retouren-deutschland.html, retrieved 2014/5/28.
Pur, S., Stahl, E., Wittmann, M., Wittmann, G., and Weinfurter, S. 2013. Retourenmanagement im Online Handel - Das Beste daraus machen, Regensburg: ibi research an der Universität Regensburg GmbH.
Shmueli, G., and Koppius, O. R. 2011. ?Predictive Analytics in Information Systems Research.,? MIS Quarterly (35:3), pp. 553?572.