Analytics towards Computational Social Science

SMU’s Office of Research and the School of Information Systems jointly organised the recent 2013 Workshop on Analytics for Business, Consumer and Social Insights (BCSI 2013). The workshop was initiated in 2012 as a venue for interdisciplinary research presentations and postgraduate research tutorials in the area of data analytics for business, consumer and social insights.

Held over three days from 3-5 August, the workshop saw about 100 faculty members, postgraduate students and industry professionals from all parts of the world gathered at this event to learn from one another. Presenters addressed topics ranging from social commerce, privacy controls, and metaheuristics applications involving data analytics to innovative ‘big-data’-based empirical and experimental designs.

Business analytics – a critical enabler for organisations

Keynote Speaker: Professor Michael Trick, Senior Associate Dean (Education) at Tepper School of Business (Carnegie Mellon University)

Recent trends in data collection, algorithms and computing have radically changed the practical role of operations research and have given rise to the phrase ‘business analytics.’ As the trends continue, business analytics methods that take advantage of techniques from Operations Research, Management Science, Marketing Science, Information Systems and practice will need to evolve to better handle closer interactions between descriptive analytics, predictive analytics and prescriptive analytics.

Professor Michael Trick spoke about the past, present and future trends of Operations Research, reflecting the perspective of a core business analytics discipline.

He said that, in today’s Operations Research, business analytics is going through an exciting time and is seen as a critical enabler for strategic and operational decision-making in organisations. He identified three trends. (1) There is a tremendous growth in the amount of data. About 90 percent of data generated from the last two years came from enterprise resource planning (ERP) and other systems, etc. (2) Computing speeds for supercomputers have continued to achieve exponential growth. (3) Algorithms are now getting faster and faster. Fast algorithms change how optimisation and analytics work though – when optimisation techniques become faster, this eventually lowers the costs of decision support. However, most of the data still are being ignored and underutilised in most organisations.

The field of Operations Research has been changing over time. Professor Trick highlighted four future trends. (1) Organisations will make better decisions by linking predictive and prescriptive analytics with the use of data. The use of business intelligence and data mining through optimisation and simulations converts data into actionable information. (2) The increase in the assessment of the robustness of analytics solutions will allow companies to handle uncertainty more effectively. (3) Parallelism in computational solution methods will be more prevalent, as the prices of computers fall further. (4) Advanced analytics are needed to stay ahead of sophisticated adversaries who also use new decision support tools.

Cross-media exposure can increase customer conversion and improve sales

Assistant Professor Hang Sang Pil City University of Hong Kong

As companies divert more funds from traditional media towards digital advertising, they are interested in understanding what effects the two channels of advertising – web advertising and mobile advertising – have on consumer choices. Assistant Professor Han has been focusing his research on the relationship between web and mobile advertisements. His research shows that cross-media exposure on the Web and via mobile phones have positive impacts on click-throughs, customer conversions and sales volume. Cross-media ad exposure also may play a role as an informative reminder for consumers. Interestingly, his experimental study further found that the total sales associated with cross-media exposure doubled relative to single media exposure, such as using only web or mobile phone-based ads.

View presentation slides
 

Going beyond location-based mobile advertising

Associate Professor Archan Misra Deputy Director of SMU’s Living Analytics Research Centre School of Information Systems, SMU

Mobile devices are already prevalent in consumer paths of purchase, from comparing prices and browsing products online to paying for goods and services at merchant points-of-sale. However, deep data analytics can leverage new data sources like mobile devices to capture insights in the physical world.

Associate Professor Misra focused his presentation on mobile analytics as a way to probe and understand human behaviour. He noted that analytics now go beyond location-based mobile marketing. The use of social media data, coupled with the use of new data sources from mobile devices, produces unique insights into human behaviour in a range of urban settings. Mobile analytics use real-time mobile sensing from consumers like group detection, queuing detection and in-store shopper behaviour to infer consumer preferences, interests and behaviour.

View presentation slides

Increased privacy leads to online social network users’ openness in sharing information

Assistant Professor Phan Tuan School of Computing, National University of Singapore

With the influx of social media, privacy concerns have somewhat taken a backseat in recent years. Nowadays, people are increasingly concerned as to how ‘public’ they truly want to be to the outside world. Thus, it is becoming an exploding concern with online social network users.

Assistant Professor Phan Tuan shared new research on how online social network users behave. He presented his research on public and private information disclosures that occurred before and after the policy change in the Facebook social networking platform in 2009, resulting in more fine-grained privacy controls. His findings show that due to improved privacy controls, online social network users who are highly sensitive to privacy became more open in sharing information. Specifically, online social network users increased their use of public posts and decreased their use of private messages.

Strategic considerations of platform providers to screen long tail sellers

Marketing Professor Shantanu Dutta Marshall School of Business, University of Southern California

Platform providers such as Google, Amazon and Tmall rely on profit and service quality screening mechanisms to filter out sellers, including niche sellers known as ‘long tail sellers'. Platform providers claim that screening helps to filter out the lowest quality service providers and encourage the surviving sellers to improve their services. Many long tail sellers claim that the screening mechanisms are costly and preclude them from participating on the platforms and reduce competition. Further, third-party reports suggest that these screening mechanisms may not enhance service provision by the sellers.

Professor Shantanu Dutta presented his findings in this area. (1) The profit screening mechanism does not always remove the long tail sellers with poor service. (2) The platform may be better off by directly filtering out a long tail seller with better service. (3) The platform screening mechanism does not always increase the average service quality. He suggested that long tail sellers with lower service quality and product differentiation might be more valuable to the platform. It is because those sellers may provide stronger incentives for mass-market sellers to maintain their provision of high quality services. Long tail sellers also may influence the platforms to increase or decrease their service efficiency.

Forward-looking view on analytics

Adjunct Professor Prabir Sen School of Information Systems, SMU Chief Management Scientist at Accenture

Economists question whether analytics can deliver the kind of wide-ranging, profound impact that the introduction of the automobile or the semiconductor chip had, and point to data showing slow productivity growth as evidence.

Adjunct Professor was another keynote speaker at the workshop. He gave a visionary presentation on how analytics can solve some pressing business, consumers and social insight problems.

He said that achieving the full potential of promising analytics for data-driven decisions, at individual and societal scale, while addressing their challenges and risks will require effective leadership. However, the potential is vast. Today, we see many rapidly evolving, potentially transformative techniques on the horizons of the physical, biological, material, social and behavioural sciences and other fields. A confluence of advances in computational speed, machine learning and natural user interfaces – voice, gestures, etc – is making it possible to automate many business activities that have long been regarded as impossible or impractical for machines to perform. This opens up possibilities for sweeping change in how businesses learn, organise and network with their customers. Sophisticated analytics techniques are useful to enable the design of well-coordinated decomposable systems in which humans and machines can interact to ‘learn’ new insights, modify their behaviour and adjust their own algorithms based on analyses of the data. This will enable them to ‘see’ relationships or links that a human might overlook.

Research Snippets

		Metaheuristics Applications in Business and Consumer Analytics Stephen Voss (University of Hamburg) View presentation slides Problems related to prediction models in business and consumer analytics have become increasingly important in recent years. The research demonstrates how metaheuristics computing like support vector machines can be applied in business and consumer analytics contexts with methods from Marketing and Operations Research. The use of classification analysis has important applications in web mining, corporate planning and other areas. Using computational intelligence approaches, metaheuristics can provide meaningful results and solutions.
		Impact of E-Commerce on Concentration of Sales in the Apparel Industry Alejandro Zentner (University of Texas, Dallas) The research focuses on how e-commerce affects the concentration of sales based on the long tail hypothesis. Focusing on the apparel industry, Associate Professor Alejandro Zentner showed that there are remarkable differences between the products that are popular in the online versus offline channels. As consumers move from offline to online markets, sales concentrate around the products that are popular online and away from products that are popular offline. His research devises a metric that seeks to gauge how the overall concentration of sales increases or decreases that accounts for differences in the location and concentration of online and offline sales. Based on that metric, he reported superstar effects for one well-known retailer and no change in concentration for another retailer as consumers moved online.
		The Demand Effects of Product Recommendation Networks: An Empirical Analysis of Network Diversity and Stability Goh Khim Yong (National University of Singapore) This research attempts to answer questions related to product recommendation networks in e-commerce. Is the demand of a product influenced by both the incoming network and outgoing network? How is the demand of a product influenced by its product recommendation network? Are network diversity and stability important? And do such effects differ between co-viewing and co-purchasing recommendation networks? Using observational data from Tmall.com, Assistant Professor Goh Khim Yong’s findings show that the market share of a product is influenced by both the incoming and outgoing networks. Also, the co-purchasing recommendation network exhibits a stronger influence than the co-viewing recommendation network relative to the impacts on market share.

How to Maximise the Value of Adwords
Zhang Defu (Xiamen University)

View presentation slides
Search engine companies such as Google, Yahoo and Bing have created the multi-billion dollar adwords market. This has revolutionised advertising, especially for smaller companies. Bidding for keywords is commonplace for advertisers to maximise their total revenue. However, current algorithms do not seem to help maximise the advertisers’ return on investment. Therefore, the proposed new algorithm – online bipartite matching using matching theory – is a new and more efficient alternative.
 

Dynamic Queue Management for Hospital Emergency Room Services
Tan Kar Way (SMU)

View presentation slides
Addressing the issue of crowding in a hospital emergency room (ER), Tan Kar Way’s research introduces intelligent dynamic patient prioritisation strategies to manage demand concurrently with dynamic resource adjustment policies to affect physician services supply. With the effective use of both historical and real-time information for decision support, the integrated dynamic queue management framework allows decision-makers to select both the demand-side and supply-side strategies to suit the needs of their ER.

Best Research Presentations

Best Doctoral Research Pricing New Products with Pay-What-You-Want Essi POYRY (Aalto University) (above) Petri PARVINEN (Aalto University) Maurits KAPSTEIN (Aalto University)		Best Faculty Research Marginal Deterrence in the Enforcement of Law: Evidence from Distributed Denial of Service Attack HUI Kai-Lung (Hong Kong University of Science & Technology) (above) KIM Seung Hyun (National University of Singapore) WANG Qiuhong (Huazhong University of Science & Technology)

BCSI 2013 Leadership

Co-chairs representing the various disciplines for the workshop include:

• Robert KAUFFMAN, Associate Dean (Research), Deputy Director (Living Analytics Research Centre), and Professor of Information Systems (IS) and Management at SIS, SMU [Email: rkauffman [at] smu.edu.sg] (Event coordinator)

• Shantanu DUTTA, Dave and Jeanne Tappan Chair in Marketing, and Professor of Marketing, Marshall School of Business, University of Southern California [Email: sdutta [at] marshall.usc.edu] (Marketing)

• Pulak GHOSH, Professor of Quantitative Methods and IS, Indian Institute of Management, Bangalore; and LARC, SMU, [Email: pulak.ghosh [at] iimb.ernet.in] (Statistics)

• GUO Zhiling, Associate Professor of IS, SIS, SMU [Email: zhilingguo [at] smu.edu.sg] (Information Systems)

• LAU Hoong Chuin, Associate Professor of IS, SIS, SMU [Email: /Users/ashachoolani/Documents/Asha%20Choolani/ORe/Newsletter/02-May%202013/hclau [at] smu.edu.sg">hclau [at] smu.edu.sg] (Computer Science)

• YANG Yinping, Programme Manager, Independent Investigator and Scientist, Department of Computing Science, Institute of High Performance Computing, Agency for Science, Technology and Research (A*STAR)

• Alejandro ZENTNER, Associate Professor of Finance and Managerial Economics, Naveen Jindal School of Management, University Texas at Dallas [Email: azentner [at] utdallas.edu] (Economics)