When the conference team was planning for the in-person event, we created many sample agendas for the different topic areas to help attendees with a starting point of which presentations to attend. We had lots of great feedback about how they help attendees to navigate the incredible amount of content that is available at the conference. As you know, the on-going situation with COVID-19 means that SAS Global Forum 2020 became a virtual event and conference attendees didn’t get to view the sample agendas; however, the conference proceedings are still available!
I'll be sharing the sample agendas that the conference team built for different topic areas to help you get started on which proceedings might interest you. The first one is Business Analytics, and there are links for a few papers, including a Forecasting and Econometrics paper, and a few Machine Learning/Predictive Modeling papers. Happy reading!
Business Analytics
Classification of regional characteristics using population composition data and POS data: In this e-Poster, Kazuma Bannai , explains how using population data from Tsukuba, Ibaraki Prefecture, a Huff-model will be created to identify the characteristics of each store in the city. In this city, the area where young people live and the area where elderly people live are relatively separated. Using SAS, the company will analyze whether there is a difference in sales between stores categorized into different categories based on store characteristics and POS data.
Using Analytics to Predict Tax Recovery and Prioritize Audits and Investigations in Canada: Jason Oliver, from Canada Revenue Agency (CRA), leverages the capabilities of SAS® Enterprise Guide® and SAS® Enterprise Miner™ in unearthing predictive patterns of interest with the clear objective of strengthening a feedback loop between tax risk assessment and the corresponding accrual of tax via audit. He examines powerful data learning techniques, as they apply to tax-based analytics, such as neural networks, decision trees, and regression analysis.
Deploying Computer Vision by Combining Deep Learning Action Sets with Open-Source Technology: Jonny McElhinney and Duncan Bain, ScottishPower Energy Retail Ltd; Haidar Altaie, SAS Institute Inc., UK, focuses on doing Automatic Meter Reading (AMR) using customer-submitted photos of their meters. The challenges in this context are two-fold. First, we need to localise the box containing the digits, and then we must classify each digit with the correct label, 0-9. Many approaches are available, but both of these challenges can be addressed by combining the Deep Learning action set in SAS® Visual Data Mining and Machine Learning (VDMML) with DLPy and Keras. Read about how when applying these models in a real-world business context, an additional challenge arises, which is how to deploy and keep track of these models in a consistent way. This paper shows how SAS® and open source tools can be used together to provide a consistent approach both to creating as well as managing and deploying models.
Empower and Inspire-Designing Reports for Mobile Experiences.docx: Khaliah H. Cothran, Ph.D., from SAS Institute Inc., presents tips for how to reduce information density with space saving-features, as well as do’s and don’ts for selecting objects, fonts, and colors that enable complex analytics to be processed quickly. Come learn how to create effective reports that scale from your desktop to your mobile device screen.
Smarter and Faster with SAS® Visual Analytics: Rick Styll, from SAS Institute Inc., writes that SAS® Visual Analytics is the smartest business intelligence tool available. Automated Explanation, the new name for automated analysis, has been rewritten and redesigned to give you smarter and clearer insights, more interactivity, and easier-to-read explanations. In seconds, you can get the analytical story for the business intelligence hidden in your data that would take you hours to do manually. On top of that, you can automatically see suggested visualizations and identify related measures. For more advanced analytical visualizations, like decision trees, you get human-friendly natural language descriptions, drawing out insights that are easy to digest.
Forecasting and Econometrics
Forecasting Hourly Electricity Prices: Joseph Perez from American Electric Power states that forecasting power prices can be more of an art than it is a science. Locational Marginal Prices (LMPs) exhibit high volatility—a phenomenon known as volatility clustering as with many financial time series data. His paper does not discuss the mean reverting properties of power prices, and assumes the reader has a basic understanding of the autoregressive nature of electricity prices. This method forecasts LMPs by preserving past volatility in the form of ratios and assigning them to corresponding load in the forecast. The purpose is to “shape” the forecasted series by utilizing historical profiles while holding to supply and demand fundamentals—high electricity usage tends to be associated with higher energy prices.
Machine Learning (Data mining/Predictive Modeling)
How to Explain Your Black-Box Models in SAS® Viya®: Funda Güneş, Ricky Tharrington, Ralph Abbey, and Xin Hunt, all from SAS Institute Inc., introduces partial dependency (PD) plots, independent conditional expectation (ICE) plots, local interpretable model-agnostic explanations (LIME), and Shapley values, and demonstrates their use in two scenarios: a business-centered modeling task and a health-care modeling task. Also shown are the two different interfaces to these methods in SAS Viya: Model Studio and the SAS Viya programming interface.
Optimizing Supply Chain Robustness through Simulation and Machine Learning: Bahar Biller and Jinxin Yi, SAS Institute Inc., introduces a supply chain simulator that has been built using SAS® Simulation Studio. The key features of the SAS® simulation technology, which enable the development of digital supply chains and the analysis of thousands of scenarios to perform risk-and-return tradeoff, are discussed. The paper concludes with a description of how computational efficiencies can be achieved through an integrated use of SAS Simulation Studio and SAS® Visual Data Mining and Machine Learning.
Human Bias in Machine Learning: How Well Do You Really Know Your Model?: Jim Box, Elena Snavely, and Hiwot Tesfaye, all from SAS Institute, explores the history of bias in models, discusses how to use SAS® Viya® to check for bias, and examines different ways of eliminating bias from our models. Furthermore, they look at how advanced model interpretability available in SAS Viya can help end users to better understand model output. Ultimately, the promise of AI and machine learning is still a reality, but human oversight of the modeling process is vital.
Exploring Online Drug Reviews using Text Analytics, Sentiment Analysis, and Data Mining Models: Thu Dinh, Goutam Chakraborty, and Miriam McGaugh, all from Oklahoma State University, aims to classify the side effect level and effectiveness level of prescribed drugs by using text analytics and predictive models within SAS® Enterprise Miner™. Additionally, the paper explores specific effectiveness and potential side effects of each prescription drug through sentiment analysis and text mining within SAS® Visual Text Analytics.
Using Machine Learning and Demand Sensing to Enhance Short-Term Forecasting for CPGs: Kedar Prabhudesai, Varunraj Valsaraj, Dan Woo, Jinxin Yi and Roger Baldridge, all from SAS Institute Inc., explains that their paper uses machine learning along with traditional time-series forecasting models to generate enhanced weekly and daily forecasts by using historical-demand signal data and point-of-sale data. The model first creates enhanced weekly forecasts, and then breaks down enhanced weekly forecasts into daily forecasts.
