Big data. Humanised.
How it can help vulnerable people get a handle on their health and their lives
What is big data anyway? There are many definitions. The simplest one I came across says that big data stands for information that goes beyond what could be squeezed into a large spreadsheet. Big data cannot be filled into a spreadsheet simply because it is no longer numbers. It includes words, visuals, pictures, videos and more. Add to that the enormous amount of information we share on social media platforms, Google searches, blog posts, online reviews, personal photographs, videos etc and you have the big mass that is big data, I was told.
At a recent seminar hosted by IIM Calcutta Alumni Association Mumbai Chapter, Professor Ram gave us a new definition. She should know; she is Anheuser-busch Endowed Professor of MIS, and Entrepreneurship & Innovation in the Eller College of Management at the University of Arizona. In her scheme of things, big data needs to be seen through just two simple lenses. One, big data has led to ‘datafication’ of what till now that was not in the realm of data. Take for example the smart watch you are wearing. By merely wearing it you are able to convert a lot of signals from your body that has always existed into data. You can count the steps you took yesterday, measure your pulse rate, and even take real time measurement of your blood pressure.
Two, big data is so defined because it comes with “time and geo stamp”. We are collecting a lot of data that is not just data but it comes with a clear marker about when it was collected and where it was collected. Take real time traffic measurement that is happening on Google maps. The data is getting collected real time with clear geographical tags (Google has some catching up to do on predicting travel times in Indian cities like Mumbai; for that it probably needs Big Big Data that goes into the realm of divinity).
The question is, how do you use all this big data. Professor Ram shared two interesting examples with us.
The university was approached by a leading Dallas hospital to see if the MIS department could predict the admissions that can be expected in the emergency room. A well-manned emergency room can save lives. But it is expensive to have additional medical or para-medical staff on standby. The data analysis took several steps. First was the simple analysis of past three years' emergency room records. To their surprise the researchers from the university found that the most critical cases were not from gun shootings or traffic accidents but from domestic accidents, cardiac and asthma cases. They narrowed down to study asthma since that department was keen on finding out more. Their search for predictive links led them to many places including Google searches, weather data, pollution data (particulate matter in the atmosphere) and even Twitter postings relating to asthma attacks.
Researchers collected tweets posted between October 2013 and June 2014 and narrowed down to 3,810 tweets that mentioned asthma attacks. They could link this data with incidence of asthmarelated emergency department visits. By mining the tweets and linking it to emergency room visits and atmosphere data, the researchers could build a robust model. As they say in the article published in IEEE Journal of Biomedical and Health Informatics, “Rapid progress has been made in gathering non-traditional digital information to perform disease surveillance. We introduce a novel method of using multiple data sources for predicting the number of asthma-related emergency department visits in a specific area. Twitter data, Google search interests and environmental sensor data were collected for this purpose... Our model can predict the number of asthma ED visits based on near-real-time environmental and social media data with approximately 70 per cent precision”.
In yet another case the researchers were posed the question of addressing student drop-outs in the under-graduate programme at the University. In the US only 60 per cent of the students graduate within six years. Current approaches such as student grades and demographic information was of limited value, especially since a student often decided to drop out within the first 12 weeks of starting at a University. How does one predict who would drop out and who needed help? The academic research had shown that two key factors predicted if student would continue or drop out -- the ability to make social connections and regularity of activity. How does one measure these two in a non-invasive manner?
Professor Ram’s team had at its disposal the data from student identity cards (smart cards) which were used for entry into the mess halls, library, dorms, class room buildings etc. This data was being collected live, real time by the university. By mapping card transactions that occur very near in time and at the same location, researchers could make inferences about a student’s implicit friends group and social networks. They could also build a model around the regularity of the students activities, in an anonymous fashion. According to Ram the model that was build was able to predict at the end of the first 12 weeks the potential for a student to drop out to the extent of 85-90 per cent accuracy. By providing selective help, the University was able to hit a retention rate of 86.5 per cent, the highest in its history. While questions of privacy remain, here is a case where big data was used to help a very vulnerable cohort manage their lives better.
As you would have learnt by now, big data is not just about Big Brother watching you. When used sensibly, it can save lives and help build better future citizens. One human bit at a time.