Student data science competition winners visit Quicken Loans headquarters in Detroit

By | Educational, General Interest, MDSTPosts, News

Earlier this year, three Data Science Team (MDST) members — winners of the Quicken Loan (QL) Lending Strategies Prediction Challenge — traveled to Detroit to visit QL headquarters, accept their prizes, and present their findings to the company’s Data Science team.

Back row left to right: Reddy Rachamallu, Alexandr, Alex, Mark Nuppnau, Brian Ball
Front row left to right: Jingshu Chen, Patrick, Alex’s wife Kenzie, Yvette Tian, Mike Tan, and Catherine Tu.

 

Alexander Zaitzeff, a graduate student in the Applied and Interdisciplinary Mathematics program won first place; Alexandr Kalinin, a Bioinformatics graduate student earned second; and Patrick Belancourt, a graduate student in Climate and Space Sciences and Engineering took third.

The goal of the competition was to create a model that would predict whether potential clients would end up getting a mortgage based on the loan product originally offered to them. In order to create this model, each participant was given access to proprietary de-identified financial data from recent QL clients. The accuracy of their models was then evaluated on one month of client data.

Alexander Zaitzeff

“Every time I participate in a competition I try out a new technique,” Zaitzeff said. “MDST puts me in competitions with other U-M students who I can team up with and learn from.”

“This was a very valuable competition because it gives people experience working with real datasets, on actual problems that companies work on day to day,” said Jonathan Stroud, organizational chair of MDST.

Brian Ball, a data scientist at QL and U-M alum, said the input from MDST students gained through the competition helped confirm the company’s hope that “our system is predictable from a mathematical standpoint.”

“In that regard, we can use the results produced and the methods used to drive good decisions to most benefit our clients,” he added. “We view this as a total success as it was our hypothesis — and underlying hope — from the beginning.”

About 20 people from QL’s Data Science team gathered to hear how the MDST winners developed their models, as well as vice presidents of the Business Intelligence unit.

The winning entry was an “ensemble model,” in which several models are synthesized into one predictive framework.

Finding that so many different kinds of models performed similarly was a confirmation that “the data tells the story,” Ball said.

“Allowing for each technique to contribute more strongly to the final score in areas where the model type performs well (referred to as “blending” or “stacking”) is an especially strong method and one we should consider moving forward,” he said.

The competition began in September and ran until the end of the Fall semester. Over 70 students competed in this challenge, including both graduates and undergraduates from several schools and departments across the University.

MDST typically runs two or three competitions each year — the current competition involves predicting the value of NFL free agents, and is being conducted in partnership with the Baltimore Ravens. For more information, please visit MDST’s webpage: midas.umich.edu/mdst

MDST Competes in the Midwest Undergraduate Data Analytics Competition

By | MDSTPosts, MDSTProjects

Authors: Weifeng Hu and Divyansh Saini

The Competition

We represented the Michigan Data Science Team (MDST) in the Midwest Undergraduate Data Analytics Competition hosted by MinneAnalytics in Minnesota.

We were provided insurance claims data for people diagnosed with Type-II diabetes. At the novice level, the goal was to use this data to find meaningful patterns in age group, gender and geolocation for patients with Type II Diabetes. This information would then be further used to generate cost-saving mechanisms.

 

Experience

The first step to work on this challenge was to understand the dataset. We had medical claims for patients who have Type II Diabetes from health care facilities, pharmacies and laboratories. The files in the dataset are described below following:

medical_training.csv: claims when patients visited a hospital and related data

confinement_training.csv: claims for when patients were confined to the hospital

rx_training.csv: pharmacies claims of drugs prescribed to the patients

labresult_training.csv: claims for laboratory experiments

member_info.csv: file contains information about gender, age, location of patients

Through our meetings with MDST members, we realized that most teams would be approaching this problem with the purpose of predicting the diagnosis or finding patterns in the comorbidity.

To stand out, we decided to focus on finding the sub-populations who have the highest cost after day 0 being of diagnosed with type-II diabetes.

 

Methodology

 We performed K-means clustering on claims in medical_training.csv to find generalized patterns among different groups.

– Data Preprocessing:

The entries in medical_training.csv contain medical claims for each hospital visit for a patient. A patient will have multiple entries if he/she visited the hospital more than once. In each entry, there were at most five ICDM-9* diagnosis codes for each visit. It contains five characters that are mostly numbers, e.g. “12345”. This code can be grouped into 19 categories of diseases and each category contains a range of codes.

The first preprocess step we performed was to “shrink” the number of rows. We used patients ID to group each diagnosis to each patient. After that we constructed a 0/1 indicator feature vectors of length 19. Each column represented a type of disease and the value on that column would 1 if the patient was diagnosed with that disease. Each row in the feature vector contains the information of a particular patient’s diagnosis.

– Hyperparameter Tuning

After making the data into feature vectors, we can know perform k-means clustering. . k-means clustering is partitioning observations into a finite number(k) of clusters or groups in which each observation belongs to a cluster with the nearest mean. However, we still need to decide the number of clusters k. We want to find the number of clusters that result in small intracluster distance but do not overfit the data. To do this we used the “elbow heuristic”, which states that if on plotting the cost of k-means(the sum of the intracluster distance) with respect to k, we should choose a k value that has a significant drop before the that point and no significant drop after that point. As in the graph above, where the x-axis is the value of k and the y-axis is the cost of [1]k-means clustering, we can see that k=19 is a good choice.

[1]

 

 

Result

From the data analysis we performed, we realized that the average cost of each visit for a patient remained the same regardless of when the hospital visit occurred, but the amounts of visits increased significantly after the day that they were diagnosed. This resulted in a significant increase in the total cost, as can be seen in the graph below.

 

Another interesting finding was that although the number of patients diagnosed with Diabetes-II increased with age, the average cost was highest for ages 20-25 and 45-50, as shown in the graph below. This trend was common among different clusters. From this data, we were able to conclude that the people who were diagnosed in those two age groups had similar diseases. This suggests that we should focus on the age group who are diagnosed with diabetes at this younger ages will have a significantly higher cumulative costs as they live their life.

 

What we learned

From this competition, we learned that it was important to style our slides for the presentation. The judges commented us as “one of the most technical solid presentation”. However, we made some mistake on the slides and some axes on the graph are not clearly labeled. As a result, we did not make to the finalis. Nevertheless, it was really encouraging to know that the judges were impressed by our analytical skills.

But beyond that, at the competition itself we were impressed by the various interesting ways that the upper level teams used to predict the highest at-risk patients. One of them used models similar to what credit-card companies use to predict credit ratings for its clients. This was definitely out of the box and I was surprised to see that it actually worked.

And finally, we realized that it is important to persist. We encountered a significant delay before we received the real data(in fact, we did not have it until 5 days before the competition). It was challenging to try to come up with good analysis in that short period of time. However, with the help of our faculty mentor Sean, we were able to find meaningful patterns in the dataset. We wish we can have more time so that we can explore trends in other dataset.

 

HPC training workshops begin Tuesday, Feb. 13

By | Educational, Events, General Interest, Happenings, HPC, News

series of training workshops in high performance computing will be held Feb. 12 through March 6, 2018, presented by CSCAR in conjunction with Advanced Research Computing – Technology Services (ARC-TS).

Introduction to the Linux command Line
This course will familiarize the student with the basics of accessing and interacting with Linux computers using the GNU/Linux operating system’s Bash shell, also known as the “command line.”
Location: East Hall, Room B254, 530 Church St.
Dates: (Please sign up for only one)
• Tuesday, Feb. 13, 1 – 4 p.m. (full descriptionregistration)
• Friday, Feb. 16, 9 a.m. – noon (full description | registration)

Introduction to the Flux cluster and batch computing
This workshop will provide a brief overview of the components of the Flux cluster, including the resource manager and scheduler, and will offer students hands-on experience.
Location: East Hall, Room B254, 530 Church St.
Dates: (Please sign up for only one)
• Monday, Feb. 19, 1 – 4 p.m. (full description | registration)
• Tuesday, March 6, 1 – 4 p.m. (full description | registration)

Advanced batch computing on the Flux cluster
This course will cover advanced areas of cluster computing on the Flux cluster, including common parallel programming models, dependent and array scheduling, and a brief introduction to scientific computing with Python, among other topics.
Location: East Hall, Room B250, 530 Church St.
Dates: (Please sign up for only one)
• Wednesday, Feb. 21, 1 – 5 p.m. (full description | registration)
• Friday, Feb. 23, 1 – 5 p.m. (full description | registration)

Hadoop and Spark workshop
Learn how to process large amounts (up to terabytes) of data using SQL and/or simple programming models available in Python, R, Scala, and Java.
Location: East Hall, Room B250, 530 Church St.
Dates: (Please sign up for only one)
• Thursday, Feb. 22, 1 – 5 p.m. (full description | registration)

Peers Health and U-M begin research partnership using disability and workers’ comp healthcare data

By | General Interest, Happenings, News, Research

Peers Health and the University of Michigan are starting a two-year research project that will apply advanced learning technologies to a proprietary global database of millions of de-identified disability and workers’ compensation cases. The goals of the project include developing a prescriptive modeling framework to facilitate development of optimal return-to-work plans for injured or ill patients.

Public policy experts have begun to connect patients’ ability to perform their productive endeavors, such as their job, to their state of general health and well-being. The findings from this project, by helping define when someone objectively has returned to health, could inform decision-making in virtually every healthcare episode.

The principal investigators in the project, Dr. Brian Denton and Dr. Jenna Wiens, are both renowned experts in medical machine learning. Dr. Denton, a professor of Industrial and Operations Engineering and Urology, and Dr. Wiens, an assistant professor of Computer Science and Engineering, are both affiliated with the Michigan Institute of Data Science (MIDAS) at U-M.

Peers Health recently announced an expanded partnership with ODG, an MCG company and part of the Hearst Health Network, to aggressively acquire new data to enhance ODG functionality and to fuel this research. Jon Seymour, MD, CEO of Peers, said, “This is a new phase in medical publishing where raw data collection is the editorial function and cutting-edge machine learning is the technology factor. We turned to the University of Michigan due to its impressive data science programs spanning multiple departments, as well as the specific experience of Dr. Denton and Dr. Wiens in medical applications. We’re confident this initiative will attract many new data contributors along the way.”

“The collaboration with Peers Health is exciting because it provides data that can help build a model that will reduce the time — from both a safety and productivity perspective — for people to return to work following sickness or injury,” Denton said. “Streaming data in from existing patients will allow our model to adapt and improve over time.”

Wiens added: “These data contain a particularly interesting training label: days away from work. We hypothesize that this will be a strong signal for the type, timing, and effectiveness of the treatments and therapies.”

The U-M partnership with Peers was established by MIDAS and the university’s Business Engagement Center (BEC).

“This partnership illustrates the power of combining data from the healthcare industry with the data science expertise of U-M faculty,” said Dr. Alfred Hero, professor of Engineering and co-director of MIDAS.

“It is energizing for the BEC to be part of these innovative collaborative relationships that create real impact in the world,” added BEC Director Amy Klinke.

 

U-M fosters thriving artificial intelligence and machine learning research

By | General Interest, HPC, News, Research

Research using machine learning and artificial intelligence — tools that allow computers to learn about and predict outcomes from massive datasets — has been booming at the University of Michigan. The potential societal benefits being explored on campus are numerous, from on-demand transportation systems to self-driving vehicles to individualized medical treatments to improved battery capabilities.

The ability of computers and machines generally to learn from their environments is having transformative effects on a host of industries — including finance, healthcare, manufacturing, and transportation — and could have an economic impact of $15 trillion globally according to one estimate.

But as these methods become more accurate and refined, and as the datasets needed become bigger and bigger, keeping up with the latest developments and identifying and securing the necessary resources — whether that means computing power, data storage services, or software development — can be complicated and time-consuming. And that’s not to mention complying with privacy regulations when medical data is involved.

“Machine learning tools have gotten a lot better in the last 10 years,” said Matthew Johnson-Roberson, Assistant Professor of Engineering in the Department of Naval Architecture & Marine Engineering and the Department of Electrical Engineering and Computer Science. “The field is changing now at such a rapid pace compared to what it used to be. It takes a lot of time and energy to stay current.”

Diagram representing the knowledge graph of an artificial intelligence system, courtesy of Jason Mars, assistant professor, Electrical Engineering and Computer Science, U-M

Johnson-Roberson’s research is focused on getting computers and robots to better recognize and adapt to the world, whether in driverless cars or deep-sea mapping robots.

“The goal in general is to enable robots to operate in more challenging environments with high levels of reliability,” he said.

Johnson-Roberson said that U-M has many of the crucial ingredients for success in this area — a deep pool of talented researchers across many disciplines ready to collaborate, flexible and personalized support, and the availability of computing resources that can handle storing the large datasets and heavy computing load necessary for machine learning.

“The people is one of the reasons I came here,” he said. “There’s a broad and diverse set of talented researchers across the university, and I can interface with lots of other domains, whether it’s the environment, health care, transportation or energy.”

“Access to high powered computing is critical for the computing-intensive tasks, and being able to leverage that is important,” he continued. “One of the challenges is the data. A major driver in machine learning is data, and as the datasets get more and more voluminous, so does the storage needs.”

Yuekai Sun, an assistant professor in the Statistics Department, develops algorithms and other computational tools to help researchers analyze large datasets, for example, in natural language processing. He agreed that being able to work with scientists from many different disciplines is crucial to his research.

“I certainly find the size of Michigan and the inherent diversity that comes with it very stimulating,” he said. “Having people around who are actually working in these application areas helps guide the direction and the questions that you ask.”

Sun is also working on analyzing the potential discriminatory effects of algorithms used in decisions like whether to give someone a loan or to grant prisoners parole.

“If you use machine learning, how do you hold an algorithm or the people who apply it accountable? What does it mean for an algorithm to be fair?” he said. “Can you check whether this notion of non-discrimination is satisfied?”

Jason Mars, an assistant professor in the Electrical Engineering and Computer Science department and co-founder of a successful spinoff called Clinc, is applying artificial intelligence to driverless car technology and a mobile banking app that has been adopted by several large financial institutions. The app, called Finie, provides a much more conversational interface between users and their financial information than other apps in the field.

“There is going to be an expansion of the number of problems solved and number of contributions that are AI-based,” Mars said. He predicted that more researchers at U-M will begin exploring AI and ML as they understand the potential.

“It’s going to require having the right partner, the right experts, the right infrastructure, and the best practices of how to use them,” he said.

He added that U-M does a “phenomenal job” in supporting researchers conducting AI and ML research.

“The level of support and service is awesome here,” he said. “Not to mention that the infrastructure is state of the art. We stay relevant to the best techniques and practices out there.”

Advanced Research Computing at U-M, in part through resources from the university-wide Data Science Initiative, provides computing infrastructure, consulting expertise, and support for interdisciplinary research projects to help scientists conducting artificial intelligence and machine learning research.

For example, Consulting for Statistics, Computing and Analytics Research, an ARC unit, has several consultants on staff with expertise in machine learning and predictive analysis with large, complex, and heterogeneous data. CSCAR recently expanded capacity to support very large-scale machine learning using tools such as Google’s TensorFlow.

CSCAR consultants are available by appointment or on a drop-in basis, free of charge. See cscar.research.umich.edu or email cscar@umich.edu for more information.

CSCAR also provides workshops on topics in machine learning and other areas of data science, including sessions on Machine Learning in Python, and an upcoming workshop in March titled “Machine Learning, Concepts and Applications.”

The computing resources available to machine learning and artificial intelligence researchers are significant and diverse. Along with the campus-wide high performance computing cluster known as Flux, the recently announced Big Data cluster Cavium ThunderX will give researchers a powerful new platform for hosting artificial intelligence and machine learning work. Both clusters are provided by Advanced Research Computing – Technology Services (ARC-TS).

All allocations on ARC-TS clusters include access to software packages that support AI/ML research, including TensorFlow, Torch, and Spark ML, among others.

ARC-TS also operates the Yottabyte Research Cloud (YBRC), a customizable computing platform that recently gained the capacity to host and analyze data governed by the HIPAA federal privacy law.

Also, the Michigan Institute for Data Science (MIDAS) (also a unit of ARC) has supported several AI/ML projects through its Challenge Initiative program, which has awarded more than $10 million in research support since 2015.

For example, the Analytics for Learners as People project is using sensor-based machine learning tools to translate data on academic performance, social media, and survey data into attributes that will form student profiles. Those profiles will help link academic performance and mental health with the personal attributes of students, including values, beliefs, interests, behaviors, background, and emotional state.

Another example is the Reinventing Public Urban Transportation and Mobility project, which is using predictive models based on machine learning to develop on-demand, multi-modal transportation systems for urban areas.

In addition, MIDAS supports student groups involved in this type of research such as the Michigan Student Artificial Intelligence Lab (MSAIL) and the Michigan Data Science Team (MDST).

(A version of this piece appeared in the University Record.)

Video available from MIDAS Research Forum

By | General Interest, Happenings, News, Research

Video is now available from the MIDAS Research Forum held Dec. 1 in the Michigan League at http://myumi.ch/6vA3V

The forum featured U-M students and faculty showcasing their data science research; a workshop on how to work with industry; presentations from student groups; and a summary of the data science consulting and infrastructure services available to the U-M research community.

NOTE: The keynote presentation from Christopher Rozell of the Georgia Institute of Technology will be available in the near future.

Yottabyte Research Cloud able to accept HIPAA-aligned data

By | General Interest, HPC, News

Advanced Research Computing – Technology Services (ARC-TS) is pleased to announce that the Yottabyte Research Cloud (YBRC) computing platform is now HIPAA-compliant. This means that YBRC and its associated services can accept restricted data, enabling secure data analysis on Windows and Linux virtual desktops as well as secure hosting of databases and data ingestion.

The new capability ensures the security of restricted data through the creation of firewalled network enclaves, allowing HIPAA-aligned data to be analyzed safely and securely in YBRC’s flexible, robust and scalable environment.   Within each network enclave, researchers have access to Windows and Linux virtual desktops that can contain any software required for their analysis pipeline.

This capability also extends to our database and ingestion services:

  • Structured databases:  MySQL/MariaDB, and PostgreSQL.
  • Unstructured databases: Cassandra, MongoDB, InfluxDB, Grafana, and ElasticSearch.
  • Data ingestion: Redis, Kafka, RabbitMQ.
  • Data processing: Apache Flink, Apache Storm, Node.js and Apache NiFi.
  • Other data services are available upon request.

YBRC is supported by U-M’s Data Science Initiative launched in 2015. YBRC was created through a partnership between Yottabyte and ARC-TS announced last fall.

These tools are offered to all researchers at the University of Michigan free of charge, provided that certain usage restrictions are not exceeded. Large-scale users who outgrow the no-cost allotment may purchase additional YBRC resources. All interested parties should contact hpc-support@umich.edu.

ARC Director Sharon Broude Geva re-elected vice-chair of Coalition for Academic Scientific Computing

By | General Interest, News

Sharon Broude Geva, the Director of Advanced Research Computing at the University of Michigan, has been re-elected vice-chair of the Coalition for Academic Scientific Computation (CASC).

Founded in 1989, CASC advocates for the use of advanced computing technology to accelerate scientific discovery for national competitiveness, global security, and economic success. The organization’s members represent 84 institutions of higher education and national labs.

The vice-chair position is one of four elected CASC executive officers. The officers work closely as a team with the director of CASC. The vice-chair also leads CASC meeting program committees, is responsible for recruitment of new members, substitutes for the chair in his or her absences, and assists with moderating CASC meetings.

Geva served as CASC secretary in 2015 and 2016, and one term as vice-chair in 2017. Her next term as vice-chair is effective for the 2018 calendar year.

The other executive officers for 2017 are are Rajendra Bose, Chair, Columbia University; Neil Bright, Secretary, Georgia Institute of Technology; and Andrew Sherman, Treasurer, Yale University. Curt Hillegas of Princeton University is immediate past chair.

The 2018 CASC brochure is available online.

U-M wraps up successful SC17 conference

By | General Interest, Happenings, HPC, News

Several University of Michigan researchers and professional IT staff attended the Supercomputing 17 (SC17) conference in Denver from Nov. 12-17, participating in a number of different ways, including demonstrations, presentations and tutorials.

U-M participation included:

  • Matt McLean, a Big Data systems administrator with ARC-TS, served as a panelist at a session titled “The ARM Software Ecosystem: Are We There Yet?” (Slides)
  • Jeff Sica, a research database administrator with ARC-TS, helped lead a Birds of a Feather session titled “Containers in HPC.” (Slides)
  • Quentin Stout (EECS) and Christiane Jablonowski (CLASP) taught the “Parallel Computing 101” tutorial.
  • Shawn McKee, U-M Department of Physics, and OSiRIS Principal Investigator, demonstrated Object Storage and Caching for Science (network topology diagrams)
  • Eric Boyd, Director of Research Networks, presented on Research Networking at the University of Michigan at the U-M exhibit booth.
  • Simon Adorf, Ph.D. Candidate, Chemical Engineering Department, U-M, presented on Simple Data and Workflow Management with Signac and GPU-Accelerated Predictive Material Design at the U-M exhibit booth.
  • ARC sponsored a networking and career networking reception put on by Women in HPC. ARC Director Sharon Broude Geva spoke at the event.
  • Amy Liebowitz, a network architect at ITS, worked on SCINet, a high-capacity network created every year for the conference. Liebowitz was on the routing team, which is responsible for installing, configuring and supporting the high performance conference network. The Routing Team also coordinated external connectivity with commodity Internet and R&E WAN service providers.

New Data Science Course – Winter 2018

By | Educational, News

Computational Data Science
(EECS 598 / BIOINF 505)

A new graduate course that provides an in-depth introduction to computational methods in data science for identifying, fitting, extracting and making sense of patterns in large data sets is now enrolling students for Winter 2018.

Lectures will typically begin with an introduction of a core data science method, followed by the student programming the method computationally with a computer assisting the student by certifying when the program is correct, interleaved with ‘just-in-time’ theory that will expose the student to the mathematics that underpin the methodology. Once the method has been correctly implemented, the students will be given a real world example or ‘success story’ to work with that illustrates when the algorithm ‘works’ as expected, followed by an instructor guided computational exploration of the various subtleties of the algorithm and its weakness.

A full course description, prerequisites and schedule are available.

Please share this announcement with students who might be interested.