The practicum is an opportunity for students to directly experience the work of a data scientist or data analytics professional. It consists of project-based learning on a significant and contributory business objective in conjunction with practicing professionals in one of many appropriate industries. Students may complete the practicum by working on a project at the student’s place of employment or by obtaining an internship in the area of data science.
Sample of Past Practicum Projects
Sponsor: Saint Joseph High School, South Bend, IN
Problem: Develop a strategic enrollment management plan in order to have a data-informed process that helps address challenges such as adapting recruitment and retention strategies to rapid market changes, responding to changing demographics, and supporting diversity.
Methods: R/RStudio, Data Preprocessing & Visualization, Databases
Outcome: An understanding of the characteristics of a successful student at Saint Joseph High School, including factors that contribute to the likelihood of graduating in 4 years.
Sponsor: Teradata, St. Louis, MO
Problem: Build a repeatable demonstration of the Teradata Unified Data Architecture. The demo should show the capabilities of the data warehouse platform and the data discovery/analytic platform provided by Teradata, as well as the connection between the two platforms.
Methods: SQL, Databases, Data Scraping from the Web, Data Cleaning, Predictive and Classification Modeling (including Random Forests, Least Angle Regression, Naïve Bayes, Principal Component Analysis)
Outcome: A database of NCAA basketball data and a model for predicting the “win/loss” outcome of a game.
Sponsor: Teradata, St. Louis, MO
Problem: Develop a discovery Proof of Concept (POC) to determine if patterns or prediction algorithms existed where the US Air Force could modify their process for maintaining an aircraft.
Methods: Hadoop, Data Warehousing, Text Analytics, Predictive Modeling
Outcome: An interactive and scalable database of US Air Force aircraft information.
Sponsor: Kellstrom Materials, Miami Lakes, FL
Problem: Develop a model to predict how much of a certain part (e.g., blades, nozzles, shrouds) in an aircraft engine will survive repair process.
Methods: R/RStudio, Data Preprocessing & Visualization, Regression (including Lasso Models), Bootstrap, Cross-Validation
Outcome: A predictive model identifying the most and least significant factors related to repair yields for aircraft engine parts.
Sponsor: Delta Dental of California, San Francisco, CA
Problem: Gain a better understanding of customer experience by analyzing electronic interactive voice recognition data resulting from customer inquiry about eligibility, claim and benefit information.
Methods: Python, R/RStudio, SQL, Tableau, Data Preprocessing & Visualization, Natural Language Processing, Text Mining
Outcome: Data visualizations highlighting certain characteristics of the customer experience.
Sponsor: ROIDNA, San Francisco, CA
Problem: Design, implement, test, and operationalize machine learning algorithms for clients using end-to-end data science methodology.
Methods: Python, R/RStudio, SQL, Data Warehousing, Data Preprocessing & Visualization, Machine Learning
Outcome: A Data Science Pipeline for online ad bidding on an airline search engine. Specifically, the pipeline automated data processing (collection, cleaning, visualization), implemented database storage from beginning to end, and applied machine learning techniques for advanced analytics.
Sponsor: Office of Institutional Research, Saint Mary’s College
Problem: Provide a better understanding of the transition patterns of undergraduate students from application, admission, enrollment, and graduation.
Methods: SQL, R/RStudio, Python, Data Preprocessing & Visualization, Predictive Modeling (Logistic Trees, Random Forests)
Outcome: Interactive visualizations mapping the undergraduate pathway from admission to graduation, that highlights changes from intended major to completed major. Additionally, a predictive model for enrollment yield was developed to identify key factors in the likelihood of an admitted student enrolling at the College.