Data Science (DSC)

Meet the Team Alvin Yao Model Training & Evaluation Developed performance diagnostics and visualized stage-by-stage model behavior. Authored the next-steps and challenge discussion. Grayson Gong Pipeline Architect Designed the three-stage…

Portfolio risk analysis & daily reporting. This portfolio risk analysis project was completed as a University of Rochester Data Science Capstone for Senior Design Day, with sponsorship from FinchAI. We…

Authors: Caelan Clayton, Veena Laks, Ramneek Nandha, Aabha Pandit Advisor: Ajay Anand, Ph.D. Sponsor: Akkadian, LLC.; Jessica Gleason, Ph.D. and Jon Schlecht, BAS Introduction Pedagogy, the study of effective teaching…

Team Members: Bryce Tyler, Carter Schmitt, Mia Alex, Tianyou Tu, Yaxun Chen Sponsor: MathWorks | Instructors: Ajay Anand, Ph.D. Cantay Caliskan, Ph.D. Abstract: The FOMC, or Federal Open Market Committee, gathers…

Capstone Team Baichuan Duan  ·  Lexiang Yang  ·  Shixin Lin  ·  Zhengyang Zhu University of Rochester × Paychex  ·  April 2026 A Capstone Project · Spring 2026 Listening Between the…

DSCC 383W Team 7: Aakanksha Dutta, Elvis Imamura, Eva Salmone, Sreejato Chatterjee, Duole Hong Sponsors: Zalliant —John Balbian, Emmanuel Mounier, Utsav Advisor: Professor Cantay Caliskan Background: Cattle health monitoring is…

This project showcases an analysis on the research output at the University of Rochester.

This project investigated whether patterns in medication use among older adults could help predict rehospitalization after home healthcare. Using data from over 6,800 patients, we analyzed active medications, diagnoses, and patient characteristics, mapping them to standardized codes like RxNorm and PhecodeX. We applied clustering techniques and machine learning models, including XGBoost and BERT-based text embeddings, to identify potential risk factors. Although some variables—like hyper-polypharmacy, reduced physical function (ADL), and certain medication classes—were associated with increased risk, no clear or consistent clusters emerged as highly predictive. Our best-performing model achieved 96% accuracy and 97% ROC-AUC, reinforcing the value of advanced methods but also underscoring the need for individualized deprescribing strategies in geriatric care.

Team members: Skye Crocker, Colin Hascup, Ahmed Malik, Lam Nguyen, Ruilin ZhangSponsor: Zalliant | Instructors: Ajay Anand, Ph.D. Cantay Caliskan, Ph.D. Drying crops before storage as hay or silage is…

Authors Zhe Chen, Luca Laport, Erjia Meng, Zhefu Qin, Diego Velázquez Sponsor Rel8ted, Ryan Hilimoniuk Instructor Professor Ajay Anand Abstract Modern supply chains face growing risks from geopolitical instability, financial…

In order to help vineyards in the Finger Lakes region with frost prediction, we built a Hidden Markov Model matching algorithm that computes log-likelihood scores between seasons. We implemented seasonal grouping in this approach.

DiscSense is aiming to advance athletes’ throwing skills through the development of a gyroscopic sensor that tracks the end conditions of throws. Throughout our capstone project, we concentrated on building a classification model that will aid athletes in recognizing patterns of successful throws and pinpoint prevalent errors.

Author Sponsor Dr. Zidian Xie Instructor Professor Ajay Anand Professor Cantay Caliskan Abstract In the digital age, Twitter has emerged as ancentral arena for public health issues, particularly those involv-ing…

Our new approach uses Major Code, lessening the total codes from 850 to 23. In Major Classification, we used input layer and embedding as well as 2 hidden layers. For Job Title Classification by Major, we used Random Forest to output based on Majority Voting.

Our hierarchical approach uses Major Code, lessening the total codes from 850 to 23. In Major Classification, we used input layer and embedding as well as 2 hidden layers. For Job Title Classification by Major, we used Random Forest to output based on Majority Voting.

Team Team Member Major Amanda Pignataro B.S. in Data Science Avery Girksy B.S. in Data Science Ryan Hilton B.S. in Data Science Vaarya Srivastava B.S. in Data Science Mentor Prof.…

Our project is to aid in the development of a pickleball analytics platform by improving ball detection and tracking. The baseline model used is a TrackNetV2 (Sun et. al. 2020) model trained on badminton, and the purpose of this project is to adapt the model by using transfer learning techniques to improve its performance in pickleball.

1. Team 2. Mentor  Georgen Institute for Data Science (GIDS) 3. Sponsor Lisa Altman 4. Abstract Due to the continuously increased demand for Data Science degrees, our school will open…

Magnetic Couple Study collected data and information from heterosexual couples who are of mixed HIV-status and recorded their prevention methods, including condom use, viral load, and new method-PrEP. This project focused on using unsupervised learning algorithms to examine the main predictors associated with protection strategies.

ML Based classification model to detect triage level for patients arriving at trauma centre, and thus allocate appropriate resources. This was achieved using patients’ data from URMC (Department of Paediatrics).

Virufy has created machine learning models that analyze coughs in order to provide a COVID-19 diagnosis. Training these models requires an even balance between COVID-positive and COVID-negative data, but they unfortunately have very little positive data. In order to combat this issue, the team hoped to generate synthetic coughs that closely resemble real coughs.

Team Academic Advisor Prof. Ajay Anand and Prof. Cantay Caliskan Project Sponsor Dr. Erika Ramsdale and the URMC Geriatric Oncology Team Introduction Studies in recent years have shown that cancer…

Team Mentor Professor Ajay Anand Dr. Preston Countryman Sponsor Corning Inc. – Data Science & Intelligence (DSI) Team Abstract Corning wants to develop a deep-time-series model to perform accurate customer-level…

Team Yihe Chen Harry Huang Junting Chen Kehan Yu Mentor Cantay Caliskan Abstract Predictive Analytics for Demand Responsive Para- transportation Vision & Goal ● Create a productive schedule for Demand…

This project uses the luminescence of the nighttime sky as a predictive features for economic activity.

This project aims to observe, visualize, and model the trends in which COVID-19 patients at the University of Medical Center were allocated ventilators. Descriptive analyses are performed to investigate the relationships between variables such as but not limited to recovery rate and length of ventilator allocation and gender, race, and age.

This project aims to build a model which detects features such as crosswalks and curb ramps at intersections in the city of Rochester.

Team Qianqian Gu (Project Manager) Wei Wu Chen Yao Hanyang Zhang Mentor Ajay Anand Abstract The Goergen Institute for Data Science (GIDS) masters admission office wants to better understand applicants’…

Team Steven Dai Zachary Mustin Uzoma Ohajekwe Duy Pham Sponsor Vnomics Corporation Matt Mayo Mentor Prof. Ajay Anand Abstract Our task is to predict imminent failures in Diesel Particulate Filters…

Team Xiaoen Ding Jiecheng Gu Sung Beom Park Joseph Smith Mentor Ajay Anand Sponsor Lisa Altman Gretchen Briscoe Abstract The Goergen Institute for Data Science wants to understand the types…

Team Yihan Shao Chuqin Wu Melanie Xue Zihe Zheng Mentor Cantay Çalışkan Abstract The goal of this project is to forecast the pest pressure of Grape Powdery Mildew at a…

This project investigates the associations between geriatric assessment based features and relative dose intensity of chemotherapy. It is at the first few phases of Wilmot Cancer Institute’s Ger Oncology Research team at University of Rochester Medical Center. The team refined the data preprocessing pipeline, built predictive models and employed feature selection on the dataset, providing insightful suggestions for future work in cancer studies.

Successfully built autoencoder models with ML Flow and Keras to predict truck failures given sensor data for a fuel optimization startup called Vnomics. The model is optimized by comprehensive time series feature engineering with TS Fresh to achieve a high recall score of 56% on unseen data.

The City of Rochester wants to understand if physical proximity to a convenience store or liquor store affects the likelihood of different types of part 1 crimes.

The Geriatric Oncology Research Team at URMC wants to better understand chemotherapy tolerability in vulnerable older adults.

The goal of the project was to identify upsell opportunities for Paychex’s 401(k) service products to their existing clients.

Rochester Monroe Anti-Poverty Initiative (RMAPI) launched a new survey to better understand the impact of COVID- 19 on community member’s income and basic needs as well as what community members need to be safe and financially secure. The goal of the project was to analyze the survey and responses to inform United Way which kind of assistance needs to be provided, and what features of living necessities are more important for the respondents.

Identify scenarios where DPF (Diesel Particulate Filter) failure is likely to happen so that the trucking customer can be alerted in advance to avoid costly roadside breakdowns.

The main objective is to identify the maximum water flow tolerance of the Moses-Saunders Dam in order not to exceed the permissible limits of Lake St. Louis.

The project supported the goal of UR Utilities and Energy Management deparment to improve the efficiency of chilled water production through predictive modeling

The goal of the project was to explore public perception on COVID-19 vaccine by analyzing social media platform data (Twitter).

The project aim was: 1) Understand how the degree of mental health issues changed over time and space during COVID-19; 2) Find out what topics are people concerned about, and 3) Infer what group of people are more likely to have mental health issues.

To spur museum membership growth, encourage donations from members, and increase overall museum revenue

The Caldwell-Fay equation (2002) attempts to model what Lake Ontario’s current water level would be if dam construction had never taken place along the St. Lawrence Seaway (i.e. the natural hydraulic state of the lake).

Newly unearthed Lake Ontario data going back to the 1860s has been discovered, and we had the rare opportunity to be the first to digitize and publicly analyze it.

Since this data set predates any dam construction it actually captures the lake’s natural state. Therefore it can be used to verify Caldwell-Fey’s equation which is being used to govern the lake’s inflow and outflow rate on a daily basis.

We were given a patient reported symptoms dataset PRO-CTCAE and applied a variety of clustering methods. The clusters were then statistically tested for associations with a selection of outcomes such as hospitalization. We found significant associations with clusters and outcomes and compared it to linear regression results.

RTS is a regional transportation authority established by New York State and the goal of the project is to find the potential reasons for preventable accidents caused by bus operators. First, descriptive and exploratory analysis is performed on all the data provided and driver-related variables and environmental-related variables. Then, frequent pattern mining is applied and conditional probabilities are calculated for the accident history of operators with high risk of accidents to extract accident patterns.

Wegmans grocery stores experience changes in consumer demand due to weather-related events which may result in item shortages. Our goal was to generate a list of items that are expected to have a huge increase in sales which would allow Wegmans to prepare beforehand. We correlated the change in consumer demand over time with weather warning data and detected anomalous behaviors in item sales.

In this project, we want to apply DSC and machine learning techniques to identify and analyze group communication and interaction patterns from the data collected, e.g. “Who interacts with whom” and “Who attended which breakout sessions”, which can function as an indicator of team performance, group intelligence and meeting efficiency. We can further use the information to increase the productivity of Un-meetings by modifying related elements.