Developing a Machine Learning Model for Detecting Fraudulent Credit Card Transactions (John Hopkins University) – For his capstone project, a student developed a machine learning model using techniques like clustering, association rule mining, and random forests to analyze transaction data and identify abnormal spending patterns that could indicate fraudulent activity. The model was able to detect fraudulent transactions with over 90% accuracy on test data, outperforming the existing rules-based system used by most banks.
Designing an AI Chatbot for Customer Service (University of Michigan) – A group of students designed and implemented an advanced natural language processing system in the form of a chatbot for handling basic customer service inquiries. The chatbot was trained on a large corpus of real customer conversations with company representatives. It used techniques like intent classification and semantic parsing to understand user requests, retrieve relevant information from a knowledge base, and respond in a helpful, conversational manner. User testing showed the chatbot could resolve over 70% of common issue types with an accuracy comparable to a human agent.
Creating an Immersive Virtual Reality Environment for Architecture Design (Georgia Tech) – For her capstone, a student developed a virtual reality application for early-stage architectural design using the Unity game engine. Key features included tools for modeling basic 3D structures and environments, a physics simulation to experience real-time lighting/shadow effects, and multiplayer capabilities for collaborative design sessions. Architects providing feedback noted it was a promising proof-of-concept for how VR could transform and streamline certain phases of building conceptualization and review.
Developing an IoT Crop Monitoring System (University of Washington) – A group of four students designed and built a wireless sensor network system for monitoring soil conditions,temperature, moisture levels and other factors in agricultural fields. Sensors nodes equipped with XBee radios reported readings back to a Raspberry Pi base station at regular intervals. Custom software visualized real-time metrics and generated alerts when readings exceeded pre-set thresholds. Early pilots with local farms found it helped optimize watering and better detect issues before sizable losses occurred.
Building a Database-Driven Photo Sharing Web App (Northeastern University) – For his individual capstone, a student developed a full-stack photo sharing application using modern web technologies. It included features like user accounts/profiles, advanced search, image uploads with resizing/compression, commenting, and likes/favorites. The app was built with Django and PostgreSQL on the back-end and React/Redux on the front-end. Usability testing and feedback from other students found it to be polished and on par with popular sites like Flickr or Imgur.
These represent just a small sample of the diverse and impactful capstone projects undertaken by computer science and engineering students to demonstrate their skills before graduating. By applying their classroom learning to real-world problems, students are able to gain practical experience that significantly strengthens their professional portfolios and readiness for industry roles. The caliber of work produced also highlights the strong hands-on learning component of these degree programs.
Andrew Stroud