• The Chowculator
Keeshawn Sun
• Classifying Hand Drawn Images“Chowculator” is derived from merging the words “chow”, referring to the colloquial term for food, and “calculator”. The purpose of this project is to create a calorie counting web application (app) that provides two unique features: a points system and rewards shop. A major issue that prevents individuals from adhering to their diet plans is motivation. Why eat a salad when hamburgers and steak are just as available? To help its users keep motivated, the Chowculator will either reward or punish them for abiding by or deviating from their diet plans. If the user abides to their diet plan, they will be rewarded a certain number of points, however, if the user deviated from their plan, they will have some of their points deducted. The points that each user accumulates can be spent in the customizable rewards shop where they can create items that fit their desires.
• Comparative Analysis of Ransomware FamiliesThis project classifies user-drawn input into one of the three datasets included: an umbrella, a bee, or a house. This is accomplished by submitting a drawing to a recurrent neural network, trained on a subset of the “Quick Draw Dataset” from Kaggle. With these three categories, the algorithm performs with around 85% accuracy in training. The entire dataset consists of 345 classes but for the purposes of this project, only three were used. Additionally, in order to mimic a practical application of this project, a website consisting of a doodle board was also designed. This doodleboard allows a user to draw an image and that image is then classified. In order to make this system, python and more specifically tensorflow was used for the backend and javascript, CSS, and HTML were used for the front end.
• League of Legends Machine Learning Based Draft Website – AiLoLRansomware is a relatively new form of malware that has become increasingly prevalent within the last decade. As the name implies, ransomware’s primary goal is to extort payment from the victim by holding their files hostage. The victim’s files are encrypted, the decryption key is held behind a paywall, and the victim is given a time window to make the payment before the key is destroyed. Payments are made via cryptocurrency or a prepaid card. Ransomware has been central in several high-profile attacks on home users and businesses alike, resulting in billions of dollars in damages. One ransomware family, WannaCry, is even believed to have originated from North Korea. While there is a great deal of variation between ransomware families, and even between samples within a family, we believe that a significant amount of ransomware shares certain characteristics that could be used to detect and prevent attacks. Towards this end we have collected data from reports on five prolific ransomware families with the goal of finding commonalities between them.
• LinkThe goal of this project is to create a website that will allow a user to leverage deep learning with a Recurrent Neural network in order to predict and improve on drafts in the video game League of Legends. We aim to create a website that allows a user to input each character in the order that they are picked by both teams, and at each pick all champions are ranked by the neural network for their suitability to be picked next. Both the website and the machine learning model must be fast enough to use during the time limit of a normal draft phase, while still being accurate enough to give the best possible results. The outcome and success of the project will be determined by testing results when using the website and comparing them to other, non-deep learning, draft assist websites.
• Movienite: A Recommendation Engine for New MoviesAt its core, Link is a project that presented an opportunity to learn new technologies that are relevant in the modern software engineering industry. Mobile applications have become a staple of daily life, and possessing the skills and knowledge to build one has become increasingly valuable. The following paper details the construction of a web application as a proof-of-concept for a mobile application that provides the basic functionality of connecting with friends, scheduling events, and real-time messaging and location tracking. The inspiration behind Link was creating an application that, like many others, makes some aspect of life more convenient for the user. In this case, the problem that Link solves is communicating whether or not the individuals involved in an event will arrive on time. The report also included an outline of the technologies used to turn concept into reality. The application is built in the Vue.js framework and relies on Google Firebase to provide a real-time database connection. Supporting features and functionality are implemented using various front end modules, back end APIs, and Vue plug-ins. While Link is still considered to be under development, a demo of the live web application can be found here.
• School Time: A Multi-Schedule Bells SystemWe are proposing a website where users will be matched to movie critics who have similar taste profiles as them, allowing us to recommend them movies that are currently in theaters. Since critics have early access to movies, there is a lot of data online containing early reviews to movies, which we can web scrape websites like Rotten Tomatoes and Metacritic (which contain information on individual critics and their ratings for specific movies weeks before a film’s release date) to build an aggregate database of movies, critics, and their reviews. With this data, we will be able to use a powerful machine learning technique called collaborative filtering to give users recommendations. These recommendations are based on the rating patterns and preferences of critics. Ultimately, this will give us the opportunity to do what no one else does: give movie recommendations on the day of release, or even earlier.
This project is centered around the desire to have a single system to control the automated bell system at a multi division school or an environment that has multiple schedules that it would like to maintain. Most systems in use currently have very minimal control, poor accessibility, and are not easy to use. Few systems are able to handle multiple different schedules/sections of a building to account for schools that span multiple divisions, such as a K-12 school. The goal of this project is to provide a system that addresses these issues, making it easy to use, customize, and able to handle several different schedules/divisions.
Created: 30 April 2019
P.Y. Wang