Fall 2018 C-Day Program

November 29, 2018

Location: Marietta Campus - Atrium (J) and Design Studio II Auditorium (I2)

An onlooker wearing a virtual reality (VR) headset while a student explains his C-Day presentation.

C-Day Home Page 

Return to the C-Day home page. 

A group of CCSE students posing with their C-Day awards.

Fall 2018 Winners

View the Fall 2018 C-Day winners.

Dean Jon Preston presenting at C-Day.

Flash Session Presentation

Download the C-Day Flash Session Presentation.

 

  • Time
    Event
  • 4:00 pm - 4:30 pm
    Student check-in time followed by set-up (presenters only) J lobby
  • 4:30 pm - 5:00 pm
    Check-in judges, industry partners,
    Networking J lobby
  • 5:00 pm - 5:35 pm
    Welcome from Dean Preston followed by Flash Session I2
  • 5:35 pm - 6:20 pm
    Judging of Student Posters and Games
    Browsing J building
  • 6:20 pm - 6:40 pm
    Pizza and Networking J152
  • 6:40 pm - 6:45 pm
    Introduction of Keynote Speaker (Dean Preston) I2
  • 6:45 pm - 7:00 pm

    Keynote Speaker: Nevarda Smith
    Vice President, Technology, MagMutual I2

    Nevarda Smith

    Innovation in an Executive Setting
     

  • 7:00 pm - 7:10 pm
    Recognition of Judges I2
  • 7:10 pm - 7:40 pm

    Presentation of Awards I2

    Sponsored by MagMutual

    • Best Undergraduate Capstone Project
    • Best Graduate Capstone Project
    • Best Graduate Research Project
    • Best Undergraduate Research Project

    Special Award "Most Impactful Cyber Project" Sponsored by SunTrust

Terabyte Sponsor: MagMutual

Kilobyte Sponsor:   SunTrust

    • Bob Cole - Managing Director - Accenture
    • Suneel Mendiratta - VP - Product Development - ADP
    • Jaspal Sagoo - CDC Chief Technology Officer - Centers For Disease Control and Prevention
    • Scott Bradshaw - Application Lead - Georgia Pacific
    • Bruce Skillin - Technology Innovator - Georgia-Pacific
    • Andrew Greenberg - Executive Director - GGDA
    • Eric Carrier - CEO - ISO Network
    • Dr. Meng Han - Assistant Professor - CCSE, Kennesaw State University
    • Evanda Remington - Director - R&D - Manhattan Associates
    • Shaun Sheppard - Lead Game Developer - Motion Reality, Inc.
    • Joe Cassavaugh - CEO/Designer/Engineer - Puzzles By Joe
    • Peter Vennel - Head of Data Management - SAFE-GUARD Products International
    • Justin Rose - IOS developer - State Farm
    • Keith Deininger - Sr. Information Security Analyst, Information Security Officer - SunTrust Banks, Inc., Enterprise Security & Resiliency (ESR)
    • Leslie Dugosh, PMP, CSM - Director of Program Management - Transaction Network Services, Inc

  • Capstone/ Undergraduate/Graduate Research scale 0 - 10 with 0 representing "Poor" and 10 representation "Exceeds Expectations"

    • Successfully completed stated project goals and reported deliverables (0-10)
    • Methodology/Approach: All required elements are clearly visible, organized, and articulated (0-10)
    • Effective verbal presentation (0-10)

    Games scale 0 - 10 with 0 representing "Poor" and 10 representation "Awesome"

    • TECHNICAL: Technically sound with appropriate visual & audio fidelity(0-10)
    • GAMEPLAY: Engaging & Fun, with an intuitive UI. Rules of play are clear. Includes a win/lose state(0-10)
    • ORIGINALITY: Sound, Art, Design, or Code(0-10)

  • * Candidates for the best undergraduate capstone project award

  • * Candidates for the best graduate capstone project award

    • *GC-01 Android "Identify" App
      by John McKinney (MSSWE), John Sineath (MSSWE), Mark Kordahi 
      Advisor: Dr. Reza Parizi
      The project developed a stand-alone Android application named "Identify" which utilizes deep learning CNN and R-CNN for the detection and classification of visual objects by analyzing either still images (photographs) or images seen thru a live camera lens.
    • *GC-02 Satellite Image Land Classification
      by Raymond Martin (MSCS)
      Advisor: Dr Mingon Kang
      This is a multiclass classification project surveying several methods. The dataset is the SAT-6 satellite image airborne dataset. It contains 404,000 samples and 3136 features per sample (RGB and infrared, 28x28 pixels). Each sample composes one of six labeled types of land (e.g. 'water', 'road'...). The project is done entirely in Apache Spark (using pyspark) on the KSU Spark server.

    • *GC-03 Toxic Comment Classification
      by Nusrat Asrafi (MSCS)
      Advisor: Dr Mingon Kang
      The aim of the project is to build a multi-headed model that's capable of detecting different types of toxicity like threats, obscenity, insults, and identity-based hate . I am using a dataset of comments from Wikipedia talk page edits. Improvements to the current model will hopefully help online discussion become more productive and respectful. I want to create a model which predicts a probability of each type of toxicity for each comment.
    • *GC-04 Predict Credit Risks
      by Liyuan Liu (Ph.D. in Analytics and Data Science), Yiyun Zhou (Ph.D in analytics and data science
      Advisor: Dr. Mingon Kang
      In general, credit risk datasets are label imbalance datasets, and logistics regression is one of the standard credit risk model in the industry for many years. In this project, I aim to conduct an experimental analysis on German Credit Risk Dataset to predict banking loan default using multiple strategies to deal with the imbalanced data, such as random oversampling, random under-sampling, SMOTE, SMOTEENN, weight-based method. After that, to examine the effects of comparing multiple strategies and different machine learning algorithms, I employ multiple machine learning algorithms: decision tree (DT), random forest(RF), logistic regression(LG), neural networks(NN), and support vector machine(SVM).
    • GC-05 CNN - Identify malign moles on Skin
      by sanjoosh akkineni (PhD Student, Analytics and Data Science)
      Advisor: Dr. Mingon Kang
      The idea is to develop a simple CNN model , and evaluate the performance to set a baseline. Data I have used is a set of images from the International Skin Imaging Collaboration: Melanoma Project ISIC. Following steps to improve the model are: Data augmentation: Rotations, noising, scaling to avoid over fitting Transferred Learning: Using a pre-trained network construct some additional layer at the end to fine tuning our model. (VGG-16, or other) Full training of VGG-16 + additional layer.
    • *GC-06 Cancer Detection-Atrous Convolution
      by NELSON Zange TSAKU (MSCS)
      Advisor: Dr. Mingon Kang
      For the past 50 years, Pathologists have had significant tedious moments providing accurate quantifications (e.g. tumor extent, nuclei count) and reduced variability between related research findings. In this work, we: - create algorithms to extract data - build models to ingest the data, - detect and classify key classes, following tested metrics through tuning and tweaking. Such an automated approach have shown to be beneficial in the context of significantly reducing such tedious processes while helping pathologists to reduce variability amongst themselves. Learn More
    • GC-07 Outcome Prediction in Intensive Car
      by Lauren Staples (PHD in Analytics and Data Science), Ryan Rimbey (MS-Applied Statistics)  
      Advisor: Dr. Mingon Kang
      This project uses an open-access database called Medical Information Mart for Intensive Care database (MIMIC3) developed from Beth Israel Deacon Hospital in 2012. The goal of this project is to predict outcomes (in this case, death within 30 days of hospital discharge) in an Intensive Care Unit Setting, with demographic data and billing/claims data. This project achieved 79% accuracy with logistic regression and 10-fold cross validation on a balanced dataset (equal number of deaths and non-deaths). This project uses a unique method of dimension reduction in handling categorical billing data codes (International Classification of Diseases, ICD-9) that achieves the same model evaluation characteristics as traditional one-hot encoding.

  • * Candidates for the best undergraduate research project award

    • *UR-01 CapsNet Traffic Light Recognition
      by Keshav Shenoy (HS Intern)
      Advisor: Dr. Selena He
      This research constructed a Capsule Neural Network and a Convolutional Neural Network to classify traffic light images by signal and looked at the differences between the two in accuracy of classification.
    • *UR-02 Phishing Analysis using ML
      by Suvan Paturi (HS Intern)
      Advisor: Dr. Dan Lo
      Various machine learning algorithms were applied to phishing website datasets to assess their performance by comparing calculated accuracy, false positive, and false positive rates for each algorithm. The testing was conducted using HPCC Systems ECL and the dataset used was obtained from the University of Huddersfield's ML Repository.
    • *UR-03 MS Office Macro Malware Detection
      by Ruth Bearden (BSCS)
      Advisor: Dr. Dan Lo
      Microsoft Office documents can contain macros, scripts designed to automate menial tasks or improve a document's user-interface. Malicious attackers, however, use macros as a means to download malicious payloads from the internet onto a host computer. Due to the ever-improving document engineering techniques these attackers employ, it is becoming increasingly difficult to visually identify a document containing macro malware, and users are susceptible to activating malicious macros. The goal of this research is to find the best way to detect macro malware automatically using machine learning. This project utilizes two kinds of data extracted from a sample of malicious and benign (safe) documents - VBA (the macro script source code) and P-code opcodes (a compiled version of this source code). Using this sample to train K-Nearest Neighbors, Random Forrest Decision Trees, and SVM machine learning models yielded a high prediction accuracy of 98% and revealed an interesting trend in the data that may help to improve this accuracy: using KNN and SVM, semantic information from the VBA data improves detection accuracy. We will use this trend to further explore classification using semantic-heavy analysis in the future.
    • *UR-04 Text-based Speaker Segmentation
      by Steffen Lim (BSCGDD), Sams Khan (BSCS)
      Advisor: Dr Chih-Cheng Hung
      Imagine a text to speech algorithm that can adjust it's voice based on contextual linguistic identifiers. Learn More

  • * Candidates for the best graduate research project award

    • *GR-01 VR/AR App for Non-Visual Navigation
      by Karis Kim (MSIT), Devan Patel (BSCGDD), Nick Murphy (MSCS)
      Advisor: Dr. Rongkai Guo
      People with visual impairments often require repetitive on-site training to memorize routes that they need to reach desired destinations. This study examined the feasibility of virtual environments as an Assistive Technology tool to supplement tradition training for daily use and pathway recall. Participants at the Center for the Visually Impaired in Atlanta were outfitted with testing equipment and asked to navigate pure virtual and mixed reality environments based on the equipment's verbal feedback. Data on the participant's movement, orientation, and confidence while using the system were collected to gain insights on the viability of virtual reality-based Assistive Technology for orientation and mobility. Learn More
    • GR-02 Customer Review Analytics
      by Jhanvi Vyas (MSIT)
      Advisor: Dr. Meng Han
      Analyzed the results obtained from experiments to analyze the impact of customer review on today's generation.
    • GR-03 Analysis of Top Grossing Apps
      by Qingliang Yang (MSIT)
      Advisor: Dr. Meng Han
      The online software distribution channels such as App Store has offered developers a powerful distribution mechanism. App store help users discover apps by providing categories and rankings. The ratings and reviews left by users in these online App store have the potential to influence new users. The motivation for this project is to find some clues to get high ratings to ensure that an app has a viable future. I think it is meaningful to the application company and developers.
    • GR-04 How Big Data Can Improve Healthcare
      by Shayan Shamskolahi (MSIT)
      Advisor: Dr Ying Xie
      This project explores the possibilities that big data offers in improving the healthcare industry. More specifically, it seeks to provide an example where big data can assist decision-makers in improving the quality and effectiveness of healthcare in hospitals across the U.S. In conclusion, the project will discuss the importance of accessing public data (specially healthcare datasets) and the role that it can play in public health.
    • *GR-05 Imbalanced Dataset
      by Wajira Abeysinghe (MSCS) 
      Advisor: Dr Chih-Cheng Hung
      Imbalanced dataset is available in many fields such as in credit card fraud detection, classification usually performs on the large number of normal transactions, to detect small percentage of fraud transactions.
    • *GR-06 Machine Learning for Fintech
      by Karl Kevin Tiba Fossoh (MSCS)
      Advisor: Dr. Dan Lo
      Knowing the importance of the detection of fraud in our banking system, especially to protect users from unlawful transactions, we decided to go for a diversified analysis of a transaction report dataset. This dataset enabled us to better understand what are the key parameters to identify for fraud detection along. The accuracy of those different classifications is based on different methods, each providing an insight on what parameters and data affect the most the detection
    • *GR-07 ML for IDS benchmark in HPCC System
      by Alexander Federico (MSCS)
      Advisor: Dr. Dan Lo
      The research looks into the performance of machine learning algorithms and how accurate the machine learning model can classify whether a packet is an intrusion or not.
    • *GR-08 Youtube8M Video classification
      by Karl Kevin Tiba Fossoh (MSCS), Maxwell Lavin (MSCS) 
      Advisor: Dr. Dan Lo
      The main motivation of the research was the capacity to summarize digital entities such as video, sounds, and images to a simple textual representation. One of the reason was to provide the ability to people to get fast and efficient summarization of long content without any risk of losing detail, information or precision. One simple approach was to try to work with videos. In effect, videos are a set of images frames associated with sound frames. The capacity to dissociate all of these features gives us a wide array to analyze each of them can affect the neural network designed to provide a classification and a captioning of the video, but also how each of these elements could be associated with another to provide a different result.
    • *GR-09 KNN Optimization with Vector Models
      by Arialdis Japa (MSCS), Daniel Brown (MSCS)
      Advisor: Dr. Yong Shi
      Optimization to the traditional implementation of the KNN algorithm by using vector space models.
    • *GR-10 Animal Identification deep learning
      by Joel Kamdem Teto (MSCS)
      Advisor: Dr. Ying Xie
      How can we outperform the best animal identification model on the SS project? Can capsule network perform better than CNNs in large and complex datasets? Can we build a capsule net that outperforms the best capsule net model on large and complex dataset?

 

 

©