Yifan Guo is a PhD student at Case Western Reserve University studying machine learning and big data. He received his B.S. from Beijing University of Posts and Telecommunications and his Master's from Northwestern University. His research projects include developing an image recognition system for identifying pill types, building a movie recommendation system using matrix factorization, and designing an algorithm for a nonlinear integer programming transportation problem.

1. YifanGuo
PhD Student at Case Western Reserve University
yxg383@case.edu
Summary
I received the B.S. Degree in Information and Computing Sciences from Beijing University
of Posts and Telecommunications on June 2013, and Master’s Degree in Computer
Science from Northwestern University on February 2016. My potential research is
expected on related topics in Machine Learning and Big Data.
Research Projects
Smart Pill Box January 2015-December2015
Supervisor: Prof. Seda Ogrenci-Memik, Department of EECS, Northwestern University
Research Field: Machine Learning and Image Processing
Brief Description: The medication-related problem often occurs for the aged when wrong medicine is
being taken due to an increase in the number of tablet type being circulated throughout our society. In order
to overcome this obstacle, this project has developed an automated image-based content recognition
system based on the Bag of Features Model with feature vector of SURF descriptor and color feature
histogram. This system could not successfully make identification on pill’s type, but also effectively
recognize on wrong medicine mixed into the bottle and send notification to users timely. In the experiment,
840 pill images from eight categories and 8 distinct pill images for testing (without used in training) have
been used to verify our system. The result has shown a perfect performance with overall accuracy of 100
percent on recognition on single type of pill when applying multi-class classifier SVM using SURF, Hue,
and Saturation Feature. And the robustness to recognition on unknown type of pills is also verified. When
dealing with mixed type of pill, the system could successfully separate it from the single type of pills and
interact with user on the information of mixed pills. In future, the system would make contributions on the
design of a vision based Smart Cap design for a pill box at Northwestern University.
My Responsibility:
 “Intelligently” recognized pill images via bag of features model, and successfully applied appropriate
features combination and multiclass machine learning method on recognition;
 Conducted presentations during the research group’s weekly meeting to report on new potential method
and the experiment results from different methods and corresponding parameter setting;

2.  Implemented the entire code via Open CV in Python environment with packages in Scikit-Learn.
Movie Box March 2015-June2015
Supervisor: Prof. Doug Downey, Department of EECS, Northwestern University
Research Field: Machine Learning (Recommendation System)
Brief Description: The Recommender Systems have become extremely common in recent years, and
are applied in a variety of applications, especially on movies, music, news, books, to predict the preference
that user would give to an item. Our task is to recommend favorite movies to users and compute the
accuracy of our prediction based on the processed data that originated from Amazon movie websites. The
main method in our recommend system is based on Matrix Factorization Model with method of Singular
Value Decomposition. The processed dataset consists of 200, 000 ratings from 500 users on more than
9,000 movies. After applying data into our models, we will generate the predicted score for the unscored
movies. Then we choose RMSE (Root Mean Square Error) to evaluate the system and compare the
performance of different via with 10-fold cross validation. Our goal is to make RMSE under 0.9. Finally, we
use the model with best performance to output the recommendations, which achieves RMSE at 0.89 under
10-fold cross validation.
My Responsibility:
 Tested major methods in Recommendation System (e.g., Matrix Factorization Model with method of
singular value decomposition, User-based recommendation, and Item-based recommendation) and
determined the effectives on recommendation by comparing its Root Mean Squared Error;
 Implemented part of Java code of process including data parsing and collection, machine learning and
evaluation; wrote report on recommending favorite movies to users and computed the accuracy of our
prediction based on the processed data that originated from Amazon movie websites
Specific Transportation Problem September 2014-December 2014
Supervisor: Prof. Wenbao Ai, Member of Operations Research Society of China (ORSC) & Society
for Industry and Applied Mathematics (SIAM)
Research Field: Algorithm Design and Integer Optimization
Brief Description: As a core issue of combinatorial optimization theory, integer programming plays an
important role on task scheduling, network design, network routing and other important applications.
Although integer programming just turns real variables of linear programming to integer variables and linear
programming is polynomial-time solvable, integer programming is a typical NP-hard problem. Consider a
class of special transportation problems with stochastic loss, since such a problem is a nonlinear
integer programming problem, completely different from the classical transportation problem, there is no
ready-made algorithm available, which needs author to design a novel algorithm. By analyzing defects of

3. some preliminary designed algorithms, we develop the algorithm based on minimizing critical factor of
actual demands. Experimental results show that the algorithm successfully overcomes "pseudo-non-
feasible" situation under the plays of the preliminary designed algorithms.
My Responsibility:
 Analyzed and designed the symbol system and established the mathematical model; discussed related
algorithms and accomplished the corresponding implementation and verification with MATLAB;
 Completed the design of the algorithm for non-linear integer programming and wrote program
independently.
Trust Management Model of Virtual Community in P2P Network
May 2011-June 2012
Supervisor: Prof.YucuiGuo, School ofScience, Beijing University ofPosts & Telecommunications
(BUPT)
Research Field: Information Security
Brief Description: The existed trust management models in Peer-to-Peer (P2P) network mainly have two
problems. For one thing, the different influences on value of trust between short-term trading and long-term
trading are usually ignored. For another, the lack of the specific risk analysis on trading resources exists.
Consequently, focusing on the quality of different nodes and its opposite risk value, this project introduced
the concept of risk factor with setting up its value and proposed a trust management model based on
evaluation of value-at-risk with changing time. From the simulation results, a higher efficiency on
resisting malicious actions in P2P network is achieved, and it has confirmed to select better traders
effectively with a deeply quantitative analysis of trade resources through the model.
My Responsibility:
 Facilitated weekly discussion on the breakthrough and innovation of the research topic; modeled and
designed algorithms; analyzed the time complexity of the algorithms; explored the feasible algorithm;
conducted quantitative study of trust management of virtual community in P2P network;
 Accomplished the simulation by using MATLAB/Simulink; collected relevant paper and academic journals,
assisted other team members; wrote the paper;
Internship
Intern of Data Annotation at Samsung Advanced Institute of Technology in Beijing
July 2012-January 2013

4. My Responsibility: Served as the leader of intern group, assigned tasks and collected daily record of work;
studied the Human Body Recognition Project; collected and arranged the required image data; completed
the image annotation for subsequent debugging; responsible for programming according to the
requirements
Publication
 Yifan Guo, Teng Li, Yucui Guo, Trust Management Model based on Value-at-risk Evaluation with
ChangingTimeinP2P Network, Journal of Computer Applications (ISSN: 1001-9081. CN: 51-1307),
Sep. 2012.
Proficiencies
 Proficientin Python,Java, MATLAB, MySQL;
 Familiar with C Language, C++, HTML, CSS, JavaScript, MATHEMATICA, R; knowledgeable in PHP,
SAS;
 Experienced in fieldof machine learning, data mining, image processing, webdesign, and database
management