“Liyang interned in our data science team at LinkedIn this summer, working on an ambitious new text mining project. I appreciate that he refined the proposal with business partners and kept everyone well informed of the progress, delivered the customer sentiment model and analysis after researching multiple alternative solutions, and meticulously verified the outputs all within the tight deadlines. I was delighted to see his growth this summer culminating in his final presentations to both technical and non-technical audiences. After seeing Liyang take on these project challenges successfully, I am happy to recommend Liyang whole-heartedly.”
About
A full-time data scientist at LinkedIn, working in LinkedIn Talent Solution and Learning…
Activity
-
It’s time to say goodbye, LinkedIn. 4,123 days—that’s how long I’ve had the privilege to grow, learn, and innovate alongside some of the most…
It’s time to say goodbye, LinkedIn. 4,123 days—that’s how long I’ve had the privilege to grow, learn, and innovate alongside some of the most…
Liked by Liyang Zhao
-
Hi all! Meet Fangfang Tan, a Silicon Valley based entrepreneur and an applied machine learning and data science leader, who is dedicated to nurturing…
Hi all! Meet Fangfang Tan, a Silicon Valley based entrepreneur and an applied machine learning and data science leader, who is dedicated to nurturing…
Liked by Liyang Zhao
Experience
Education
Projects
-
Spotify Playlist Recommender
-
In this project, we worked with the one million playlists dataset provided by Spotify and attempted to build three different recommender systems to predict songs for playlists. We used the interaction matrix between tracks and playlists alone to train our models, and we evaluated the performances of each model using r-precision and discounted cumulative gain. Finally, we concluded that the K-Nearest-Neighbor-based ranking algorithm is the best performing model, achieving an average overlap…
In this project, we worked with the one million playlists dataset provided by Spotify and attempted to build three different recommender systems to predict songs for playlists. We used the interaction matrix between tracks and playlists alone to train our models, and we evaluated the performances of each model using r-precision and discounted cumulative gain. Finally, we concluded that the K-Nearest-Neighbor-based ranking algorithm is the best performing model, achieving an average overlap score of 0.7956 and an average DCG score of 0.4153 on test data. In addition to being accurate, the K-Nearest-Neighbor-based ranking model can generate 10 track predictions for a seeded playlist using 2.8 seconds on average, which is reasonably fast.
-
Eigenvalue Solver
-
- Implemented the QR algorithm with Wilkinson shift that finds eigenvalues of real symmetric matrices with cubic convergence
-
Time Series Prediction
-
- Performed method comparisons between SARIMA and ARCH/GARCH methods, and predicted the next 104 google trend-like data based on a given training set.
- Conducted cross-validation to select the model with the lowest MSE
-
Ant Colonies Extrapolation
-
- Implemented regression diagnostics on the ant data set by analyzing Standardized Residuals and Cook’s Distance.
- Performed model selections by comparing Adjusted R-square, AIC, BIC, Mallow’s Cp and selected the variables that may influence the mass of ants by k-fold cross-validation.
- Predicted Colonies of ants based on the selected variables by logistic regression and drew reasonable conclusions
Honors & Awards
-
Dorothea Klumpke Roberts Prize in Mathematics
University of California, Berkeley
Awarded to seniors who have demonstrated truly exceptional scholarship in mathematics
-
QuantEdge Award for Academic Excellence
University of California, Berkeley
Awarded in recognition of exceptional and sustained academic excellence among students of senior academic standing.
-
Kraft Award for Freshmen
University of California, Berkeley
For the academic achievement of the Top 3% of Berkeley’s freshman class
Test Scores
-
Exam P - Probability
Score: Passed
-
Exam FM - Financial Mathematics
Score: Passed
Languages
-
Chinese
Native or bilingual proficiency
-
English
Full professional proficiency
Recommendations received
1 person has recommended Liyang
Join now to viewOther similar profiles
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Liyang Zhao
-
Liyang Zhao
Wissenschaftsjournalistin | Moderatorin
-
Liyang Zhao
Senior Software Engineer
-
Liyang Zhao
A to Z Enterprise Solutions CEO, PCP Certified Payroll & HR Specialist
-
LIYANG ZHAO
Student at City University of New York-Baruch College
142 others named Liyang Zhao are on LinkedIn
See others named Liyang Zhao