Resume

Sign in

Data Scientist, machine learning

Location:
LIC, New York, United States
Posted:
January 15, 2018

Contact this candidate

QING ZHANG

*-** **** ***, *** *E, Long Island City, NY 11101 917-***-****

ac31nh@r.postjobfree.com https://www.linkedin.com/in/qingzhang1990/ SUMMARY

Ph.D. candidate in Physics with strong analytic ability and mathematics background, proficient programming skills

(Python/SQL), experience with machine learning and deep learning, and work experience in the financial industry. PROFESSIONAL EXPERIENCE

Qizhiwanbang Beijing Science & Technology co., ltd., New York, NY Sep 2017-Present Intern

Machine Learning: Explored trading strategies [Python, Pandas, NumPy, SciPy, LightGBM]

• Built the Gradient Boosting Machine (GBM) model to explore trading strategies

• Improved results with tuning: practiced grid search to try a set of different learning rates

• Calculated the Profit and Loss (PNL) for the test data, got the correlation value 0.92 and the average PNL relative difference 21% between real trader action and model prediction Data Analysis: Improved stock recommendation accuracy [Python, Pandas]

• Calculated the correlation between stocks and their sector’s ETF

• Computed the proximity of companies by Breadth-First Search (BFS) Data Visualization: Visualized intraday stock data [d3.js]

• Created stock chart based on the theory of data visualization (Gestalt Principles of Perception, Pre-attentive Attributes, Data-Ink Ratio, Lie Factor, Rankings of visual encodings) City College & Graduate Center, CUNY, New York, NY Sep 2012-Expected May 2018 Deep Learning (Binary Classification): Binary Classification of Sonar Return Data [Python, TensorFlow, Keras]

• Built the neural network model to predict the type of objects from sonar return data (Accuracy: 82%)

• Improved the model performance with data preparation (84%), by tuning layers and neutrons in the model (86%) and using dropout on the visible layer (86%)

Deep Learning (Classification): Handwritten Digit Recognition [Python, TensorFlow, Keras]

• Built the baseline model with multilayer perceptrons (Accuracy: 98.27%)

• Built the small (99.00%) and large (99.17%) Convolutional Neural Network

• Improved model performance with image augmentation (ZCA whitening, random rotation, shifts, and flips) Machine Learning (Classification): Pima Indians Onset of Diabetes Dataset [Python, Pandas, NumPy, SciPy]

• Evaluated Classification Algorithms by Accuracy: Logistic Regression (Accuracy: 77.0%), Linear Discriminant Analysis (77.3%), k-Nearest Neighbors (72.7%), Naive Bayes (75.5%), Classification and Regression Trees (69.3%), Support Vector Machines (65.1%)

• Improved Performance with Ensembles:

Bagged Decision Trees (77.1%), Random Forest (77.1%), Extra Trees (76.0%), AdaBoost (76.0%), Stochastic Gradient Boosting (76.4%), Voting Ensembles (72.9%) Machine Learning (Regression): Investigated the Boston House Price dataset [Python, Pandas, NumPy, SciPy]

• Evaluated algorithms by Mean Squared Error before/after data standardization: Linear Regression (Mean Squared Error: 21.4/21.4), LASSO Regression (26.4/26.6), ElasticNet (27.5/27.9), Classification and Regression Trees (23.6/23.4), Support Vector Regression (85.5/29.6), k-Nearest Neighbors (41.9/20.1)

• Improved results with tuning: practiced grid search to try a set of different neighbors of k-Nearest Neighbors (18.1)

• Improved results with Ensemble Methods:

AdaBoost (15.0), Gradient Boosting (10.0), Random Forests (13.7), Extra Trees (11.5)

• Improved results by tuning the Gradient Boosting: tried a set of different boosting stages (9.4) Database: Social Networking Site [MySQL]

• Developed Social Networking workflow with functions of signing up, editing profile, searching friends, sending friend requests, accepting or refusing requests, adding status update and viewing friends’ status EDUCATION

City College & Graduate Center, CUNY, New York, NY Sep 2012-Expected May 2018 Ph.D. Physics, GPA 3.80/4.0

Fudan University, Shanghai, China Sep 2008-July 2012 BS Physics

TECHNICAL LANGUAGES & SKILLS

Programming Languages: Python, Matlab, Java, JavaScript, PHP Technical Skills: MySQL, TensorFlow, Keras, Algorithms, phpMyAdmin



Contact this candidate