Hongling Yang

PROJECT 1: Fashion Product Image Classification with Convolution Neural Network

GitHub - hyang78227/CapstoneProjectTwo

• Data Augmentation: Leveraged ImageDataGenerator to bolster the representation of the minority classes in the dataset
• Deep Learning: Achieved >95% accuracy with a 5-layer CNN via Transfer Learning. Leveraged VGG16 to extract 20 features and built a CNN-based recommendation system for 4000+ fashion products.
• Optimization: Implemented the Hyperband algorithm for hyperparameter optimization, resulting in a predictive accuracy enhancement from 95% to 97.5%

PROJECT 2: A Google App Store Educational Apps Rating Analysis

GitHub - hyang78227/capstone-project3

• Data Exploration: Conducted EDA following data wrangling, pre-processing, and visualization techniques
• Classification: Employed Decision Tree (85% accuracy), Random Forest (92% accuracy), and Gradient Boosting (94% accuracy) classifiers to predict the rating tiers of educational apps
• Imbalance Resolution: Utilized the imbalanced-learn module to mitigate class distribution discrepancies, elevating minority class representation from an initial 10% to a balanced 45%
• Optimiation: Employed the Hyperband algorithm to systematically optimize hyperparameters in classification models

PROJECT 3: Big Mountain Ski Resort Ticket Pricing Study

GitHub - hyang78227/DataScienceG

• Modeling: Utilized Multivariate Linear Regression and Random Forest Regression techniques, resulting in value enhancements of $11 and $19 per ticket, respectively
• Pipeline Architecture: Formulated an optimized pipeline for data preprocessing, regression, tuning, and model selection, encapsulated in a singular Python notebook
• Optimization: Utilized GridSearchCV for hyperparameter tuning, enhancing model training accuracy from 92% to 96%.

OTHER PROJECTS

GitHub - hyang78227/Springboard

• COVID-19 Patient State Classifications: Using the South Korean COVID-19 dataset, a Random Forest Classifier was deployed to delineate patient states—'isolated', 'released', and 'deceased'—achieving a classification accuracy of 92%
• Flight Departure Delay Prediction: Employed Light GBM for the prediction of flight delays surpassing 15 minutes, attaining a model accuracy of 94%. The precision was augmented through Bayesian Optimization for hyperparameter tuning and strategic feature engineering
• Cigarette Sales Time Series Analysis: Performed a detailed analysis of Cowboy Cigarettes' historical sales datasets. Through time series forecasting, projected sales trends with a Mean Absolute Percentage Error (MAPE) of 3.5%
• Wine Customer Segmentation: Employed K-means clustering on wine customer datasets, segmenting customers by their responses to wine offers. The clustering exhibited a silhouette score of 0.75, indicating well-defined customer segments

• District Study: Analyzed associations between alcohol use and IPV; explored the alcohol-IPV-HIV infection relationship using structural equation modeling
• Alcohol & IPV Intervention: Tested a 2-arm pilot intervention for men, estimating required effect sizes for broader application
• Baltimore HIV Risk Study: Used mixed methods to study how neighborhood factors affect forced sex rates among African American women, influencing HIV risk
• Stress & HIV Risk Analysis: Employed a multilevel design to understand stress-related pathways between forced sex and HIV
• Risk behaviors.

• Spearheaded medical research, mentoring residents throughout the process
• Originated research concepts and outlined sampling strategies
• Enhanced the resident program with expertise in Statistical Computing
• Partnered with physicians in executing clinical trials.

• Geographic Information System (GIS) Research: Played an instrumental role in the 'Ride8 Project', targeting Ozone pollution analysis in El Paso, TX
• Collaborative Research: Teamed up with fellow researchers, culminating in multiple published papers
• Academic Endeavors: Imparted knowledge as an educator in mathematics and statistics courses.

• Academic Endeavors: Imparted knowledge as an educator in mathematics and statistics courses.