Ree Jung Kim

Machine Learning / Deep Learning Applications

Semantic segmentation

View repository

• Objective: Seek to answer what's in the image, and where in the image is the class located
• Model/Algorithm used: pascal voc, ade20k

Object detection

Application for object detection View repository

• Objective: detect and classify objects
• DL framework: OpenCV
• Model/Algorithm used: YOLO v3

Time series

View repository

• Model/Algorithm used: LSTM Keras View repository
• Model/Algorithm used: Prophet View repository



• Model/Algorithm used: Seasonal decomposition View repository

Bankruptcy prediction

View code

• Objective: Predict brankruptcies
• Model candidates: random forest, gradientboost, xgboost, lightGBM classifier
• Model selected: XGB classifier
• Feature reduction algorithm: Recursive feature elimination with cross validation (RFECV)
• Hyperparameter optimization: RandomizedSearchCV
• Evaluation metrics: confusion matrix, ROC, accuracy, precision, recall score, f1-score, support
• Model interpretation: feature importance, Shapley values, Morris Sensitivity

Housing price prediction

View repository

• Objective: Predict housing prices
• Data collection: Kaggle data set
• Data Preprocessing: label encoder, robust scaler
• Model/Algorithm used: random forest regression, support vector regression, gradient boosting regression
• Hyperparameter optimization: RandomizedSearchCV
• Evaluation metrics: mean absolute error (mae), mean squared error (mse), mean absolute percentage error (MAPE)
• Challenge: dimensionality reduction
• Model interpreation: tree feature importance, Shapley Additive Explanations(SHAP), Local Interpretable Model-Agnostic Explanations(LIME)

Spam email detection

View repository

• Objective: Spam classification based on NLP
• Data collection: Kaggle data set
• Data Preprocessing: spelling correction, remove stopwords, stem extraction (lemmatization), tokenizer, padding
• Model/Algorithm used: Tensorflow - Sequential, Embedding, Simple RNN

Fraud detection

View repository

• Objective: Identify fraudulent credit card transactions based on credit card transaction data
• Data collection: Kaggle data set
View code: supervised learning model
• Model/Algorithm used: random forest classifier
• Optimization:RandomizedSearchCV



View code: unsupervised learning model
• Model/Algorithm used: one-class support vector machine

Outlier detection

View repository

• Objective: detect/remove outliers
• Model/Algorithm used: Cook's distance

Market basket analysis

View repository

• Objective: Recommend items based on retail purchase data
• Data collection: UCI
• Model/Algorithm used: apriori function

Exploratory Data Analysis / Visualization

EDA & Visualization

• Uber pick up Visualization



• Funnel analysis



• Geographic sales analysis
• Corhort retention & CLV



• Instacart user purchase behavior
• 
• Korean Text analysis

Data collection method: web scraping reviews from Google Play Store View code





• Network analysis

Vision

View repository

Optimization

View repository

demo:



• Objective: Enable users to choose data range and easily get the optimal weight of portfolio
• Data collection: Stock data from yahoo using pandas data reader
• Model/Algorithm used: Markowitz portfolio theory
• Deployment method: Heroku, Streamlit

Experience