University of Washington, Seattle Course-Projects

STAT 527 Nonparamatric Regression and Regression
Project Description:

1. Spam classification by building Classification Tree and pruning the tree;
2. Spam classification by building Ridge regression;
3. College data regression and selected features by the forward stepwise selection
4. Handwritten digit data identification by K-NN

STAT 534 Statistical Learning
Project Description: Hidden Markov Model with Multiple Sequence of Observations

1. Derived the iterated formula based on EM algorithm and Lagrange Multiplier for multiple sequences; Introduced scaling factor to avoid underflow appeared in the process of forward and backward; Optimized the number of hidden states by grid search and cross validation
2. Built the self defined class “HMM” without using built-in function
3. Applied Viterbi Algorithm (Dynamic Programming) to predict the most possible hidden and output state at the next node

STAT 536 Analysis of Categorical and Count Data
Project Description:Log-linear models, Hierarchical model, Feature Selection

STAT 554 Statistical models for Spatial analysis Project Description: Death of drugs in Ohio - Spatial Analysis

1. The data includes the health information of 88 counties in Ohio State. We want to explore the relationship between death due to drug abuse and HIV rate.
2. Matched FIPS codes in Ohio State shape files, detected spatial dependence by mapping the residuals from OLS fit and by Moran’s I test
3. Trained spatial lag model, improved the model by redefined the neighbors and introducing weighted spatial lag model

STAT 570 Advanced Regression Methods for Independent Data
Project Description: MLE, K-L divergence, Simulation