My name is Tiffany Chan. I am a data scientist with a passion to apply statistics and artificial intelligence
to find solutions for scientific and business problems.
Please see below for my resume.
The rest of this page is dedicated to my projects on Artificial Intelligence, Statistics, Epidemiology, Forensic Science, and Drug Surveillance.
This project explores the most liked films by viewer ratings.
Key Items Assessed: Exploratory Data Analysis and Data Visualization using Python and Associated Libraries like Pandas, Seaborn, etc.
This is an assessment on the way we can improve banking marketing strategies to target reliable clientele.
Key Items Assessed: Logistic Regression, Confusion Matrix, Predictions based on a Classification Model, Data Wrangling/Cleaning
Using multiple machine learning models, this project juxtaposes the strength of supervised learning techniques to strongly identify potential customers for the next campaign.
Key Items Assessed: Decision Tree, Bagging, Boosting, Prediction Success, Dealing with Outliers and Data Anomalies.
A hard look at the possibilities of strengthening concrete by using discipline specific ratios, model tuning, and feature selection for dimension reduction.
Key Items Assessed: Feature Selection, Hyperparameter Tuning, K-Fold Cross Validation, Support Vector Machine, Support Vector Regression, K Nearest Neighbors, Gradient Boost, GridSearch CV
Without a target variable in hand, unsupervised learning was the proper way to approach this problem. K-Means and hierarchical clustering was employed to segment groups of promising customers and predict new customers based on similar properties.
Key Items Assessed: K-Means, Hierarchical Clustering, Silhouette Scores, Dendograms
Neural networks can be very powerful, especially for artificial intelligence. This is a project that explores the use of neural networks in solving business problems. It identifies the strengths and weaknesses of neural networks.
A lot of seedlings look alike. We can use computer vision to help to classify seedlings into their respective categories by getting the computer to learn recurring patterns from image edges and applying the learned patterns to untested data.
Natural language processing can be a useful tool. In this project, we comb through the Twitter Tweets to identify customer feelings about an airline's performance. This can be especially useful for improving a company's business performance.
Looking for causality is one of the most important goals in epidemiology. Type of insurance can affect HPV vaccine completion. In this project, there are also multiple confounders that must be acknowledged and handled. Logistic regression is used in a different way than it is in artificial intelligence to better understand how insurance affects general health.
Based on BRFSS data, we examine the effects of homeownership on mental health. Since the target variable is call data, we can use Poisson Regression modeling. Additionally, we also look at the application of logistic regression, negative binomial and zero-inflated negative- binomial regression models.
Survival analysis is useful to gage the efficacy of a treatment, methodology, or clinical factors in prolonging patient health. This project evaluates models consisting of different clinical variables of breast cancer to predict the prognosis (overall survival) of affected patients. The model will also assess if Oncotype Dx score adds any prognostic value to the model.
This project emphasizes on developing a mixed model that can summarize maternal anxiety levels of pediatric bone marrow patients across time. The advantage of the mixed model can display the overall distribution and detect a difference between mothers.
The Port Project is a collaborative project between the Center of Forensic Science Research and Education, US Customs and Border Patrol, and US Department of Justice. This project mainly informs the government of the drug composition trends that are coming into the country through our port borders. This is an ensemble of graphs that were made for the initiative.
Novel psychoactive substances like benzodiazepines have plagued America for a long time. This is a public health infographic that displays the number of driving incidents and deaths that were influenced by benzodiazepine intake in Pennsylvania.
Drug identification laboratory instruments are expensive, and not every country has the means of purchasing this equipment. The International Toxic Adulterant Database maps out the most common drug adulterant combinations in the world. This is one example from Argentina, graphed in R with UpSet Plot.