Yan-Ting (Vicky) Liau

aditar@r.postjobfree.com 480-***-**** www.linkedin.com/in/vicky-liau/ github.com/vickyting0910 Richardson, Texas Ph.D. in Geospatial Information Sciences (GIS), Specialty in Spatial Statistics SUMMARY

Data Enthusiast, Agile Person, and Challenge Solver with 10+ experience specializing in spatiotemporal data collection, processing, mining, analysis, prediction, visualization, and possessing multi-disciplinary training. Seek a full-time opportunity to work as a full-stack Data Scientist. SELF-LEARNING PROJECTS

R & Python AL/ML Libraries (e.g., Tensorflow, PyTorch, pyspark); Azure AI; AWS Analytics tools; Tableau Autocorrelation Adjusted Algorithms to Enhance Prediction with Missing Data November 2020 – Present

Conducting pioneering literature reviews on how spatial autocorrelation contributes to machine learning algorithms

(e.g., Neural Network, Random Forest, SVM, KNN, AdaBoost, Bagging, Gradient Boosting, Reinforcement Learning).

Establishing optimization and regularization by long-memory autocorrelation, expecting reducing 50% biased results. Developing Spatial-Temporal Matching Algorithms to Predict Crime Incidents June – August 2018

Obtained high-dimensional data from Zillow, Yellow Page, Yelp, and Google Maps (705,874 incidents).

Invented the matching filter, incorporating cross-decomposition algorithms and its reverse function, accelerating the adjusted R squared values up to 0.5, even with anomalies.

Implemented the matching selection to seek the correlated structure by integrating elastic-net with clustering to regularize 484 spatial features for the hierarchical specifications in the generalized linear mixed model.

Designed the matching segmentation by implementing the Multivariate Adaptive Regression Spline with spatiotemporal lagged terms to handle long memory spatiotemporal series, upgrading to 1-hour forecast.

R & Python Statistical Libraries & SAS (e.g., generalized linear model, lasso, stepwise regression, exploratory spatial data analysis, geostatistics (Gaussian process), spatial autoregression, geographically weighted regression, eigenvector spatial filtering, spatially sampling, visualization(ggplot2, matplotlib)) Implications of Spatial Imputation-based Measurement Error Models January – November 2019

Convinced three agencies to obtain fine-resolution, diverse-sourced datasets (1TB big data), managed by PostgreSQL.

Constructed Spatial Imputation-based Measurement Error Models, extending from Econometrics literature.

Pioneered A/B tests to assess the impacts of using spatial imputations in regression, validating the theoretical models.

Improved regressional performance by spatially balanced samples, saving 50% data collection and computation costs.

Cooperated with my advisor to reexamine the model’s effectiveness using SAS, compared to the results using R.

Python Libraries of Web Scraping (e.g., selenium) & NLP (e.g., re, NLTK, fuzzywuzzy, python-Levenshtein) Automatically Identify Deals to Promote Sales via Slickdeals September 2020

Set up the system by web scraping and Natural Language Processing to promote 13 Frontpage and 17 Hot Deals from 9/4 to 9/25 with up to 40,080 views or 106 deal scores in a single post, and potentially pushed new product.

Python Image Processing Libraries (e.g., Scikit-Learn, Scikit-Image, Scipy.ndimage, OpenCV); ENVI; eCognition Examining Multiple Algorithms to Classify Muddy Water November 2019

Evaluated segment-based classification algorithms and improved at least 5% performance with spatial autocorrelation. Yan-Ting (Vicky) Liau

aditar@r.postjobfree.com 480-***-**** www.linkedin.com/in/vicky-liau/ github.com/vickyting0910 Richardson, Texas Environmental Gradient Segmentation to Improve Species Identifications January – August 2013

Devised the locally random walker segmentation over the most widely used software (0 found), saving $5200 a year.

WinBUGS; R & Python Libraries for Bayesian Analysis (e.g., MVPLN) & Parallel Processing (e.g., snow); SAS Comparisons of Model Specifications to Enhance Bayesian Prediction October – November 2017

Adapted Bernardinelli (1995) spatiotemporal Bayesian disease risks model for assessing prediction specifications. Multivariate Spatial Crime Analysis March – April 2017

Extended Principal Component Analysis to improve at least 0.1 R squared values and handled multicollinearity.

Implemented the Eigenvector Spatial Filtering in Python for further implementations of machine learning algorithms.

Excel; SPSS (e.g., factor analysis, imputations); Minitab (e.g., ARIMA model); ArcGIS Geodatabase Reconsider Missing Data Imputation for Time-Series Forecasting October – November 2012

Adapted imputations for Time-Series species dynamical models but indicated challenges of Missing Not At Random. Spatiotemporal Pattern Recognition Approach for Large, Infrequent Disturbances January – July 2009

Managed spatiotemporal data (10 GB) from over 400 archives and 10 institutes across 50 years.

Pioneered predictions of long-term landscape changes because 0 existing geographical modeling can handle extremes. WORK EXPERIENCES

Graduate Teaching Assistant - The University of Texas at Dallas September 2015 – June 2020

Led 2 team members for 3D visualization using ESRI ArcGIS CityEngine in 2016 GIS Day and won high appreciation from the National Geospatial-Intelligence Agency, deciding to visit us every year after this event.

Advised the bottom 25 % of students by gaining their confidence to finish labs independently and learn new techniques.

Served as the guidance counselor for colleagues to support their research and job seeking and students to fulfill their curiosity and assist their self-learning projects, especially regarding Python, R techniques, and statistical modeling. National Water Center Research Fellow - National Oceanic & Atmospheric Administration June 2016 – July 2016

Formulated a locally moisture-structural filter (a Newly Formed Moisture Index + Principal Component Analysis + Structural Similarity Indices) for detecting flooding extents with 5% and 10% accuracy improvements.

Provided counsels of image processing and classification skills to at least 10 research fellows with CS, EE background. Graduate Research Associate for NIJ Project - University of Oklahoma April 2014 – September 2014

Wrote SQL to extract 10 TB data (GPS tracking of offenders’ electronic tagging) from a PostgreSQL database.

Mined and detected crime patterns by Dynamic Time Warping, focusing on recurrent crimes (i.e., for anomalies).

Refined the geocoding methodology by text matching, 20% more match rates, and 80% less time, saving $3000 a year. Technical Workshops

Azure AI Engineer Workshop - UT Dallas Department of Computer March 2020 Architecting on AWS: Hands-on Workshop - UT Dallas Department of Computer Science November 2019 In-Class Practices using Numpy, SciPy, and PySAL for GitHub open-sourced Projects August – December 2011 Various Workshops in GIS, Statistics, Earth Sciences, Sociology, Epidemiology, etc., Taiwan 2005 – 2010 EDUCATION

Ph.D. in GIS (Spatial Statistics) UT Dallas, Richardson, Texas August 2015 – December 2020 Master in Geography, Arizona State University, Tempe, Arizona August 2011 – December 2013

Contact this candidate