My data analytics portfolio
Hello, my name is Eric. This is my data anlytics portfolio: portfolio_Chi-Hsun_Chang.pdf. It contains a couple of projects I am working on recently and several research I conducted during grudate school. I hope it shows my interests and experience in using various statistical and data analytics technique to tackle complex problems, and present actionable insights as a succinct, powerful story.
This is part of the Brainfeed project. I scraped reddit posts and built a content-based recommendation model that recommends posts similar to a post an user reads using NLP, as well as built a content-based classification/prediction model that predicts reddit posts' topics using a Naive Bayes classifier. For more details, please see the GitHub project repository and my portfolio, pages 4-12.
In this project, I identified features and information of Korean dramas/shows (K-drama/shows) predictive of their ratings using correlation, regression, wordcloud, and data visualization. The findings may help people to find good K-dramas/shows. The ultimate goal is to build a K-drama recommendation app that can recommends good dramas/shows to users baesd on dramas/shows' information, or recommends dramas/shows based on users' previous viewed dramas/shows. For more details, please see the GitHub project repository and my portfolio, pages 15-25.
This is part of my Master project. I developed a data-driven image reconstruction pipeline that can recreate face images a person sees and recalls from memory. To acquire data, I designed a data acquisition software, Face Similarity Rating Task, which collects human observers' similarity judgments of faces in the laboratory and online. The similarity ratings were used as inputs for image reconstruction. The results demonstrate that image reconstruction is capable to recreate images of unfamiliar faces human observers perceive, those learned by participants during experiments, and images of celebrities subjects recollected from memory. The image reconstruction methodology may serve as the basis for future application, such as an automated sketch artist. For more details, please see my portfolio, pages 26-34
In my PhD, I further applied the image reconstruction technique to a large, diverse population varying in age, brain health, and personality disorder traits. I showed that the appearance of facial identity and expression, as well as specific facial features (e.g., eye shape, skin colour) were successfully recovered from perception and memory. The success of reconstruction elucidates the nature of the impact of various factors has upon face processing, confirms several theories in the field, and validates the applicability and utility of image reconstruction. For more details, please see my portfolio, pages 35-45