Metis is an immersive 12 week data science bootcamp in NYC focusing on machine learning, statistical analysis, project design,visualizations in Python, D3, and Flask.
These are the five projects I worked on while at Metis, starting with the most recently completed.
Project 5: Predicting Likes on Instagram (ConvNet)
Built a computer vision, convolutional neural network (ConvNet & Deep Learning) to analyze picture quality and predict likes based on a user's historical photos. Classified quality of training set photos based on expected likes per user, while considering follower count. Scraped images and meta data using Scrapy.
Project 4: Community Survey Natural Language Processing (NLP)
Natural Language Processing (NLP) on community survey long answer questions. Topic modeled with LDA and visualized in D3.js. Created Flask application to return similar words using word2vec. See first entree in Project section.
Project 3: Correlates of War, Predicting Hostility Levels
Predicted hostility level of word wide disputes using a time series random forest classification model, with data from the Correlates of War.
Project 2: Movie Series Analysis
Developed regression model to predict gross revenue, adjusted for inflation, with data scraped from IMDB using Scrapy.
Project 1: Exploring MTA Turnstile Data
Using the MTA publicly available subway turnstyle data we, in a team of four, analyzed and recommended subway stations to target for a street canvasing campaign.