PORTFOLIO

Utilizing Tweets to Predict Future Trends in BTC

    Whether you’re bullish or bearish on Bitcoin you can agree on one thing – the commodity is volatile. Prices can drop 10-20% within an hour and then be up 30% from the original price 4 hours later. If there was a way to accurately predict the future trend of the commodity you could reap rewards many times over within a day.

    In this project I’ve scraped twitter by the hashtag #BTC to gauge sentiment on the volatile commodity and used it to get a prediction of the trend by the hour.

A Look at Weather and Traffic

    “LSTW is a large-scale, country-wide dataset for transportation and traffic research, which contains traffic and weather event data for the United States. In terms of traffic, we have several types of events including accident, congestion, construction, etc. In terms of weather events, we have several types including rain, snow, storm, cold weather event, etc. This dataset is continuously being collected from August 2016, and today it contains about 29.5 million traffic and weather events.”

    Traffic congestion is an inescapable problem of the modern world. I don’t believe there is a sane person fond of sitting in traffic.  In this project I utilize the LSTW data in a clustering algorithm to see if there are relationships in the data we can learn from.

Learning Through Gameplay

    In PBS KIDS Measure Up!, children ages 3 to 5 learn early math concepts focused on length, width, capacity, and weight while going on an adventure through Treetop City, Magma Peak, and Crystal Caves. Children navigate a map and complete various levels, which may be activities, video clips, games, or assessments. Each assessment is designed to test a child’s comprehension of a certain set of measurement-related skills.

    The ability to predict how a user will perform on an assessment at the end of the world could be used to change how a user is to interact within that world continuously. For example if a user is predicted to perform well less time could be spent within that learning segment so that they could move quicker to those they would perform worse on, leading to quicker overall learning.  In this project I’ve created and assessed several different supervised learning models to predict performance in user assessments.