Bitcoin Sentiment Analysis
Addison Mcfeely, Jordan Fujioka, Myles Lewis, and Ning Li
In this paper, we try to analyze how the sentiments of the public, remarks of some celebrities, and news reports influence the bitcoin price. We build models based on the dataset of bitcoin price and text dataset from the source of Kaggle, Twitter, and OpenBlender. At last, we conclude that perdition models with sentiment analysis perform better than onefold price prediction. The bitcoin price is very hard to predict, and the predictive models are not so reliable to do the forecast for bitcoin price considering the accuracy rate of the models.
Does Happiness Matter?
The aim of this project was to build a deep learning model, in order to predict employee turnover based on employee’s self-reported happiness and satisfaction. The data comes from 34 companies in Barcelona that utilize HappyForce, an informative system that aims to collect data on employee satisfaction and daily happiness.
Energy in California: The Path to Renewable Energy
Jenna Vogelsang, Christa Copenhaver, Peichen Geng and Nabil Syed
California is currently taking an initiative to adopt clean renewable energy resources by 2035. Our project aims to examine the future of energy usage in the state of California and how different methods of electricity generation can impact the state in upcoming years. Our data was extracted from the U.S. Energy Information Administration and cleaned using Python. The researchers chose to do a time series analysis in hopes of predicting how energy usage will change as more renewable methods are adopted by 2035. The model chosen was the prophet model created by Facebook in 2017. This model was chosen because it accounts for seasonality and holiday effects which both have an impact on energy usage. The findings of our study indicate a steady upward trend of both electricity consumption and generation with the largest sources being natural gas and solar energy.
FOMO: Can Twitter Sentiment Predict Crypto Prices?
RJ Copeland and Robert Spring
The intent of our Capstone project is to observe the effect of Twitter sentiment on Crypto prices. The coins of interest are Bitcoin, Ethereum, and Litecoin. The data for sentiment analysis is pulled from Twitter. The Crypto data is comprised of static files showing minute based Crypto price changes relative to the coins listed above. The model used is a Long Short-Term Memory model (LSTM). Both the model and exploratory analysis are implemented in Python.
Forecasting Web Traffic
This project looks at the problem of forecasting future values of time-series data. Wikipedia has over 145,000 articles and their view counts available to analyze. Following the Exploratory Data Analysis (EDA) process I was able to develop an Auto-Regressive Integrated Moving Average (ARIMA) model that showed a prediction of view counts for the following 45 – 60 days. Web forecasting is gaining popularity and has many applications including load balancing for cloud services, and understanding user behavior.
Major League Soccer Analysis
With highly influential salary restrictions in Major League Soccer, organizations must navigate limited high-dollar signings and maximize effectiveness of lower cost players while translating their signings to positive team performance. My goal was to use historical player performances, salary data, and team results to find trends and understand the key metrics that result in successful team performances. This analysis can be leveraged by MLS organizations and scouts to sign the most effective players and increase chances of team success.
MLB Optimizing Outfield Alignments
This project serves to provide a program that uses variable Euclidean distance means model to create optimized outfield shifts for MLB teams. Utilizing variables such as hang time of the ball, outfielder’s reaction, jump, and speed. The result is a GUI developed in R that could be used both in a gameday setting and used by front offices to determine which players fit in to their defensive schemes better.
NFL Combine Research
Taking results from combine participants from several years and testing the data in a few different methods to find what kind of scores are needed to perform successfully in the NFL.
Last Updated March 24, 2022