Using Machine Learning and Google Searches to Predict Stock Market Changes

MA Thesis - Jarred Glaser

Spring 2018

This repo holds the code for my MA thesis: "Using Machine Learning and Google Searches to Predict Stock Market Changes".

MAThesisR.R

The main script that I used to generate features, run random forest and SVM models, and plot data.

Packages Used in this script:

• ggplot2
• tidyverse
• fpp2
• TTR
• pracma
• data.tables
• caret
• mlbench
• randomForest
• gridExtra
• ggpubr
• mlbench
• e1071

GoogleSearchVolume.py

A Python script I created and used to collect and normalize daily Google search data. Special thanks and credit goes to a module called pytrends, a pseudo API created by GeneralMills for grabbing data from Google Trends. See the module here.

Modules used in this script:

• pytrends
• datetime
• pandas
• matplotlib
• time
• random