March Madness 2018

Prediction of March Madness 2018 Men's Tournament

The detailed project report can be viewed here.

Installation Instructions

1. Install Apache Spark and Apache Hadoop
2. Download and install R v3.4.3
3. Download and install RStudio v1.1.383
4. In case you have never used the libraries I used in my project, open the R console in RStudio and type the following lines:
   • install.packages("caret")
   • install.packages("SparkR")
   • install.packages("dplyr")
   • install.packages("magrittr")
   • install.packages("tidyr")
   • install.packages("ggplot2")
5. Open FullTeamData.rmd in RStudio
6. Go to line 11 (it will contain: currentYear <- 2016)
7. Set currentYear to the value of 2010 (it should look like: currentYear <- 2010)
8. Knit the file, this will create a FullTeamData.html file in the current directory that shows the results and a FullTeamData2010.csv file in the Data folder
9. Repeat steps 8 and 9 for values 2011, 2012, 2013, 2014, 2015, 2016, 2017, and 2018
10. Open Testing.rmd in RStudio
11. Knit the file (this will take about 20 minutes to complete), this will create a Testing.html file that shows the results
12. Open Submission.rmd in RStudio
13. Knit the file, this will create a Submission.html file that shows the results and submission_v1.csv, submission_v1_forBracket.csv, submission_v2.csv, and submission_v2_forBracket.csv in the current directory
14. submission_v1.csv and submission_v2.csv were the files submitted to the Kaggle competition, submission_v1_forBracket.csv and submission_v2_forBracket.csv were used to create brackets for the NCAA March Madness Bracket Challenge
15. To view the results of the three R scripts (FullTeamData.rmd, Testing.rmd, and Submission.rmd), open the respective html files that were created in any browser

* Note *

FullTeamData.rmd creates 9 datasets (FullTeamData2010.csv, ..., FullTeamData2018.csv) that combines all of the data from the 52 datasets given by Kaggle into one dataset for each year. Testing.rmd tests six different Logistic Regression models on every year's data, which comes from the datasets created in FullTeamData.rmd. Submission.Rmd uses the two best Logistic Regression models on the 2018 data to create output files used in the two competitions (Kaggle and NCAA Bracket Challenge).