Project Instructions
The final project is an individual assignment and will be a chance for you to perform an applied data mining analysis on a real data set. We will be working with the Washington Dulles Domestic 2016 data set for this project. The description of this data and the variables contained in it appear on the second tab of the Excel file.

The final project has the following four key deliverables:
1. The R notebook file that produces the data mining results and analysis
2. Exploratory Data Analysis (EDA)
3. Predictive Modeling – Regression and Classification
4. Summary of Results and Recommendations to Decision Makers

Washington Dulles Domestic 2016 (Excel File)
The Washington Dulles Domestic 2016 dataset contains data on over 30,000 outbound domestic flights from Washington-Dulles International Airport in 2016.

Executives at Dulles airport have hired you as a data science consultant to identify the factors that lead to overall flight delays and to recommend data mining algorithms that can predict whether a flight will be delayed or not once it has taken off from Dulles airport. Specifically, the broad questions that the Washington-Dulles International airport is trying to answer include:

1. What are the factors that contribute to overall (gate-to-gate) flight delays on outbound domestic flights? (measured by the overall_delay_min and overall_delay variables)
a. Are there particular seasons, time of day, or days of the week that have greater delays?
b. Are certain destination airports (city, state, or region) more prone to delays? If so, why?
c. Are there issues with the plane taxing process?
2. Can we accurately predict the number of minutes that a flight well be early or delayed once it takes off from Dulles airport?
3. Can we accurately predict whether a flight will be delayed (Yes/No) once it takes off from Dulles airport?

R Notebook File
Download the R Notebook Template.rmd file from the Final Project content folder. This is an R notebook file and you will be writing your R code and analysis in this file. This file must be opened with RStudio for full functionality. R notebooks are an easy way to communication results from data mining projects because you can write R code in special sections called chunks, followed by plain text outside of chunks for comments or analysis.
Any analysis that you need to write can be typed in this file outside of the R code chunks. Your R code needs to be placed within code chunks that you can create by typing “Ctrl” + “Alt” + “I” or by selecting “R” under the Insert button on the top right of the file in RStudio.

Exploratory Data Analysis and Variable Importance Analysis
This portion will represent the first two sections of your R notebook file.
1. Exploratory Data Analysis (EDA)
a. You must think of at least 8 relevant questions that explore the relationship between overall_delay_min and overall_delay and the other variables in the data set
i. An example would be “Are there particular months that experience a larger proportion of overall flight delays?”
ii. Are there certain airports more prone to delays? Why?
iii. What days of the week have the most delayed flights and which days have the least?
iv.
b. You must answer each question and provide supporting data summaries with either a summary data frame (using dplyr) or a plot (using ggplot2) or both
c. In total, you must have at least 5 plots and 5 summary data frames for your EDA section

Predictive Modeling
This portion will represent the third and fourth sections of your R notebook file. For ALL the predictive models that you try, you must use cross validation to study prediction accuracy. This means you must split your original data into a training and test set and perform the necessary model building steps as outlined in the course lectures and tutorials.
For both the regression and classification models, use the following variables as predictors:
month_text, distance, sch_dep_hour, sch_arr_hour, dep_delay_min, and taxi_out

1. Regression Analysis – Predicting overall_delay_min
a. Try at least 2 regression algorithms to predict the response variable, overall_delay_min
i. Algorithms you can try include:
1. Multiple Linear Regression
2. Decision Trees/Random Forest
3. KNN
b. Analyze the results from each model
i. Make a plot visualizing the R2 that you obtained on the test set for each model you try
ii. Write a short analysis of the model performance on the test data
p
2. Classification – Predicting overall_delay
a. Try at least 2 classification algorithms to predict the response variable, overall_delay
i. Algorithms you can try include:
1. LDA, QDA, Naïve Bayes
2. Decision Trees/Random Forest
3. KNN
b. Analyze the results from each model
i. Make a confusion matrix for each model using cross validation techniques (either apply the model to a test set or use a package’s built-in cross validation features such CV = TRUE in linear discriminant analysis)
ii. For at least one of your models, try a couple of different cut-off points (or different values of k for KNN) and write a short analysis of your false positive and false negative rates. Also, talk about which error rate you think is more severe in this setting.

Project Instructions The final project is an individual assignment...