Predicting airplane arrival time

R has built-in functionality for splitting up a data frame between training and testing sets, building a model based on the training set, predicting results using the model and the testing set, and then visualizing how well the model is working.

For this example, I am using airline arrival and departure times versus scheduled arrival and departure times from http://stat-computing.org/dataexpo/2009/the-data.html for 2008. The dataset is distributed as a .bz2 file that unpacks into a CSV file. I like this dataset, as the initial row count is over 7 million and it all works nicely in Jupyter.

We first read in the airplane data and display a summary. There are additional columns in the dataset that we are not using:

df <- read.csv("Documents/2008-airplane.csv")summary(df)...CRSElapsedTime      AirTime          ArrDelay          DepDelay       Min.   :-141.0   Min.   :   0     Min.   :-519.00   Min.   :-534.00   1st Qu.:  80.0   1st Qu.:  55     1st Qu.: -10.00...