tf.flights.delay.nn

Create and train a neural net to classify flights

This script tries to train a neural net to classify the delayed/canceled/diverted status of a flight based

on a set of input features which represent the characteristics of a flight as described in the US department of transportation flight database

The script can run in two modes : A prepare mode which pre-processes the data and splits it into train, validation and test set.

A train mode which creates and trains a neural net model which can be monitored with tensorboard.

In order to run this script, the root directory must be modified to match the directory where the data is placed.

Data can be downloaded at https://drive.google.com/drive/folders/19oVuHTETul5BCPxUWU0aWQ_MTN3MAdmC?usp=sharing

It can be started for example with the command "python tf.flights.delay.nn.py --model_dir=Momentum_0.9_bs200_lr0.001 --learning_rate=0.001 --momentum=0.9 --batch_size=200"