Skip to content

Latest commit

 

History

History
90 lines (56 loc) · 3.04 KB

README.md

File metadata and controls

90 lines (56 loc) · 3.04 KB
Raw

Flight_Delay_Prediction

A Big Data Assignment regarding Spark, with Airbus data fetched and linear regression model

The Full report could be found here: REPORT

Getting Started

Dependencies

here are all the dependencies needed for the project

  • Here an easy script to download pySpark and java8. remember your path for the installation_folder
!apt-get install openjdk-8-jdk-headless -qq > /dev/null
!wget -q http://mirrors.viethosting.com/apache/spark/spark-2.4.7/spark-2.4.7-bin-hadoop2.7.tgz
!tar xf spark-2.4.7-bin-hadoop2.7.tgz

import os
os.environ["JAVA_HOME"] = "/usr/lib/jvm/java-8-openjdk-amd64"
os.environ["SPARK_HOME"] = "/installation_folder/spark-2.4.7-bin-hadoop2.7"
  • Clone this repo:
git clone https://github.com/LorenzoFramba/Flight_Delay_Prediction.git
cd Flight_Delay_Prediction

Install dependencies: Let's finish with running the setup.py function, to download any uninstalled library

python3 setup.py install

To Start the program

  • Select the --path at which the Airbus dataset is saved. If --path is not specified, the program assumes the Airbus is in the same folder as the project itselves. Make sure the name is 'year.csv' and year is a 4 digit number from 1987 to 2008.
python3 main.py --dataset 'year.csv'
  • You also have the option to choose the train/test split (default is 75 / 25), and the dataset sample size for training and testing with --dataset_size.

you also the ML model type between 'linear_regression', 'gradient_boosted_tree_regression', 'decision_tree_regression' and 'random_forest' (default : linear_regression).

The all option will train and test all the models, compare their respective R2 and select the best performing one.

python3 main.py --dataset 'year.csv' --model 'linear_regression' --split_size_train 75 --dataset_size 100000

Variable Selection

The selection of the variables is done by analyng patterns and correlation matrix ( select --view True to watch it). We selected this following variables together

  • "X1": ['DepDelay', 'TaxiOut']
  • "X2": ['DepDelay', 'TaxiOut', 'HotDepTime']
  • "X3": ['DepDelay', 'TaxiOut', 'HotDayOfWeek', 'Speed']
  • "X4": ['DepDelay', 'TaxiOut', 'HotDayOfWeek', 'Speed', 'HotMonth']
  • "X5": ['DepDelay', 'TaxiOut', 'Speed', 'HotDepTime', 'HotCRSCatDepTime', 'HotCRSCatArrTime']

By default, the model will run with the easier variable: X1. You have the option to use X5, which is the best performing one, by selecting "best" on --variables. You can also select "all" to try everything.

python3 main.py --dataset 'year.csv' --model 'all' --split_size_train 75 --variables 'best' --view True

if you have any doubts, use

python3 main.py --help