Airfare prediction and classification

A machine learning exercise presenting a simple model for

• (1) prediction of airfares from the data contained in the search query using linear support vector regression implementation sklearn.SVM.LinearSVR.
• (2) classification of flight itinerary search results by likelihood the returned fare will ultimately be purchased by the consumer, using simple Naive Bayes implementation in sklearn.

Getting Started

The discussion is contained in the two ipython notebooks

• fare_regression.ipynb
• booking_classifier.ipynb.

These files contain code that will mostly work out of the box, but that depends on the existence of three csv files containing datasets in the working directory:

• airport_data.csv
• booking_data.csv
• search_data.csv

Due to github filesize restrictions, search_data.csv was partitioned and zipped. To recover the original search_data.csv file, simply run the following to generate search_data.csv

me@myshell $ unzip search_data_1.zip && unzip search_data_2.zip && cat search_data_1.csv search_data_2.csv > search_data.csv && du -h

The output should look like

search_data.csv
Archive:  search_data_1.zip
  inflating: search_data_1.csv
Archive:  search_data_2.zip
  inflating: search_data_2.csv
581M	search_data.csv

You should be ready to go.

me@myshell $ jupyter notebook fare_regression.ipynb &
me@myshell $ jupyter notebook booking_classifier.ipynb &