vipulnj / football-ftr-prediction

football-ftr-prediction

Quick guide

• We first explore the data given to us which is available in original_data directory. This is covered in the 1_exploration.ipynb notebook file.
• As part of the same file, we check for missing values and possible discrepancies in the data. To fill in the missing values, we have downloaded data from the internet and saved it in the downloaded_data directory. It contains the information for all the five leagues from the 2008-09 season to 2017-18 season. The script to download these CSVs is 0_downloadMoreData.py. Once done, we use the imputed_data directory to save the imputed train and test dataframes for use in the next steps.
• Finally, now that there are no null values, we perform feature-engineering on the train and test dataframes keeping in mind the models we might use. This is covered in 2_featureEngineering.ipynb
• Further feature-engineering steps which are specific to the model we will be using is covered in the three 3_modelBuilding_*.ipynb files.
• The predictons for each of these three models are saved in predictions directory.

More information about steps taken in each steps can be found in the Ipython notebook files.