Bank Australia is awesome, but they aren't quite there when it comes to the data they provide to manage your budget. So, to help manage the home budget I hacked together this R script that takes a category file and performs a fuzzy string match on transaction description to your category search criteria and returns the best result as a category. It's a simple cobbling together of libraries that uses Jaro-Winkler distance for matching, then outputs the best category match data appended to the original transaction into the data directory for further use.
- Clone the repository into your R working directory
- Create a "data" folder in the repository root
- Download the transactions you'd like categorised, and move the file into your data folder named as 'transactions.csv'
- Create your cateogry file with a 'searchValue' column as the first column and save as a csv in the data directory as 'transaction-categories.csv'
- Open the budget-manager.R script in your R IDE of choice
- Run, and enjoy the newly categorised data!
You'll need to review and tune your categorisation inputs as the file takes the 'best' match. I have a soft distance threshold of 0.4 - anything over that has proven to be a poor selection in the model I use for category searching.
I may introduce a more direct threshold value and default everything over that set value to "uncategorised" to allow you to tune more effectively. If you've got other thoughts, pop them in as an issue or submit a PR!