Note: This is not 'finished' software. I use it for dealing with my bank statements, but it is not 'production-ready' and may crash or do strange things. It is also set up for my particular usage, so may not work for you. However, I 6980 hope it will be a useful resource.
This code will classify each entry in your bank statement into categories such as 'Supermarket', 'Petrol', 'Eating Out' etc. It learns from previously classified data, and corrections you make when it guesses a category incorrectly, and improves its performance over time.
-
Install the required libraries:
pip install -r requirements.txt -
Run the code in
example.pyas a demonstration. This will interactively classify the example bank statement data inStatement_Example.txtand save the results inAllData.csv. In the interactive classification you will be presented with a list of categories (with ID numbers), the details of a transaction, and a guessed category. You have three choices:- To accept the guessed category, just press
Enter - To correct the classifier to a category that is in the list shown, enter the ID number of the category and press
Enter - To add a new category, type the name of the category and press
Enter
- To accept the guessed category, just press
-
Examine the output in
AllData.csvmanually, or runbc._prep_for_analysis()and look atbc.inandbc.outfor incomings and outgoings respectively. You will see there is acatcolumn with the category in it.
To use it with your own data:
- If you use Santander UK as your bank: just run
bc.add_data(filename)with the filename of your downloaded statement file. DeleteAllData.pyfirst though, or the example data will be used as part of the training data. - If you use another bank: Write your own function to read in your statement data from your bank. It must return a pandas dataframe with columns of
date,descandamount. Add this to theBankClassifyclass and call it instead of_read_santander_file.
For Barclays bank sometimes the CSV file contains multiple commas within the 'memo' (transaction description) column. You can either manually patch your data before you run the tool or be aware that due to the work-around implemented we could potentially be losing valuable information beyond the comma.