Someone may launder money through bitcoin gambling. This project is to predict which addresses are gambling accounts.

We divided these accounts into three kinds: pool, exchange, and gamebling.

Users can refer to the results after predicting as the protect themselves from joining money laundering indirectly.

Python Versrion : 3.9+

Use this command on terminal to install necessary modules :

pip3 install -r requirements.txt

Use this command on terminal to execute program file:

python3 main.py

After you execute this python file, you will see the message:

Please insert the algorithms you want to use:

This time you can insert keyword of algorithm.

If you want to execute more than two algorithms, please segment them with commas.

And if you want to execute all the algorithm, please insert "all".

Then, you will also see the message:

Do you want to predict? (y / other value)

If you want to use these algorithms to predict, please insert "y".

If not, pleasw insert any other value except "y".

You will see some images and csv files created.

classify_tx_data.py

=> get_tx_data.py

=> get_tx_info.py

get_address_classfied.py

=> get_features

merge_data.py

This process takes about 40 minutes to ETL and generate features data. (tested in M1pro environment)

Fee to in ratio is fee divided by total IN amount in a transaction. And fee is the total IN amount minus total OUT amount.

Each account may join in serveral transactions, this feature (column) collect the maximun fee to in ratio among these transaction for each account.

This feature collect the 80% percentile value (fee to in ratio) among the transactions where a certain account join.

You can look up the keyword of algorithm in this file.

Beside, if you want to stop some algorithms from executing, comment out them in this file.

You can look up the column explanation in this file.