Randomly simulated data is particularly useful when it's real world counterpart is hard access due to complexity, privacy and security reasons. Moreover, randomly simulated data has additional benefits including reproducibility, scalability and controllability.
This application aims to simulate telecommunication payments using random number generation. It includes typical transaction level relationships and behaviours amongst the user, device, ip, and card entities. It can be used in place of real world telecommunication payments for prototyping solutions and as an education tool.
A stable master version of the Random Telecom Payments data can be found on Kaggle here:
The underlying data model present in the simulated telecommunication payments is displayed below.
For a more detailed account of each column in the dataset see the data dictionary:
The following command can be used to execute the application:
python3 RandomTelecomPayments/scripts/main.py --factor 0.5 --randomseed 1 --nitr 3
The application works by first generating user level telecom payments data. Once generated, the user level data is exploded to transaction level, and any inconsistencies within the data model are removed. Finally, the transaction status and error codes are generated using underlying features within the transaction level data.
- factor - float, the factor to increase or decrease the number of randomly generated users by; a factor of 1.0 denotes a standard size of randomly generated users
- randomseed - integer, whether to run the random telecom payments data generation with or without a random seed set for reproducible results; must be 0 or 1.
- nitr - integer, the number of random telecom payments data batches to generate; must be at least 1. The python multiprocessing library is used to run each in parallel across all available cores.
The latests image can be found on dockerhub here: