INDECS 20(2), 162-178, 2022 DOI 10.7906/indecs.20.2.8 Full text available in pdf and xml formats.

Received: 16th January 2022. Accepted: 25th March 2022. Regular article

ABSTRACT

Using open data and artificial intelligence in providing innovative public services is the focus of the third generation of e-Government and supporting Internet and Communication Technologies systems. However, developing applications and offering open services based on (open) machine learning models requires large volumes of private, open, or a combination of both open and private data for model training to achieve sufficient model quality. Therefore, it would be beneficial to use both open and private data simultaneously to fully use the potential that machine learning could grant to the public and private sectors.
Federated learning, as a machine learning technique, enables collaborative learning among different parties and their data, being private or open, creating shared knowledge by training models on such partitioned data without sharing it between parties in any step of the training or inference process. This paper provides a practical layout for developing and sharing machine learning models in a federative and open manner called Federated Learning Open Model. The definition of the Federated Learning Open Model concept is followed by a description of two potential use cases and services achieved with its usage, one being from the agricultural sector with the horizontal dataset partitioning and the latter being from the financial sector with a dataset partitioned vertically.

KEY WORDS
e-Government, open data, machine learning, federated learning open model

CLASSIFICATION

JEL:

D80