JMIR Publications

ABSTRACT

Background:

Machine learning(ML) is widely deployed in our everyday lives. Building robust ML models require a massive amount of data for training. Traditional ML algorithms require training data centralization, which raises privacy and data governance issues. Federated learning(FL) is an approach to overcome this issue. We focused on applying FL on vertically partitioned data, where a person's record is scattered among different sites

Objective:

This study aimed to perform federated learning on vertically partitioned data to enable the performance comparable to centralized models without exposing raw data.

Methods:

An overcomplete autoencoder was used for altering the original data into latent representations. We used three different datasets (Adult, Schwannoma, and eICU) and vertically divided each dataset into different pieces. Following the vertical division of data, each site independently performed overcomplete autoencoder-based model training. Following training, each site's latent data were aggregated for training. A tabular neural network model with categorical embedding was used when training. A centrally-based model was used as a baseline model and was compared to that of federated learning in terms of accuracy and area under the receiver operating characteristics curve (AUROC).

Results:

The autoencoder-based network successfully transformed the original data into latent representations with no domain knowledge applied. These altered data were different from the original data in terms of feature space and data distributions, indicating appropriate data security. The loss of performance was minimal; accuracy loss was 1.2%, 8.89%, 1.23%, and AUROC loss was 1.1%, 0%, 1.12% in Adult, Schwannoma, and eICU dataset, respectively.

Conclusions:

We proposed an overcomplete autoencoder based ML model for vertically incomplete data. Since our model is based on unsupervised learning, no domain-specific knowledge is required in individual sites. Under the circumstances where direct data sharing is not available, our approach may be a practical solution enabling both data protection and building a robust model.