Federated learning (FL) and split learning (SL) are two emerging collaborative learning methods that may greatly facilitate ubiquitous intelligence in the Internet of Things (IoT). Federated learning enables machine learning (ML) models locally trained using private data to be aggregated into a global model. Split learning allows different portions of an ML model to be collaboratively trained on different workers in a learning framework. Federated learning and split learning, each have unique advantages and respective limitations, may complement each other toward ubiquitous intelligence in IoT. Therefore, the combination of federated learning and split learning recently became an active research area attracting extensive interest. In this article, we review the latest developments in federated learning and split learning and present a survey on the state-of-the-art technologies for combining these two learning methods in an edge computing-based IoT environment. We also identify some open problems and discuss possible directions for future research in this area with the hope of arousing the research community’s interest in this emerging field.

In the current field of disease risk prediction research, there are many methods of using servers for centralized computing to train and infer prediction models. However, this centralized computing method increases storage space, the load on network bandwidth, and the computing pressure on the central server. In this article, we design an image preprocessing method and propose a lightweight neural network model called Linge (Lightweight Neural Network Models for the Edge). We propose a distributed intelligent edge computing technology based on the federated learning algorithm for disease risk prediction. The intelligent edge computing method we proposed for disease risk prediction directly performs prediction model training and inference at the edge without increasing storage space. It also reduces the load on network bandwidth and reduces the computing pressure on the server. The lightweight neural network model we designed has only 7.63 MB of parameters and only takes up 155.28 MB of memory. In the experiment with the Linge model compared with the EfficientNetV2 model, the accuracy and precision increased by 2%, the recall rate increased by 1%, the specificity increased by 4%, the F1 score increased by 3%, and the AUC (Area Under the Curve) value increased by 2%.

In the context of population aging, to reduce the run on public medical resources, nursing homes need to predict the health risks of the elderly periodically. However, there is no professional medical testing equipment in nursing homes. In the current disease risk prediction research, many datasets are collected by professional medical equipment. In addition, the currently researched models cannot be run directly on mobile terminals. In order to predict the health risks of the elderly without relying on professional medical testing equipment in the application scenarios of nursing homes, we use the datasets collected by non-professional medical testing equipment. Based on transfer learning and lightweight neural networks, we propose a disease risk prediction model, Diplin (disease risk prediction model based on lightweight neural network), applied to nursing homes. This model achieved 98% accuracy, 97% precision, 96% recall, 95% specificity, 97% F1 score, and 1.0 AUC (area under ROC curve) value on the validation set. The experimental results show that in the application scenario of nursing homes, the Diplin model can provide practical support for predicting the health risks of the elderly, and this model can be run directly on the tablet.

With the innovation of technologies such as sensors and artificial intelligence, some nursing homes use wearable devices to monitor the movement and physiological indicators of the elderly and provide prompts for any health risks. Nevertheless, this kind of risk warning is a decision based on a particular physiological indicator. Therefore, such decisions cannot effectively predict health risks. To achieve this goal, we propose a model Lidom (A LightGBM-based Disease Prediction Model) based on the combination of the LightGBM algorithm, InterpretML framework, and sequence confrontation network (SeqGAN). The Lidom model first solves the problem of uneven samples based on the sequence confrontation network (SeqGAN), then trains the model based on the LightGBM algorithm, uses the InterpretML framework for analysis, and finally obtains the best model. This paper uses the public dataset MIMIC-III, subject data, and the early diabetes risk prediction dataset in UCI as sample data. The experimental results show that the Lidom model has an accuracy rate of 93.46% for disease risk prediction and an accuracy rate of 99.8% for early diabetes risk prediction. The results show that the Lidom model can provide adequate support for the prediction of the health risks of the elderly.