**5. Conclusions**

This study compared eight machine-learning models for predicting the morbidities of six main gastrointestinal diseases from food-contamination data. Experiments on the datasets from ten cities/counties in central China demonstrate that the DNN models achieved significantly higher accuracies than the classical MLR and shallow ANN models, and the DDAE model with evolutionary learning exhibited the best prediction performance. Results also indicate that model accuracies are generally higher on acute gastrointestinal diseases than on other diseases, but it is difficult to predict the morbidities of gastrointestinal tumors. Moreover, a drawback of DNN models is that it takes significant effort to tune the structural parameters of the networks.

The studied deep-learning models could be utilized for the morbidity prediction of many other diseases whose influencing factors are large and complex. However, DNNs typically need to be trained on a large amount of labeled data, but disease- and health-related data are often very limited. Thus, we are currently studying unsupervised and transfer-learning technologies [34] for adapting the models from some well-known diseases to other diseases with insufficient data. Our future work also includes integrating the deep-learning models with fuzzy systems to handle uncertain information in the data [35,36], and utilizing the morbidity-prediction results for improving medical services, such as for medical-resource preparation and drug-procurement planning [37]. We believe that the combination of emerging deep-learning and intelligent decision-making technologies can significantly improve our society's healthcare services.

**Author Contributions:** Conceptualization, Q.S. and Y.-J.Z.; methodology, Y.-J.Z.; software, Y.-J.Z.; validation, Q.S.; formal analysis, J.Y.; investigation, Q.S.; resources, J.Y.; data curation, Q.S. and Y.-J.Z.; writing—original draft preparation, Q.S.; writing—review and editing, Y.-J.Z.; visualization, Q.S.; supervision, J.Y.; project administration, J.Y.; funding acquisition, Y.-J.Z.

**Funding:** This research was funded by the National Natural Science Foundation of China under grants number 61872123 and 61473263.

**Acknowledgments:** The authors would like to thank the Institute of Yichun Agricultural Science and Yichun University, China, for their help in data acquisition and processing.

**Conflicts of Interest:** The authors declare no conflict of interest. *Int. J. Environ. Res. Public Health* **2019**, *16*, 838
