Related Experiment Video
Updated: Jun 9, 2025
An R-Based Landscape Validation of a Competing Risk Model
Published on: September 16, 2022
Development and validation of a potential risk area identification model for hand, foot, and mouth disease in
Xu Guang1, Yifei He1, Zhigao Chen2
1School of Public Health and Emergency Management, Southern University of Science and Technology, Shenzhen, China.
Abstract:
Maximum Entropy model (MaxEnt), as a machine learning algorithm, is widely used to identify potential risk areas for emerging infectious diseases. However, MaxEnt usually overlooks the influence of the optimal selection of spatial grid scale and the optimal combination of factor information on identification accuracy. Furthermore, the internal level information of factors is closely related to the potential risk of disease occurrence but is rarely applied to enhance MaxEnt's accuracy. In this study, the Optimal Parameters-based Geographical Detectors-Information Value-MaxEnt (OPGD-IV-MaxEnt) was first proposed to identify the potential risk areas of hand, foot, and mouth disease (HFMD) in Shenzhen and compared its identification accuracy with that of OPGD-MaxEnt and MaxEnt. Firstly, the optimal grid scale and optimal combination of factor information were determined by OPGD. Secondly, the contributions of factors' internal level information to the potential risk of HFMD occurrence were quantified and incorporated by IV. Lastly, the spatial patterns of potential risk areas and their main driving factors were elucidated. Results showed that: (i) Area under the curve (AUC) of single MaxEnt were 0.638, 0.688, 0.763, 0.796, and 0.757 at 100 m, 250 m, 500 m, 750 m, and 1000 m scale, respectively, and 750 m were deemed the optimal scale. (ii) At the optimal scale, OPGD-IV-MaxEnt (AUC = 0.868) identified potential risk areas more accurately than MaxEnt (AUC = 0.796) and OPGD-MaxEnt (AUC = 0.827). (iii) Resident (r = 0.61, q = 0.39) and Market (r = 0.61, q = 0.36) were the primary factors affecting the identification of potential risk areas. (iv) Potential high-risk areas of HFMD were mainly distributed in northwestern, southwestern, and central Shenzhen, with dense resident and market distribution. Such insights are instrumental in devising targeted infection prevention and control measures for emerging infectious diseases and provide references for improving the identification accuracy of similar machine learning algorithms.
Insights
A new OPGD-IV-MaxEnt model improves infectious disease risk mapping by optimizing spatial scales and incorporating factor information. This method enhances accuracy in identifying high-risk areas for diseases like hand, foot, and mouth disease (HFMD).
Area of Science:
- Environmental science and public health
- Machine learning applications in epidemiology
- Spatial analysis of infectious diseases
Background:
- Maximum Entropy (MaxEnt) is a common machine learning tool for identifying disease risk areas.
- MaxEnt's accuracy is often limited by suboptimal spatial grid scales and factor combinations.
- Internal factor information, crucial for disease occurrence, is underutilized in MaxEnt models.
Purpose of the Study:
- To propose and evaluate the Optimal Parameters-based Geographical Detectors-Information Value-MaxEnt (OPGD-IV-MaxEnt) model.
- To enhance the accuracy of identifying potential risk areas for hand, foot, and mouth disease (HFMD).
- To compare the performance of OPGD-IV-MaxEnt against traditional MaxEnt and OPGD-MaxEnt.
Main Methods:
- Determined optimal spatial grid scale and factor combinations using Optimal Parameters-based Geographical Detectors (OPGD).
- Quantified and incorporated internal factor level information using Information Value (IV).
- Applied the integrated OPGD-IV-MaxEnt model to map HFMD risk in Shenzhen.
Main Results:
- The optimal spatial scale for MaxEnt was identified as 750 m (AUC=0.796).
- OPGD-IV-MaxEnt achieved the highest accuracy (AUC=0.868), outperforming MaxEnt (AUC=0.796) and OPGD-MaxEnt (AUC=0.827).
- Resident and market distribution were primary factors influencing HFMD risk areas, predominantly in northwestern, southwestern, and central Shenzhen.
Conclusions:
- The OPGD-IV-MaxEnt model significantly improves the accuracy of identifying potential infectious disease risk areas.
- Optimizing spatial scales and incorporating factor internal information are critical for enhancing MaxEnt's predictive power.
- Findings provide a basis for targeted public health interventions and inform improvements in similar machine learning models.
