| dc.description.abstract | Bachground: East Azerbaijan province has a second highest GC prevalence in Iran. Despite the efforts and advances in the technology, GC remains the most lethal cancer. Therefore, modeling of GC can play an important role in identification of susceptible areas and planning and prioritization.
Objective: This study was conducted with the aim of preparing a GC risk model in East Azarbaijan province using techniques, environmental factors, GIS and ML.
Method and material: The addresses of GC patients during the years 2015 - 2019 were received from East Azerbaijan Population Based Cancer Registry program. Then, by using Google Earth software, the addresses of 2884 patients were converted into geographic coordinates, and by using GIS, the geographical distribution map of GC was obtained. By studying related sources, 9 parameters affecting GC were selected, which were entered into WEKA software for modeling with LR and LMT and weight extraction. Finally, the zoning map of the risk of GC was obtained in the GIS software environment. AUC and RMSE values were used to evaluate the performance of the models.
Results: The proportion of GC in men compared to women was 2.09. According to the geographical distribution map of GC, foci of concentration were identified in the central areas. According to LR and LMT models, 8 counties were shown as high-risk and susceptible areas and and 5 counties were shown as low-risk areas. In this survey, the risk performance of LR model was better than LMT.
Conclusion: The use of machine learning models can identify high-risk areas of GC and help managers and policy makers to prioritize and plan facilities. This study is the first research executed in the field of GC risk, Therefore, more research is needed.
Key words: Stomach neoplasms, zoning, Geographic information systems, Logistic regression, Logistic model tree, East Azerbaijan, Iran | en_US |
