Advanced

A Comparison of four methods of diseases mapping

Yang, Chao LU (2016) In Lund University GEM thesis series NGEM01 20161
Dept of Physical Geography and Ecosystem Science
Abstract
Disease susceptibility mapping can produce risk maps showing predictive distribution of disease incidences. Hence, it is a useful tool for disease prevention. Data-driven approaches for disease susceptibility mapping model relationships and patterns from experience data. Therefore, using data-driven methods can facilitate a simple and direct approach towards disease susceptibility mapping.

In this study, four data-driven models: logistic regression (LR), backpropagation neural network (BPNN), radical basis functional link nets (RBFLN) and general regression neural network (GRNN) for disease susceptibility mapping were implemented using Python. The performances of those four models were tested by a case study of visceral leishmaniosis... (More)
Disease susceptibility mapping can produce risk maps showing predictive distribution of disease incidences. Hence, it is a useful tool for disease prevention. Data-driven approaches for disease susceptibility mapping model relationships and patterns from experience data. Therefore, using data-driven methods can facilitate a simple and direct approach towards disease susceptibility mapping.

In this study, four data-driven models: logistic regression (LR), backpropagation neural network (BPNN), radical basis functional link nets (RBFLN) and general regression neural network (GRNN) for disease susceptibility mapping were implemented using Python. The performances of those four models were tested by a case study of visceral leishmaniosis (VL). Seven VL occurrence related factors, which are temperature, proximity to river, precipitation, proximity to nomadic villages, land cover, altitude, proximity to health-centers, were fed into the models as the input data.

In the results, the area under the receive operating characteristic curve (AUC) was used to test the discrimination ability of the models, and the BPNN generated the best AUC of 0.942. The GRNN classified 70% of the validation data into the right class. The AUC generated by RBFLN, LR and GRNN are 0.938, 0.899 and 0.88 respectively. Meanwhile, BPNN, RBFLN and LR correctly classified 80% of the validation data. All of them they predict that the northern and south-eastern part of the study area have high susceptibility in VL incidence. (Less)
Popular Abstract
Disease susceptibility mapping can produce risk maps showing predictive distribution of disease incidences. Hence, it is a useful tool for disease prevention. Data-driven approaches for disease susceptibility mapping model relationships and patterns from experience data. Therefore, using data-driven methods can facilitate a simple and direct approach towards disease susceptibility mapping.

In this study, four data-driven models: logistic regression (LR), backpropagation neural network (BPNN), radical basis functional link nets (RBFLN) and general regression neural network (GRNN) for disease susceptibility mapping were implemented using Python. The performances of those four models were tested by a case study of visceral leishmaniosis... (More)
Disease susceptibility mapping can produce risk maps showing predictive distribution of disease incidences. Hence, it is a useful tool for disease prevention. Data-driven approaches for disease susceptibility mapping model relationships and patterns from experience data. Therefore, using data-driven methods can facilitate a simple and direct approach towards disease susceptibility mapping.

In this study, four data-driven models: logistic regression (LR), backpropagation neural network (BPNN), radical basis functional link nets (RBFLN) and general regression neural network (GRNN) for disease susceptibility mapping were implemented using Python. The performances of those four models were tested by a case study of visceral leishmaniosis (VL). Seven VL occurrence related factors, which are temperature, proximity to river, precipitation, proximity to nomadic villages, land cover, altitude, proximity to health-centers, were fed into the models as the input data.

In the results, the area under the receive operating characteristic curve (AUC) was used to test the discrimination ability of the models, and the BPNN generated the best AUC of 0.942. The AUC generated by RBFLN, LR and GRNN are 0.938, 0.899 and 0.88 respectively. Meanwhile, BPNN, RBFLN and LR correctly classified 80% of the validation data. The GRNN classified 70% of the validation data into the right class. All of them had the ability to highlight the main spatial distribution of the risk of VL. (Less)
Please use this url to cite or link to this publication:
author
Yang, Chao LU
supervisor
organization
course
NGEM01 20161
year
type
H2 - Master's Degree (Two Years)
subject
keywords
artificial neural networks, Physical Geography and Ecosystem analysis, Logistic regression, disease susceptibility mapping, visceral leishmaniosis
publication/series
Lund University GEM thesis series
report number
13
language
English
id
8886305
date added to LUP
2016-07-01 16:08:14
date last changed
2016-07-01 16:08:14
@misc{8886305,
  abstract     = {Disease susceptibility mapping can produce risk maps showing predictive distribution of disease incidences. Hence, it is a useful tool for disease prevention. Data-driven approaches for disease susceptibility mapping model relationships and patterns from experience data. Therefore, using data-driven methods can facilitate a simple and direct approach towards disease susceptibility mapping. 

In this study, four data-driven models: logistic regression (LR), backpropagation neural network (BPNN), radical basis functional link nets (RBFLN) and general regression neural network (GRNN) for disease susceptibility mapping were implemented using Python. The performances of those four models were tested by a case study of visceral leishmaniosis (VL). Seven VL occurrence related factors, which are temperature, proximity to river, precipitation, proximity to nomadic villages, land cover, altitude, proximity to health-centers, were fed into the models as the input data.

In the results, the area under the receive operating characteristic curve (AUC) was used to test the discrimination ability of the models, and the BPNN generated the best AUC of 0.942. The GRNN classified 70% of the validation data into the right class. The AUC generated by RBFLN, LR and GRNN are 0.938, 0.899 and 0.88 respectively. Meanwhile, BPNN, RBFLN and LR correctly classified 80% of the validation data. All of them they predict that the northern and south-eastern part of the study area have high susceptibility in VL incidence.},
  author       = {Yang, Chao},
  keyword      = {artificial neural networks,Physical Geography and Ecosystem analysis,Logistic regression,disease susceptibility mapping,visceral leishmaniosis},
  language     = {eng},
  note         = {Student Paper},
  series       = {Lund University GEM thesis series},
  title        = {A Comparison of four methods of diseases mapping},
  year         = {2016},
}