References | Pros | Cons |
---|---|---|
Number of participants is adequate The performance of the model is provided with different measurements | Compared different machine learning techniques The used data are not enough to predict asthma attacks Population is not generalized Environmental risk factors were not considered | |
[35] | Number of datasets is adequate Model accuracy is high | Some personal risk factors were ignored Environmental risk factors were not considered |
[24] | Used exhaustive dataset Compared different machine learning techniques | The study proposed only a protocol without the results of the prediction process |
[33] | Applied real-time dataset Compared different machine learning techniques | The tweets were taken from the English language only The ED visits data were taken from one hospital only The accuracy of the model is low |
[26] | Used sufficient dataset | Air pollution was not considered |
[27] | Dataset has many records | Data was lack of patient demographic and residential location information |
[28] | Used sufficient dataset | Air pollution factors were limited |
[34] | Dataset has many records | Weather factors were limited |
[31] | Used exhaustive personal and environmental predictors Compared two different machine learning techniques Increased the model performance by applying a feature selection algorithm | Bio-signals were daily recorded by users The interpretation of the study is difficult for users No services after prediction |
[23] | Used exhaustive environmental factors | The prediction result is only three states, without any suggestions for treatment or precautions |
[36] | Used exhaustive environmental factors Compared different machine learning techniques The performance of the model is provided with different measurements | Patient medical history and bio-signals were not used Validation results were not expressed by numbers |
[25] | Model accuracy is high Map of polluted and safe areas was introduced | Study was performed on three participants only |
[37] | A complete framework was introduced to predict attacks, alarm used, and view polluted sites | The model is based on environmental data only It is a proposed framework only |
[32] | The proposed framework considering real-time prediction and warning users with risks | Weather factors were not considered Personal sensor does not give an accurate reading Difficult to employ |