Random tree algorithm-based pediatric pneumonia screening using heart rate, respiration rate and temperature

Dagdanpurev, Sumiyakhand; Tsogzolmaa, Udval; Sun, Guanghao; Choimaa, Lodoiravsal; Hakozaki, Yukiya; Baatar, Khishigmaa; Matsui, Takemi

doi:10.31546/jbrci.1003

Cited by 1 publication

(1 citation statement)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It then combines random sub-trees. It is noteworthy that this solution is not always responsive and slows down the calculation process depending on the number of created random forests (24). The following settings were applied to improve model performance:…”

Section: Random Tree Modelmentioning

confidence: 99%

Comparison of the performance of decision tree types in determining the most important factors associated with the death of patients with Covid-19 hospitalized

Karimzadeh

Eshrati

Ghaderi

et al. 2022

Preprint

View full text Add to dashboard Cite

BackgroundNumerous factors are associated with mortality of people with COVID-19. This study aimed to investigate the most important of these factors related to the death of hospitalized patients with COVID-19 using decision tree algorithms.MethodsThe study data included the information of 8295 patients admitted to the hospitals of Iran University of Medical Sciences, who definitely had COVID-19. After cleaning the data, the most important factors related to the death of patients with COVID-19 were examined based on the indicators of sensitivity, specificity, accuracy and the area under the ROC curve (AUC) using decision tree algorithms (CHAID, C5.0, Random Tree, QUEST, and CART) with a ratio of 70% training and 30% testing. Analyzes were performed using IBM SPSS Modeler software version 18.ResultsAnalysis of the study results based on the testing data showed in terms of the area under the ROC curve (AUC), CHAID algorithm with 99.4%, in terms of sensitivity, CART algorithm with 99.5%, in terms of specificity, CHAID algorithm with 87.3%, in terms of accuracy, CHAID algorithm with 97.7% and finally in terms of the positive predictive value, CHAID algorithm with 98% had the best performance.Also, the most important factors affecting death (four main factors) were isolation, hospitalization in the ICU, and age, respectively, according to CHAID algorithm; age, symptoms up to hospitalization, occupation, and isolation, respectively, in RF algorithm; isolation, age, diarrhea, and the province of residence, respectively, in C5.0 algorithm; hospitalization in the ICU, age, isolation and dyspnea, respectively, in QUEST algorithm; and hospitalization in the ICU, treatment protocols, age and isolation, respectively, in CART algorithm. According to these results, the best performance was related to the CHAID algorithm.ConclusionDecision tree algorithms, especially CHAID algorithm, showed good performance in determining factors related to death of patients hospitalized due to COVID-19. Accordingly, isolation, hospitalization in the ICU, age and underlying diseases such as cardiovascular diseases and diabetes were the most important factors associated with death of these people.

show abstract

Section: Random Tree Modelmentioning

confidence: 99%