Kawasaki disease (KD) is a febrile disease that affects children under 5 years of age and leads to serious cardiovascular complications such as coronary artery disease. The development of markers that can predict early is important to reduce the under- and misdiagnosis of KD. The aim of this research was to develop a diagnostic predictive model to differentiate Kawasaki disease (KD) from other febrile diseases using eosinophil-to-lymphocyte ratio (ELR) and other biomarkers. We recruited a total of 190 children with KD and 1604 children with other febrile diseases. We retrospectively collected clinical information from the children, which included laboratory data on the day of admission, such as white blood cells (WBC), hemoglobin (HGB), calcitoninogen (PCT), hypersensitive c-reactive protein (CRP), snake prognostic nutritional index (PNI), peripheral blood neutrophil–lymphocyte ratio (NLR), platelet-lymphocyte ratio (PLR), and ELR. We performed analyses using univariate analysis, multivariate logistic regression, and column line plots, and evaluated the diagnostic parameters of the predictive models. ELR was significantly increased in patients with KD. After multivariate logistic regression, WBC, HGB, CRP, NLR, ELR and PNI were finally included as indicators for constructing the prediction model. The ROC curve analysis suggested that the C-index of the diagnostic prediction model was 0.921. The calibration curve showed good diagnostic performance of the columnar graph model. The cut-off value of ELR alone for KD was 0.04, the area under the ROC curve was 0.809. Kids with KD show highly expressive level of ELR compared to children with febrile disease, which can be used to diagnose KD, and column line graphs constructed together with other indicators can help pediatricians to identify KD more effectively from febrile children.