The aim of this study was to develop a model to predict the extent of acute kidney injury (AKI) in children with idiopathic nephrotic syndrome (INS) using multicenter retrospective cohort studies. An internal cohort of 3,390 from a children’s hospital in Chongqing, China and an external cohort of 356 from 3 independent Chinese hospitals were used to verify model effectiveness. Of the internal cohort data, 70% was used for model training, 10% for testing, and 20% for internal validation while all of the external data was used for testing. Patients were included if they fit diagnostic criteria for INS and were between 6 months and 17 years of age without congenital or secondary nephrotic syndromes. AKI and chronic kidney disease (CKD) were determined by serum creatinine (SCr) measurements and glomerular filtration rate (GFR). Various demographic, clinical, and laboratory data were collected, while any data with a missing value ratio greater than 25% was removed. Each metric was analyzed through six different statistical methods and identified biomarkers closely related to AKI. Using 15 of the top variables, 5 machine learning (ML) models were created and then used to create a stacking integrated learning model. This stacking model had the best performance in both internal and external datasets. Sensitivity (0.789), specificity (0.815), accuracy (0.858), area under curve (AUC, 0.888), area under precision-recall curve (AUCPR, 0.892), positive predictive value (PPV, 0.816), negative predictive value (NPV, 0.787), and balanced accuracy of the model were superior in comparison to the other models. Out of 3,390 INS patients (internal cohort), 437 were diagnosed with AKI in which a number of features were significantly different compared to the non-AKI group. The top 5 features that influenced the ML algorithms in predicting AKI were nephrotoxic antibiotic exposure, cyclophosphamide exposure, respiratory tract infection, urine pH levels, and number of admissions. The study also noted that increased age, prolonged hospitalization, increased number of admissions, hypertension, and increased levels of blood phosphorus, potassium, and uric acid were associated with the AKI group. Overall, the study created a ML model demonstrating strong predictive performance in identifying children with INS likely to develop AKI and noted which variables were most significantly different and contributing to the ML prediction. The main limitation of the study was the specific region of China that most of the data originated from, potentially leading to bias.