This study examines a cohort of pediatric epilepsy patients in Ho Chi Minh City, Vietnam, specifically focusing on those with epilepsy attributed to infectious, autoimmune, or developmental epileptic encephalopathy etiologies. An artificial neural network (ANN) model is proposed to aid in the diagnosis, classification, and prognosis of treatment efficacy for this population. A retrospective analysis was conducted on 165 pediatric epilepsy cases admitted to City Children’s Hospital in Ho Chi Minh City, Vietnam, with etiologies including autoimmune encephalitis, infectious encephalitis, and epileptic encephalopathy. In the autoimmune encephalitis subgroup, no infectious agents were identified, with over 95% of patients exhibiting favorable treatment outcomes. Varied infectious markers significantly influenced treatment responses (p < 0.05). Treatment modality correlated with specific antigens (p < 0.05), with NMDAR antibodies being most prevalent and immunoglobulin IV therapy, corticosteroids, and plasma separation proving most effective. Prolonged hospitalization (>16 days) correlated with poorer treatment outcomes. Among pediatric patients diagnosed with infectious encephalitis, 22/31 cases exhibited pathogen detection in cerebrospinal fluid via PCR. Developmental epileptic encephalopathy typically manifested early and presented diverse EEG and MRI findings, often with “normal” recordings. Sudden death due to epilepsy accounted for 4/30 cases, with 17/30 cases (56.7%) exhibiting de novo complications. Distinct clinical and laboratory features aid in the early diagnosis of pediatric epilepsy associated with various brain pathologies. The proposed ANN model offers valuable support for emergency physicians in expediting classification, diagnosis, and treatment prediction for pediatric epilepsy patients, thereby enhancing survival rates.