This study investigates the categorization of toddlers with Autism Spectrum Disorder (ASD) through the use of a wide range of machine learning models, such as Random Forest, LSTM, and different ensemble techniques. Performance evaluation included measures including F1 score, recall, accuracy, and precision. A notable performer with promising accuracy, precision, recall, and F1 scores was the LSTM-RF Fusion model. Notably, excellent accuracy and precision were also demonstrated by the hard and soft voting ensemble models. Furthermore, hybrid models that combined various basic classifiers demonstrated efficacy, demonstrating the potential of merging various approaches. These results highlight the value of early detection using cutting-edge machine learning techniques, which will eventually support early intervention initiatives and improve outcomes for people with ASD and their families. The work emphasizes the value of using machine learning to address the worldwide neurodevelopmental issue that Autism Spectrum Disorders (ASDs) provide. It also highlights the contribution of the LSTM-RF Fusion model to increased toddler classification accuracy.