<![CDATA[Artificial intelligence has brought great innovation in the world of education, one of which is through chatbots based on natural language processing. This technology allows computers to understand human language naturally. However, educational chatbots generally only provide question-and-answer-based information without considering students' psychological aspects, such as math learning anxiety, which can have a significant impact on learning outcomes. Therefore, this study aims to compare the performance of Long Short-Term Memory and Neural Network models in the development of a chatbot to support learning mathematical opportunities based on CBT emotional solutions. The Cognitive Behavioral Therapy (CBT) approach, a psychological technique aimed at changing negative mindsets into positive ones, provides an opportunity to address this challenge by integrating it into educational chatbots. The method used is through data preprocessing with NLP techniques, training and model assessment utilizing metrics such as accuracy, precision, recall, and F1-score, the results show the NN model is superior. Training the NN model resulted in an accuracy value reaching 99.15% with 88.98% validation, higher than the LSTM which recorded 96.88% and 87.29% respectively. The NN model offers more effective responses and emotionally supports students in interactive learning. These results indicate that NN and CBT-based chatbots have great potential to enhance the mathematics learning experience if further developed, especially on the topic of chance, by supporting concept understanding and reducing student anxiety.]]>
