This study aims to analyze the application of the deep learning approach in improving the learning outcomes of fourth-grade students at Anuban Tapraya School, Songkhla Province, Thailand, regarding plant-related topics and plant types. The research method used was Classroom Action Research (CAR) based on the Kemmis and McTaggart model, conducted in two cycles, each consisting of the planning, action, observation, and reflection stages. The research subjects were all 12 fourth-grade students. Data collection instruments included a learning outcome test that had been validated for validity and reliability, observation sheets, and documentation. Success was defined as individual achievement of a score of ≥75 and a class achievement rate of ≥75%. The research results show that the application of the deep learning approach can improve student learning outcomes, with the class average score increasing from 70.33 in Cycle I to 76.25 in Cycle II, and the classical mastery rate rising from 41.67% to 75%. In addition, students’ learning activities improved in terms of mindful learning, joyful learning, and meaningful learning. Thus, the deep learning approach can serve as an alternative instructional strategy to improve learning outcomes and active student engagement in science education in elementary school, although these findings are still limited to the context of a single classroom and require further, broader research.