Background/purpose: Mathematics education develops critical thinking and problem-solving skills, yet middle school students often face challenges, including didactical, epistemological, and ontogenic obstacles, that influence their academic orientation. This study investigates how Self-Organizing Maps (SOMs) can address these challenges by clustering students based on shared characteristics and designing targeted interventions. Materials/methods: A mixed-methods design was used. Phase One analyzed data from 200 high school students to explore the link between middle school mathematics performance and academic orientation. Phase Two focused on 60 middle school students completing a mathematics assessment and orientation questionnaire. SOMs identified learning patterns and guided peer-learning interventions. Pre- and post-intervention performances were compared. Results: SOM clustering identified four distinct student groups. Targeted interventions significantly reduced mathematical errors, with a 45% overall decrease. Post-intervention, there was a 20% increase in students choosing the scientific track, reflecting improved confidence and interest in mathematics. Conclusion: SOM-based clustering effectively identifies and addresses learning obstacles, improves mathematical proficiency, and positively influences students' academic orientation. This approach highlights SOM's potential for customized education and its implications for teaching strategies and policy.