Personalized emotion regulation and psychological intervention still face the challenge that fails to match users' immediate psychological state and long-term preferences at the same time in the digital music platform. In order to solve this problem, this study puts forward the Personalized Affective Music Transformer (PAMT) model; it is a personalized emotional music generation framework based on generative artificial intelligence (AI) and Transformer. Accurate emotional music generation for different users can be realized by integrating users' text evaluation, historical listening records, and individual psychological characteristics. PAMT unifies feature coding and personalized emotion embedding mapping; it also introduces a closed-loop feedback mechanism to dynamically adjust the music generation process, thus achieving emotional adaptation at both the conversation and user levels. The study takes 214 adult users aged 18-45 as the object, and collects their music behavior data and text evaluation in the past half year for experimental verification. The experimental results show that compared with the current mainstream Transformer generation models such as Generative Pre-trained Transformer (GPT)-3.5, GPT-4o, and GPT-4o Mini, PAMT has obvious advantages in many core indicators: its Valence Score and Arousal Score are above 0.81 and 0.75 on average; On the preference similarity and session emotional improvement index (SEII), it is about 0.8 and 0.4, respectively. This indicates that PAMT has higher stability and consistency in music preference matching and emotional improvement effects. Further analysis shows that PAMT maintains stable performance in different age groups, gender, and mental health groups, and can make adaptive adjustments to users' behavior habits and daily AI tool usage frequency. The above results verify the overall advantages of PAMT in personalized emotional music therapy compared with existing methods. This study aims to provide quantifiable and customizable music therapy solutions for music platform developers, mental health service institutions, and individual users. It can realize the combination of immediate emotional adjustment and long-term psychological intervention, and enhance the personalization and therapy value of digital music experience.
Dongyun Chang (2026) studied this question.