Music has a significant impact on individuals’ emotional state and can serve as a valuable tool to influence their emotions. This paper outlines a personalised music suggestion system based on the user’s emotions and context. The main aim is to build and train personalised artificial intelligence models that can help users transition from an initial emotional state to a desired one. The proposed model has as inputs the user context data, the user current emotional state and a target emotional state. Emotional states are represented as a pair of Valence and Arousal values. The model outputs a pair of Valence and Arousal values that is subsequently used to choose a song intended to lead the user from the current to the target emotional state. To achieve this, there is the need for previously rated songs according to their emotional content measured as Valence and Arousal values. Therefore, Valence and Arousal regression models have been also developed. The user’s context data is also used as a way to better understand what can influence the user’s emotional state. The results obtained serve as a proof of concept of the proposed approach, as it was possible to obtain different recommendations for different users with the same initial emotion and similar contexts.