This document proposes algorithms to improve music recommendation in session-based collaborative filtering by incorporating temporal context. It summarizes that previous approaches did not use temporal features extracted from listening sessions. The paper presents temporal context-aware algorithms that cluster sessions based on temporal features like time of day, and represent sessions as mixtures of topics using LDA. An evaluation on a dataset of over 19 million listening records found the LDA-based approach significantly outperformed traditional session-based CF, increasing hit ratio by over 200%. Future work could combine temporal properties with other context features and conduct a user study.