Utilizing Human Memory Processes to Model Genre Preferences for Personalized Music Recommendations

TL;DR

This paper presents a psychology-inspired approach to modeling music genre preferences by leveraging ACT-R memory activation. It introduces two predictors: $BLL_u$ (base-level activation for past frequency and recency) and $ACT_{u,a}$ (full activation with associative context from the most recently listened artist), and evaluates them on the Last.fm-derived LFM-1b dataset across three mainstreaminess groups. The results show that both memory-based models outperform traditional baselines, with $ACT_{u,a}$ achieving the strongest performance and offering explainability grounded in human memory processes. The work advances transparent, memory-based personalized music recommendations and lays the groundwork for future integration of procedural memory and broader contextual factors.

Abstract

In this paper, we introduce a psychology-inspired approach to model and predict the music genre preferences of different groups of users by utilizing human memory processes. These processes describe how humans access information units in their memory by considering the factors of (i) past usage frequency, (ii) past usage recency, and (iii) the current context. Using a publicly available dataset of more than a billion music listening records shared on the music streaming platform Last.fm, we find that our approach provides significantly better prediction accuracy results than various baseline algorithms for all evaluated user groups, i.e., (i) low-mainstream music listeners, (ii) medium-mainstream music listeners, and (iii) high-mainstream music listeners. Furthermore, our approach is based on a simple psychological model, which contributes to the transparency and explainability of the calculated predictions.

Figures (6)

• Figure 1: Schematic illustration of ACT-R. In our work, we focus on the activation equation of the declarative memory module.
• Figure 2: Calculation of the BLL equation's $d$ parameter. On a log-log scale, we plot the relistening count of the genres over the time since their last LEs. We set $d$ to the slopes $\alpha$ of the corresponding linear regression lines.
• Figure 3: Example illustrating the difference between $BLL_u$ (left panel) and $ACT_{u,a}$ (right panel) based on trattner2016modeling. Here, unfilled nodes represent target genres $g_1$ and $g_2$, and black nodes represent genres of the last artist listened to by the target user (i.e., contextual genres). For $g_1$ and $g_2$, the node sizes represent the activation levels and for the contextual genres, the node sizes represent the attentional weights $W_c$. The association strength $S_{c,g}$ is represented by the edge lengths. While $BLL_u$ determines a higher activation level for $g_1$ than for $g_2$, $ACT_{u,a}$ gives a higher activation level to $g_2$ than to $g_1$ by also considering the associative association based on the current context.
• Figure 4: Recall/precision plots for $k = 1 \ldots 10$ predicted genres of the baselines and our $BLL_u$ and $ACT_{u,a}$ approaches for the three user groups LowMS, MedMS, and HighMS. $ACT_{u,a}$ achieves the best results in all settings.
• Figure 5: Average pairwise user similarity for LowMS, MedMS, and HighMS. We calculate the user similarity using the cosine similarity metric based on the users' genre distributions. While users in the LowMS group show a very individual listening behavior, users in the HighMS group tend to listen to similar music genres.