A neuromarketing recommender based on EEG sensing and LSTM with attention

Conventional EEG-based recommendation systems, such as those employing standard LSTM, bidirectional LSTM, or attention-augmented LSTM, have demonstrated potential in decoding consumer preferences. However, these models often fail to account for irregular temporal intervals between decision-making events and tend to underutilize critical neural dynamics, leading to suboptimal performance in neuromarketing applications where timing and cognitive state transitions are pivotal. To address these limitations, we propose a distributed attention-aware long short-term memory (ataLSTM) framework designed for EEG-driven recommendation systems. This approach explicitly models temporal irregularities and leverages a distributed attention mechanism to prioritize EEG features most predictive of consumer “like” or “dislike” responses. Evaluated on an open-access neuromarketing dataset of EEG recordings from neurologically healthy adults (aged 18–38) using the 14-channel EMOTIV EPOC + headset (configured according to the 10–20 system), the framework analyzed brain activity during real-time assessments of 14 consumer product categories. Through a robust evaluation protocol employing stratified 10-fold cross-validation with five repeated runs, ataLSTM achieved a classification accuracy of 92.42%, surpassing LSTM, biLSTM, and attention-enhanced LSTM baselines by an average of 3.2% across categories. Optimal performance was observed with the 10–20-window configuration and 30% overlap, highlighting the model’s ability to capture nuanced neural variations. The distributed attention mechanism significantly reduced false negatives and enhanced preference discrimination. By embedding cognitive-state modeling into the recommendation pipeline, ataLSTM enables precise, real-time personalization of product suggestions, thereby improving targeted advertising, cross-selling, and customer conversion rates. This work establishes a new standard for brain–computer interface applications in consumer analytics, bridging neuromarketing research and practical recommender systems.