Scholarlink

<p>Recommendation system has emerged as a core component of contemporary content platforms as they offer relevant and personalized recommendations based on the interest and needs of users. <i>Sequential recommendation</i> aims to determine and recommend items based on the sequential dependencies of interactions in the past. The major issues with the existing approaches are that they do not capture the union level and skip behavior of sequential patterns. Some existing approaches consider these factors (union level and skip behavior) however they have failed to capture long-range dependent sequential patterns. In this paper, we developed the LSTM-based Embedding Sequential Recommendation model (LESeR) to address these issues. LESeR approach models three critical aspects of the sequential recommendation process: (1) general or long-term preferences of users, (2) short-term or sequential preferences of users, (3) synergies (union level effects) among items, and skip behavior of sequential patterns of users. LESeR integrates LSTM to capture sequential patterns and a latent space representation of users to obtain general behavior of users. We compared the LESeR model with recent state-of-the-art approaches on four benchmark datasets. Our empirical results indicate that the LESeR model performs better than the state-of-the-art.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

LESeR: LSTM-based embedding sequential recommendation

  • 期刊论文

  • 发表日期：

    2025年7月8日

查看全文
  • Garima,
  • Somaraju Suvvari

摘要

Recommendation system has emerged as a core component of contemporary content platforms as they offer relevant and personalized recommendations based on the interest and needs of users. Sequential recommendation aims to determine and recommend items based on the sequential dependencies of interactions in the past. The major issues with the existing approaches are that they do not capture the union level and skip behavior of sequential patterns. Some existing approaches consider these factors (union level and skip behavior) however they have failed to capture long-range dependent sequential patterns. In this paper, we developed the LSTM-based Embedding Sequential Recommendation model (LESeR) to address these issues. LESeR approach models three critical aspects of the sequential recommendation process: (1) general or long-term preferences of users, (2) short-term or sequential preferences of users, (3) synergies (union level effects) among items, and skip behavior of sequential patterns of users. LESeR integrates LSTM to capture sequential patterns and a latent space representation of users to obtain general behavior of users. We compared the LESeR model with recent state-of-the-art approaches on four benchmark datasets. Our empirical results indicate that the LESeR model performs better than the state-of-the-art.