Scholarlink

<p>Recommendation systems (RS) are used to make suggestions to users based on multiple criteria. The abrupt increase in network information made it harder for users to accurately obtain information about movies, music, education, popular culture, and other fields. To address this issue, RS is used to extract the most pertinent information from the big data, enabling personalized services. The creation of an efficient RS has been the subject of several studies. Nevertheless, in order to provide Top-N recommendations, the current approaches had to overcome various obstacles. For top-N item recommendation, this paper proposes a new technique called the hybrid optimal deep learning (HODL) approach. The major intention of the proposed HODL is to provide a robust top-N-items recommendation based on the classified sentiments (Positive, negative, and neutral). Tokenization, spell checking, stemming, duplicate removal, and missing entry filling are first used in the pre-processing stage of the process to remove unnecessary, redundant information from the input text data. To achieve higher classification accuracy, sentiment-based features are extracted in the feature extraction stage using term frequency-based modified inverse class frequency (LTF-MICF) and glove modeling. Additionally, the sentiments are classified using the extracted features, and a novel Hybrid Optimal Cross-Layer Attention-based Convolutional Bidirectional Gated Recurrent Unit (HOCLA_CBiGRU) model is proposed to accomplish top-N-items recommendation. In this case, the proposed method uses ratings to rank the top-N items and categorizes the sentimental patterns. Also, a new Modified Mud Ring (MMR) algorithm is used to optimize the parameters in the proposed model in order to minimize the loss function. The proposed approach is implemented in the Python platform using the movie lens 100&#xa0;K dataset and assessed the performance in terms of precision, recall, f-measure, mean absolute error (MAE), and root mean square error (RMSE). Additionally, the performance of a proposed approach is compared with that of recent existing methods. Compared to the current methods, the proposed HOCLA_CBiGRU approach yields a better RMSE of 0.69 for top-10 item recommendations.</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

An efficient top-n-items recommendation system using a hybrid optimal deep learning model

  • 期刊论文

  • 发表日期：

    2025年5月7日

查看全文
  • Chetan J. Awati,
  • Suresh K. Shirgave,
  • Sandeep A. Thorat

摘要

Recommendation systems (RS) are used to make suggestions to users based on multiple criteria. The abrupt increase in network information made it harder for users to accurately obtain information about movies, music, education, popular culture, and other fields. To address this issue, RS is used to extract the most pertinent information from the big data, enabling personalized services. The creation of an efficient RS has been the subject of several studies. Nevertheless, in order to provide Top-N recommendations, the current approaches had to overcome various obstacles. For top-N item recommendation, this paper proposes a new technique called the hybrid optimal deep learning (HODL) approach. The major intention of the proposed HODL is to provide a robust top-N-items recommendation based on the classified sentiments (Positive, negative, and neutral). Tokenization, spell checking, stemming, duplicate removal, and missing entry filling are first used in the pre-processing stage of the process to remove unnecessary, redundant information from the input text data. To achieve higher classification accuracy, sentiment-based features are extracted in the feature extraction stage using term frequency-based modified inverse class frequency (LTF-MICF) and glove modeling. Additionally, the sentiments are classified using the extracted features, and a novel Hybrid Optimal Cross-Layer Attention-based Convolutional Bidirectional Gated Recurrent Unit (HOCLA_CBiGRU) model is proposed to accomplish top-N-items recommendation. In this case, the proposed method uses ratings to rank the top-N items and categorizes the sentimental patterns. Also, a new Modified Mud Ring (MMR) algorithm is used to optimize the parameters in the proposed model in order to minimize the loss function. The proposed approach is implemented in the Python platform using the movie lens 100 K dataset and assessed the performance in terms of precision, recall, f-measure, mean absolute error (MAE), and root mean square error (RMSE). Additionally, the performance of a proposed approach is compared with that of recent existing methods. Compared to the current methods, the proposed HOCLA_CBiGRU approach yields a better RMSE of 0.69 for top-10 item recommendations.