首页|融合显隐式反馈的协同过滤推荐模型

融合显隐式反馈的协同过滤推荐模型

扫码查看
原文链接
  • 万方数据
  • 维普
融合显式和隐式反馈已被应用于提升推荐模型的性能，但是，现有的此类推荐模型未能保留显式反馈中反映用户偏好程度的信息，且现有研究认为拥有显式反馈的数据和仅拥有隐式反馈的数据对于模型具有同等影响，未能充分发挥显式反馈的优势。针对这些问题，提出一种新的融合显式和隐式反馈的协同过滤推荐模型(CEICF)。首先，所提出模型提取显式反馈中的特征得到用户/物品的全局偏好向量;然后，从隐式反馈中提取用户/物品的潜在向量，进而将两种向量进行融合得到用户/物品的偏好向量;最后，使用神经网络预测用户与物品交互的可能性。在训练模型时，定义一种加权的二进制交叉熵损失函数，加强显式反馈对模型的影响来增强模型捕获用户偏好的能力。为了验证所提出模型的有效性，在覆盖不同领域的现实数据集上进行实验，实验结果表明，CEICF可有效地融合显式和隐式反馈，且推荐效果相对于基线模型有显著提升。
A collaborative filtering recommendation model combining explicit and implicit feedback
The recommendation performance can be effectively improved by combining explicit and implicit feedback,but the existing recommendation methods fail to retain the information reflecting the degree of user preference in explicit feedback.Moreover,existing studies consider that the data with explicit feedback has the same impact on the model as the data with only implicit feedback and fail to give full play to the advantages of explicit feedback.To solve these problems,this paper proposes a new collaborative filtering recommendation model combining explicit and implicit feedback(CEICF).Firstly,the feature of explicit feedback is extracted to get the global preference vector of user/item,and the latent vector of user/item is extracted from implicit feedback.Then,these two vectors are integrated to get the preference vector of user/item.Finally,a neural network is used to predict the possibility of user's interaction with items.And when training the model,a weighted binary cross-entropy loss function is defined,which strengthens the influence of explicit feedback on the model to enhance the model's ability to capture user preferences.In order to verify the effectiveness of the proposed model,this paper conducts experiments on real datasets covering different domains.The results show that the CEICF can effectively integrate explicit and implicit feedback,and the recommendation effect is significantly improved compared with the baseline models.

information overloadpersonalized recommendationcollaborative filteringexplicit feedbackimplicit feedbackneural network

欧朝荣、胡军

展开 >

重庆邮电大学计算智能重庆市重点实验室,重庆 400065

信息过载 个性化推荐 协同过滤 显式反馈 隐式反馈 神经网络

国家自然科学基金国家自然科学基金重庆市教委重点合作项目重庆市自然科学基金重庆市自然科学基金

6193600162276038HZ2021008cstc2019jcyjcxttX0002cstc2021ycjhbgzxm0013

2024

10.13195/j.kzyjc.2022.0076

控制与决策
东北大学

控制与决策

CSTPCD北大核心
影响因子：1.227
ISSN：1001-0920
年,卷(期)：2024.39(3)
  • 20