面向浏览者的图像推荐模型

要求互联网中的浏览者为每一幅图像表示其反馈是困难的,需要通过分析浏览者的行为,隐式地获取其评价。通过对浏览者的阅读、收藏和下载等行为的分析,度量用户对图像的关注度,以此作为用户反馈,分析其关键字偏好和图像特征偏好,进一步设计了用户偏好的遗忘策略和学习策略,实现用户偏好的动态更新,通过关键字相似性分析和图像特征相似性分析两方面,为用户选择推荐的图像。以准确度和召回度作为评价标准,实验表明,所提出的方法具有较高的性能。

Image recommendation model for web browser

It is difficult to ask browser to provide feedback for every image. Through analyzing user＇ s browsing behaviors in Web, such as reading, collecting and download, the attention degree of user about image can be measured. It indicates user＇ s feedback about this image. So, keyword preferences and image feature preferences are gained. Furthermore, forgetting strategy and learning strategy are designed to refresh user＇ s preferences. So, the proposed model provides recommendation from both of keywords and image features. Taking Precision and Recall as evaluation, experimental results show the proposed model is effective and has higher performance.