基于经食管超声心动图记录和基于项目协同过滤的混合推荐方法

发布时间：2018-05-11 22:43

本文选题：TEE超声记录 + 用户交互记录　；参考：《计算机应用》2017年S1期

【摘要】：为提升经食管超声心动图(TEE)超声记录的临床应用价值,提高医师的工作效率,提出基于TEE超声记录和基于项目协同过滤(Item CF)的混合推荐方法,结合TEE超声记录特征和医师用户交互记录实现病例系统的个性化推荐。首先依托TEE病例查询系统平台采集数据并进行预处理,包括TEE记录数据和医师用户记录数据的清洗,标出记录表中可能被作为推荐因素的项目并提取特征。然后针对不同医师查询习惯,分析医师-TEE记录交互数据,运用基于内容和Item CF的混合推荐,训练模型,同时根据医师反馈信息和TEE记录数据集的变化,动态更新推荐结果。经实验测试,形成了较稳定的推荐模型,并且在相似记录数目为50、推荐记录数目为10的情况下,推荐精度较高。
[Abstract]:In order to improve the clinical application value of transesophageal echocardiography and improve the work efficiency of physicians, a mixed recommendation method based on TEE ultrasound recording and item collaborative filtering item CFS was proposed. The personalized recommendation of the case system is realized by combining the TEE ultrasonic recording feature and the physician user interaction record. Firstly, the data are collected and preprocessed on the platform of TEE case query system, including the cleaning of TEE record data and physician user record data, and the items that may be used as recommended factors in the record list are marked and the features are extracted. Then, according to different doctors' inquiry habits, we analyze the interactive data recorded by the physician and use the mixed recommendation and training model based on the content and Item CF, and update the recommendation results dynamically according to the changes of the physician feedback information and the changes of the TEE record data set. The experimental results show that a more stable recommendation model is formed and the recommendation accuracy is higher when the number of similar records is 50 and the number of recommended records is 10.
【作者单位】：中国科学院成都计算机应用研究所;中国科学院大学;
【基金】：中科院西部之光人才培养计划项目(2012SZ0133)
【分类号】：R540.45;TP391.3

【相似文献】