智能导医系统中TF-IDF权重改进算法研究

发布时间：2019-08-12 12:05

【摘要】：面向患者的智能导医系统通过人工智能技术,依据患者症状计算可能疾病,引导患者准确挂号。目前智能导医系统多采用患者输入描述自身症状或者提问的方式,该方式易出现患者输入与医学专业症状词不匹配的问题,导致计算出的疾病可信度较低。针对这一问题,提出重心后移和医学专业语料库相结合的方法,对同义词匹配,映射出与患者症状对应的症状词;根据症状不论重要与否在每一疾病中仅出现一次的特点,提出基于患者关注度的症状词频计算方法;针对传统TF-IDF算法在待分类疾病类中数量分布不均时提取疾病效果差的问题,提出基于疾病类间分布的症状权重改进算法。实验结果表明,改进算法在疾病推荐正确率和可信度两方面具有更好的效果。
[Abstract]:The patient-oriented intelligent medical guidance system uses artificial intelligence technology to calculate the possible diseases according to the symptoms of the patients and guide the patients to register accurately. At present, the intelligent medical guidance system mostly uses the patient input to describe its own symptoms or ask questions, which is prone to the problem that the patient input does not match the symptom words of medical specialty, which leads to the low reliability of the calculated diseases. In order to solve this problem, a method of combining backward shift of center of gravity with medical corpus is proposed to match synonyms and map the symptom words corresponding to the symptoms of the patients, and according to the characteristics that the symptoms appear only once in each disease, a method of calculating the frequency of symptom words based on the degree of attention of the patients is put forward. In order to solve the problem that the traditional TF-IDF algorithm has poor effect on extracting diseases when the quantity distribution of disease classes is uneven, an improved symptom weight algorithm based on disease distribution is proposed. The experimental results show that the improved algorithm has better effect on the accuracy and reliability of disease recommendation.
【作者单位】：昆明理工大学信息工程与自动化学院;昆明理工大学云南省计算机技术应用重点实验室;
【基金】：国家自然科学基金(No.81360230) 科技部科技型中小企业技术创新基金资助项目(No.13C26215305404)
【分类号】：R-05;TP18;TP391.1

【相似文献】