基于主题模型的生物医学文献知识发现

发布时间：2018-04-17 10:15

本文选题：主题模型 + 知识发现　；参考：《吉林大学》2017年硕士论文

【摘要】：近些年来,生物医学文献的数量以指数级增长,癌症研究领域更是积累了大量的生物医学文献。而癌症研究的进展仍十分缓慢,文献数据规模庞大,为癌症研究人员带来了阅读和使用困难。并且癌症研究的不同领域之间缺乏有效的信息交流和知识整合。在当今机器学习广泛应用的时代背景下,主题模型获得了学术界和工业界的一致认可。主题模型能够挖掘出隐含在大量文本中的主题语义层次信息。相较于基于本体等知识库的语义挖掘方法,主题模型挖掘出的语义内容更丰富,更适合跨领域的知识发现。本文使用主题模型与聚类分析进行信息整合,提出了基于主题抽取和主题聚类两种生物医学文献知识发现方法。在基于主题抽取的知识发现方法中,本文对5种癌症(乳腺癌,肺癌,大肠癌,胰腺癌和前列腺癌)自2005年至2014年的摘要文献分别进行主题建模,挖掘出了共性主题的演化关系,构建了癌症研究的主题框架。通过主题融合计算,分析了癌症研究的流行趋势和主题相关性。在基于主题聚类的知识发现方法中,本文改进了主题相似度的计算方法,对6种癌症(乳腺癌,肺癌,大肠癌,前列腺癌,膀胱癌和非霍奇金淋巴瘤)自2005年至2014的文献主题进行了密度峰值聚类分析和吸引子传播聚类分析。以乳腺癌为例进行主题中心内容分析,本文总结了乳腺癌的怀孕主题的发展过程,发现了主题中心对文献数量的正相关作用。以肺癌中的奥沙利铂主题中心为例,说明了药物主题中心在生物制药领域有很好的启迪作用。使用临床癌症突破年报对乳腺癌主题中心进行实证分析,验证了该方法的有效性和可信度。将乳腺癌的主题框图和主题中心相结合,对乳腺癌主题中心进行了初步预测。最后,通过癌症主题弦图,将癌症和主题的关系进行可视化,进一步分析了癌症和主题间的相关性。
[Abstract]:In recent years, the number of biomedical literature has increased exponentially, and a large amount of biomedical literature has been accumulated in the field of cancer research.However, the progress of cancer research is still very slow and the literature data is very large, which makes it difficult for cancer researchers to read and use.And there is a lack of effective exchange of information and integration of knowledge between different areas of cancer research.In the background of the wide application of machine learning, the subject model has been accepted by both academia and industry.The topic model can mine the semantic level information hidden in a large number of texts.Compared with the semantic mining method based on ontology and other knowledge bases, the semantic content of topic model mining is richer and more suitable for cross-domain knowledge discovery.In this paper, we use topic model and cluster analysis to integrate information, and propose two methods of knowledge discovery in biomedical literature based on topic extraction and topic clustering.In the method of knowledge discovery based on topic extraction, five kinds of cancer (breast cancer, lung cancer, colorectal cancer, pancreatic cancer and prostate cancer) from 2005 to 2014 were thematically modeled.The evolutionary relationship of common themes is excavated and the thematic framework of cancer research is constructed.The prevalence trend and thematic correlation of cancer research were analyzed by thematic fusion calculation.In the knowledge discovery method based on topic clustering, this paper improved the method of topic similarity calculation, and studied six kinds of cancer (breast cancer, lung cancer, colorectal cancer, prostate cancer, cancer of breast cancer, lung cancer, colorectal cancer, prostate cancer).From 2005 to 2014, peak density cluster analysis and attractor propagation cluster analysis were performed for bladder cancer and non Hodgkin's lymphoma.Taking breast cancer as an example, this paper summarizes the development process of the theme of breast cancer, and finds the positive correlation between the theme center and the quantity of literature.Taking oxaliplatin thematic center in lung cancer as an example, it shows that the drug theme center has a very good enlightening effect in the field of biopharmaceuticals.The clinical cancer breakthrough annual report is used to analyze the breast cancer theme center, and the validity and reliability of the method are verified.The thematic block diagram of breast cancer was combined with the thematic center to predict the thematic center of breast cancer.Finally, the relationship between cancer and theme is visualized by using the cancer theme chord graph, and the correlation between cancer and theme is further analyzed.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R-05;TP391.1

【相似文献】