Log And Model Alignment Method Based On Clinical Path Optimization

To analyze the process of execution,the method of Process Mining is used to acquire the process knowledge.The knowledge is from the data generated and stored in the information system.In terms of clinical path detection and optimization,process mining had a good results.However,the clinical pathways in the currently obtained process models have problems such as single structure,weak robustness,non-standard and non-standard pathways,which cannot reflect the actual diagnosis and treatment process,high final implementation and maintenance costs,and lack of systematic and efficient evaluation methods.Therefore,this article intends to conduct the research of this article from the perspective of topic optimization,and through the topic optimization,and then better mining model.However,the existing research methods include problems such as semantic differences and the inability to obtain semantically consistent concepts when extracting activities;problems such as strong associations between the same medical item and different topics.Therefore,research on clinical pathways can solve the above problems on the one hand,and can also guide medical practice.In order to obtain a high-quality process model,this article is based on the clinical diagnosis and treatment data of sepsis patients,and researches on data-driven clinical path optimization and model alignment.The main content is: a topic optimization method for joint domain knowledge maps: KG-LDA（Knowledge Graph Augmented LDA Model for topic optimization）and The topic optimization method based on word embedding:WE-LDA（Word Embedding Augmented LDA Model for topic optimization）,through these two methods,the topic is constrained to obtain a more coherent topic.On this basis,the IM（Inductive Miner）algorithm is used for model mining.After the model is obtained,the existing log model alignment algorithm is used to align,and finally,the effectiveness of this algorithm is verified.In the experiment part,the real data set: Sepsis Case-Event Log is used to conduct experiments.First,the parameters used in topic optimization: the similarity threshold parameter θ and the optimized topic parameters α are experimented to obtain the corresponding values of the parameters.On this basis,the proposed method is compared with the existing algorithms: CPM-LDA（Clinical Pathway Miner-LDA）,Alpha+,IM and ILP（Inductive Logic Programming）algorithm by the experiment.It shows our methods’ four aspects,such as accuracy,recall,F1-score and fitness.The value on the quality index is better than the existing algorithm.