Research On The Construction Of Intelligent Guidance Model Based On Deep Learning

Objective: Currently,most of the existing intelligent guidance platforms at home and abroad are selective platforms,where patients need to choose relevant medical information options to register and make appointments.However,most patients do not have professional medical knowledge reserves,which leads to blindness and uncertainty in the selection of patients when registering on the guidance platform,and may even result in incorrect registration,At the same time,there are also some patients who directly choose a certain department or physician for registration based on the recommendation of their surroundings,which can easily lead to a shortage of diagnosis and treatment resources in some departments,or a waste of diagnosis and treatment resources in some departments,and even delay their own or other patients’ treatment opportunities,leading to the emergence of more high-risk conditions.Therefore,helping patients seek effective,efficient,and accurate medical treatment,helping hospitals balance the allocation of diagnosis and treatment resources,and helping doctors improve consultation efficiency are the starting points for this article’s research on the construction of intelligent guidance models.Methods: This study used Named-entity recognition technology throughout the construction of the intelligent guidance model,effectively solving the problems of patient registration errors,low efficiency of medical treatment,and effectively improving the allocation of medical resources in departments.The research work was divided into the following modules:(1)Build a medical information corpus.The asynchronous crawler method is used to obtain the interactive information of each department on the medical website.After the data preprocessing of the interactive information,the department corpus is constructed through<Chief Complaint Text,Department>;Constructing a disease corpus through<departments,diseases,symptoms>;Build a doctor information corpus,use jieba segmentation to segment the areas of expertise of doctors.The final doctor information corpus includes data information from five aspects: doctor name,professional title,department,field of expertise,and segmentation results,providing strong data support for the construction of intelligent guidance models;Based on the characteristics of the main complaint text,data augmentation methods were used to improve the generalization ability of the model and avoid overfitting caused by data scarcity.(2)Build a department recommendation classification model.To ensure the efficiency and verifiability of the model,this study selected the most suitable deep learning model for the medical field as the basic model through comparative experiments.Based on the characteristics of the main complaint text,the model was optimized to solve the problem of inaccurate entity recognition and improve the accuracy of the department’s recommended classification model output.(3)Build an intelligent guidance model for doctor recommendation classification.Firstly,the method of disease similarity matching is used to search for symptoms that are similar to the main complaint text entity in the recommended classification model of the department;Secondly,determine the possible diseases that the patient may suffer from based on the number of matched symptoms from most to least;Finally,the disease similarity matching method is used to calculate the similarity matching between the results of the segmentation of the domain of the TCM students in the department and the name of the disease,and output the relevant information of the recommended doctors according to the similarity from high to low.Result: The main achievements of this study include:(1)A medical information corpus has been constructed.We have constructed a department corpus,which includes 23 types of secondary departments and 9874 main complaint texts;We have constructed a disease corpus,which includes 23 types of secondary departments,520 disease names,and 2600 symptom words;A physician information corpus has been constructed,containing relevant information from 200 physicians.(2)A progressive data augmentation technology has been proposed.By analyzing the characteristics of the department corpus,four thresholds are set,and four intervals are divided.Progressive data expansion is carried out according to the size of the interval threshold of the entity label.On the premise of ensuring the integrity of the original data set,the generalization ability of the intelligent guidance model is further improved.(3)A department recommendation classification model combining CTRM model and classification voting algorithm based on deep learning is proposed.Comparative experiments have found that BERT-Bi LSTM-CRF is a deep learning model with higher recognition advantages in the medical field,and is suitable as the basic model for entity recognition in the medical field;The proposed CTRM model has an accuracy improvement of 4% compared to the basic model,and the F1 value has reached 94%;By further integrating the CTRM model with the classification voting algorithm,the generalization ability of the department recommendation classification model is stronger.(4)A doctor recommendation classification model has been constructed.The similarity matching method was used to achieve effective matching between patients and suitable doctors.Conclusion: This study constructs an intelligent guidance model based on deep learning,which effectively improves the accuracy of patient visits and to some extent improves the utilization of department diagnosis and treatment resources.The proposed progressive data augmentation method provides strong data support for the construction of intelligent guidance models and valuable data resources for the research of intelligent healthcare through the construction of a medical information corpus;The proposed department recommendation classification model that integrates the CTRM model and classification voting algorithm improves the accuracy of department recommendation,making the model more in line with the practical application scenarios of guidance,improving the accuracy of patient visits,and thus improving the reliability of intelligent healthcare;The proposed disease matching algorithm and disease matching algorithm have improved the efficiency of patients and doctors,and can to some extent alleviate the problems of tight medical resources in hospitals and imbalanced utilization of departmental diagnosis and treatment resources.