Research On Patient Treatment Recommendation Based On Collaborative Filtering

With the development of industrialization and urbanization in China,as well as the continuous improvement of the medical and health system,there is an increasing interest in the medical field.Treatment plan recommendation is a complex process and is one of the important processes of clinical care.The traditional decision-making process for treatment options is prone to misdiagnosis and irreversible consequences to the patient’s body due to the differences in the level of knowledge and experience accumulated by physicians and the lack of effective sharing of relevant data and knowledge among hospitals.In addition,in the context of information overload in the medical field,how to use computers to obtain valuable information from complicated data and recommend reasonable treatment plans for physicians has become an urgent problem to be solved.In order to solve the above problems,this paper constructs a collaborative filtering-based patient treatment plan recommendation model,which works as follows.(1)To address the prognostic issues of concern in the process of treatment plan selection,a BP neural network-based prognostic prediction model was constructed for predicting the survival rate of patients under different treatment modalities within 5 years to provide an auxiliary aid for subsequent treatment plan recommendations.In addition,the BP neural network prognostic prediction model is used for feature selection,extraction of prognostic prediction features,and selection of appropriate parameters to make the prediction results more consistent with the requirements of clinical decision-making.(2)A model-based collaborative filtering model,i.e.,a random forest-based treatment plan recommendation model,is constructed for patients with common medical conditions.First,a clustering-based evolutionary random forest model is constructed to improve the accuracy of recommendations under large data samples;second,two random forest classifiers are constructed for recommending surgery types and detailed treatment plans under known surgery types,respectively,considering physician decision-making interventions;finally,the effectiveness of the method is validated on a clinical dataset of breast cancer patients.(3)For patients with atypical symptoms,a memory-based collaborative filtering model,i.e.,a treatment plan recommendation model based on user collaborative filtering,is constructed.Considering the characteristics of cancer with multiple treatment stages,multiple symptoms,and multiple pathogenesis,as well as the problems of data sparsity and low recommendation accuracy of traditional collaborative filtering,a fuzzy clustering integration method is used to classify patients,and the collaborative filtering method is used on the basis of the clustering results.In addition,to improve the recommendation accuracy,the patient similarity calculation is improved using the value function,and the idea of group decision making is used for score prediction.Finally,the effectiveness of the method is demonstrated through experiments with clinical data of breast cancer patients.