Research And Implementation Of Doctor Recommendation Algorithm Based On Improved SVD

Nowadays,the world is an era of information explosion,and it is difficult for patients to quickly and effectively find useful information in a mass of miscellaneous medical information.Therefore,a recommendation system that can effectively analyze and process a large amount of medical information is of great significance for intelligent medicine.In the medical industry,the symptom data of patients and the evaluation data of doctors are very important.By mining the effective information in these data,doctors can be comprehensively evaluated,and then the processed information can be used to provide service for patients.At present,due to the increasing medical information data there are many problems in the medical recommendation system such as high computational complexity,low recommendation accuracy and difficult processing.These reasons lead to a large amount of time in the implementation of the algorithm.In order to solve these problems,the following two aspects are studies in the paper:Aiming at the problems of high time complexity and low recommendation accuracy,this paper introduces the density-based clustering algorithm to cluster patients according to their symptoms and find the neighbor sets of patients similar to it,so as to obtain the doctors rated by similar patients and form the patient-doctor scoring matrix R.An improved SVD model was introduced to decompose the score matrix R of patients to doctors into the matrix P composed of patients’ implicit features and the matrix Q composed of doctors’ implicit features.In order to reduce the time complexity of matrix decomposition,gradient descent method is introduced to solve P matrix and Q matrix.At the same time,the similarity coefficient between patients was introduced into the matrix P,and the similarity coefficient between doctors was introduced into the matrix Q to correct the similarity information between patients and doctors lost in the matrix decomposition process.Simulation results show that this algorithm achieves higher accuracy in recommendation than other algorithms.When dealing with huge amounts of data,the matrix decomposition algorithm encounters many problems,such as low processing speed,and too many computing resources.With the aim to solve these problems,the advantages of Spark memory and iteration calculation are used in this paper,and the parallelization of collaborative filtering recommendation algorithm based on matrix decomposition under Spark framework is realized.At the same time,Spark platform is used to design a doctor recommendation system based on collaborative filtering,and the implementation process of this recommendation system is described in detail:data analysis,data storage and preprocessing,modeling,recommendation model evaluation and doctor recommendation.Simulation experiments show that the system can improve recommendation accuracy and reduce computing time,which can provide reference and help for further research on Recommendation Algorithm of large data platform.