189results about How to "Improve accuracy and stability" patented technology

The invention provides a method used for simultaneous detection of three food-borne pathogenic bacteria based on multicolor upconversion fluorescence labeling. According to the method, three upconversion materials with differentiable fluorescence spectrums are used for forming multicolor upconversion fluorescent nanoprobes via respective connection with aptamers of staphylococcus aureus, vibrio parahaemolyticus, and salmonella, and complementary oligonucleotide single chains of the aptamers are connected with magnetic nanoparticles so as to form nano-composites. When bacteria to be tested are in a detection system, double chain unwinding is realized because of specific binding of the pathogenic bacteria with corresponding aptamers; it is possible to realize simultaneous quantitative determination of staphylococcus aureus, vibrio parahaemolyticus, and salmonella by monitoring upconversion fluorescence signal strength at 477nm, 550nm, and 660nm, detection linear range ranges from 50 to 1000000cfu/ml, and detection limits are 25cfu/ml, 10cfu/ml, and 15cfu/ml respectively. The method is used for detection of pathogenic bacteria, is high in sensitivity, is rapid and convenient, and can be used for detection of the three pathogenic bacteria in food such as milk and shrimp meat; and results are accurate and reliable.

The invention discloses a mobile pollution source emission concentration prediction method based on deep learning, and provides a convolutional long-short-term memory neural network prediction methodbased on an attention mechanism according to regional space-time distribution characteristics of mobile pollution source pollutants. Firstly, a Granger causal relationship between stations is analyzedand a hyper-parameter Gaussian vector weight function is developed to determine a spatial autocorrelation variable as a part of an input feature; secondly, extracting time-space characteristics of data used by the LSTM network by using a convolutional neural network, and meanwhile, attention models are respectively used for weighting a characteristic graph and a channel so as to enhance the effectiveness of the characteristics; finally, a time series predictor based on deep LSTM is used to learn long-term and short-term dependency of the atmospheric pollutant concentration. According to the method, inherent useful characteristics are extracted from historical atmospheric pollutant data, and auxiliary data are incorporated into a proposed model to improve the performance, so that the concentration prediction method is realized.

The invention discloses a light field image depth estimation method based on a hybrid convolutional neural network. The light field image depth estimation method includes the construction of a training data set, the training of a convolutional neural network model, and the generation of a depth estimation map of a light field image. According to the method, the light field depth calculation problem is converted into the classification problem and the relationship between pixel depths in a local area is effectively utilized; and light field data is represented by a four-dimensional parameter. By utilizing the fact that the slope of a straight line in an EPI image of the light field image is proportional to the depth of the scene, the EPI image is used as a medium to map a four-dimensional light field image into a two-dimensional image. By extracting EPI block regions corresponding to pixels in a center-viewing angle image of the light field image, a novel light field image depth estimation training data set is constructed by using a mode of polar line graph block region pairs. According to the light field image depth estimation method, the advantages of deep learning in feature abstraction is utilized, and the accuracy and stability of depth estimation of the method are more advantageous than those of an ordinary conventional methods.

The invention provides a parameter estimating method and a parameter estimating device of an embedded permanent magnet synchronous motor, wherein the parameter estimating method comprises the steps of S1, searching a stator winding resistance change table along with temperatures according to a stator winding temperature, obtaining a corresponding stator winding resistance, searching a rotor permanent magnet flux linkage change table along with temperatures according to an estimated rotor temperature, and obtaining a corresponding rotor permanent magnet flux linkage; and S2, using a stator winding resistance and the rotor permanent magnet flux linkage as parameters of an embedded permanent magnet synchronous motor state equation, and identifying a d-axis inductance and a q-axis inductance by means of a recursion least square method. According to the parameter estimating method and the parameter estimating device, table searching results of the stator winding resistance and the rotor permanent magnetic flux linkage are used as given parameters of the recursion least square method, thereby realizing relatively high accuracy and relatively high stability according to a real-time working condition change, reducing algorithm execution time and realizing easy practical application.

The single cell sensor for ectocytic action potential measurement is developed on the basis of multiple light source cell micro physiological gauge, and has light addressing potential sensor without limitation in cell culture and light source moving addressing capable of tracking and positioning randomly cultured single cell for measurement. The present invention has also improved light source system with raised precision and stability for locating the measurement in single cell or even in single ion passage, so as to measure precisely the action and action target point of medicine on single cell. The single cell sensor can realize the fast real-time monitoring of outer action potential of nerve cell and the qualitative analysis of cell excitation and inhibition of different medicine in different concentration, and is used mainly in medicine screening, analysis and estimation.

The invention relates to a human body upper limb functional rehabilitation training implement method based on muscle tone signals. A piezoelectric sensor is arranged on the surface of muscle skins to collect human body physiological random muscle tone signal monoclonal gamma globulin (MMG), collected human body physiological random muscle tone signals contain muscle contraction motion signals which are amplified and filtered by a preprocessing circuit, analog digital conversion is performed on the muscle contraction motion signals through an analog/digital conversion circuit, then the muscle contraction motion signals are led into a personal computer (PC) machine to perform characteristic extraction and pattern recognition to output control signals on behalf of different motion modes to control three-dimensional human body upper limb model motion in virtual environment, subjects can perform different model view controller (MVC) % motion modes in real time on human body upper limb models according to view feedback in the virtual environment, and finally the purpose of functional rehabilitation training is achieved. The human body upper limb functional rehabilitation training implement method is high in accuracy and stability, can achieve the purpose of industrialization and can widen scope of applicable persons.

The invention provides a phishing website detection method and system based on an adaptive heterogeneous multi-classification model. The method is characterized by for a multiple-base classification algorithm, through linear addition, constructing the adaptive heterogeneous multi-classification model; training the multi-classification model, wherein a model input is the input of each base classification algorithm and an output is a sample label, and each base classification algorithm extracts a corresponding characteristic from a sample record and is taken as the input; and using a machine learning algorithm to solve a model parameter, adopting a test set to test and optimize, and finally acquiring the detection model of the type of a phishing website. The system comprises a domain name morpheme characteristic classifier, a subject index characteristic classifier, a content similarity characteristic classifier, a structural style characteristic classifier, a visual rule characteristicclassifier, a linear addition training module, an integrated classifier, a training data set management module, and a detection and alarm module. In the invention, the phishing website can be detectedin real time, and the accuracy and the stability of phishing website detection are increased.

The invention relates to a reference-free damage identification method for a composite material plate-shaped structure, belonging to the field of structure health monitoring. The method comprises the following steps of: uniformly distributing a sensor array formed by piezoelectric elements on the composite material plate-shaped structure; sequentially exciting each piezoelectric element through a low-frequency and wide-wave signal and collecting corresponding signals of the residual piezoelectric elements when each piezoelectric element is excited; then calculating a damage index on each exciting sensing path according to a signal difference on a similar exciting sensing path; and finally, imaging damages by utilizing a possibility imaging method. According to the reference-free damage identification method disclosed by the invention, online damage identification and imaging of the composite material plate-shaped structure can be conveniently realized, so as to guarantee the safety of the composite material structure in a use process; and the accuracy and the stability are high.

The invention relates to a fashion garment image segmentation method based on depth learning. The fashion garment segmentation method based on depth learning comprises the following steps of the construction of a depth neural network garment model, the loss function design of reverse error propagation and a model training strategy, wherein the depth neural network garment model comprises a featureextraction module, a garment semantic information extraction module and a garment segmentation prediction module, the loss functions comprise a regression function of a key point position, a key visibility loss function, a cross entropy loss function of a garment prediction category with weights and a regression loss function of a garment position; and the model training strategy comprises a weight parameter initialization method, the data preprocessing, an optimization algorithm and a training step. The method has the advantages of being able to automatically segment and recognize the upperbody clothing, lower body clothing and whole body clothing collocation in complex images, and being conducive to the deep learning and network training for fashion clothing design.

The invention discloses a spam filtering method and system based on deep learning. The spam filtering method based on deep learning comprises the first step of processing mail samples to generate a first vector space model and establishing a deep belief network, the second step of processing test mails to generate a second vector space model, the third step of detecting the second vector space model through the established deep belief network, and the fourth step of outputting a detection result. According to the spam filtering method based on deep learning, as the mode that the deep belief network is established and the test mails are detected through the established belief network is adopted, the accuracy and stability of spam recognition are improved, and the time and labor for labeling a large number of samples are saved.

A control system which is capable of enhancing both the stability and the accuracy of control when the output of a controlled object is feedback-controlled by a plurality of control inputs. An ECU 2 of a control system 1 controls engine speed NE during idling by an ignition control input Usl_ig and an intake control input Usl_ar. The ECU 2 calculates a target engine speed NE_cmd according to an engine coolant temperature TW and the like (step 3), and determines the ignition control input Usl_ig and the intake control input Usl_ar with a plurality of predetermined target value filter-type two-degree-of-freedom sliding mode control algorithms [equations (1) to (12)] sharing one switching function σne therebetween, such that the engine speed NE converges to the target engine speed NE_cmd (steps 4 to 7 and 9).

The invention discloses a power battery pack anti-reburning automatic fire extinguishing device and method. The power battery pack anti-reburning automatic fire extinguishing device is characterized in that an outer shell of the fire extinguishing device is divided into an inner shell body and an outer shell body, wherein a fire extinguishing module is arranged in a gap at the top of the part between the inner shell body and the outer shell body, the fire extinguishing module comprises a flame sensor and a fire extinguisher spray head, and the outer end of the fire extinguishing module is connected with the two high-pressure fire extinguishing agent tanks which are correspondingly provided with a compression cooling fire extinguishing agent and a flame retardant extinguishing agent. When the flame sensor of the fire extinguishing module detects a flame signal in the area, the flame signal is transmitted to a controller, the controller issues an instruction to sequentially open a valveof the flame-retardant fire extinguisher tank and a valve of the cooling fire extinguisher tank at intervals, and the fire extinguisher spray head in the fire extinguishing module of the fire area isstarted at the same time, so that the purposes of early rapid fire extinguishing and continuous cooling and extinguishing are achieved, and further diffusion of a fire accident of a battery pack is effectively controlled; and the shell and the battery pack are independent, so that the shell is not limited by the type of batteries, and the application range is wide.

The invention discloses an element recognition method. The method includes following steps: a to-be-matched target picture and at least two template pictures of a to-be-recognized element are read; template matching of the at least two template pictures and the to-be-matched target picture is performed, and at least two corresponding groups of matching values are obtained; operation of the at least two groups of matching values is performed according to a preset calculation rule, and a group of matching results are generated; and each matching result in the group of matching results is compared with a preset threshold, and the position of the to-be-recognized element on the target picture is recognized according to the relation of the matching results and the threshold. The invention also discloses an element recognition apparatus. According to the element recognition method and the apparatus, high recognition accuracy is guaranteed when the target picture is stained, incomplete, or deformed.

The invention provides a gas sensor probe based on a multi-point reflected spiral light path and a detecting device. A light source module, a spiral light path module and a detector module are included. The light source module is used for emitting parallel light signals. The spiral light path module comprises a plurality of reflection mirrors and a plurality of linear light paths. Each linear light path and the adjacent linear light path are connected in a head to tail mode at a certain angle and are spiral inwardly, and at least one reflection mirror is detachably fixed between the two adjacent linear light paths. The spiral light path module is used to increase a light path length of the light signals, and the light signals pass through the spiral light path module, then enter into the detector module and are converted into electrical signals. Through the design of the multi-point reflected spiral light path, a volume of a sensor probe absorption pool is reduced and a measurement light path is increased, and simultaneously the stability and the reliability of the light path are greatly improved. Debugging difficulty during production is reduced and production and processing costis decreased.

The present invention discloses a 3D model retrieval method and a 3D model retrieval apparatus based on slow increment features. The method comprises: carrying out slow increment feature extraction to a preprocessed view set by applying a supervised slow increment feature analysis method; acquiring a sorting result of the slow increment features according to the extracted slow increment features, screening the slow increment features according to the sorting result and generating a slow increment feature library of a 3D model; and carrying out retrieval matching on the slow increment feature library of the 3D model by using a nearest neighbor algorithm to acquire and output an object that is similar to a candidate model. The apparatus comprises: an extraction module, an acquisition module, a generation module and a matching and outputting module. According to the method and the apparatus, the feature extraction difficulty of a nonrigid model is reduced, the stability and accuracy of feature extraction are improved, a good condition is provided for the subsequent 3D model retrieval, and the retrieval result is guaranteed to be more efficient and accurate.

The invention relates to a laser feedback interferometer which comprises a microchip laser device, a spectroscope, a first frequency shifter, a second frequency shifter, a converging lens, at least two photoelectric detectors, an electric signal processing system and a data collecting and processing system, wherein the microchip laser device can output at least two parallel light beams in a single longitudinal and single transverse mode simultaneously, and the at least two parallel light beams can be used as measurement light for measuring information of a plurality of degrees of freedom of an objective to be measured.

The invention relates to the environment perception field and especially relates to a water surface target detection and tracking method based on a laser radar. The method comprises the following steps of (1) acquiring point cloud data of a t-th frame sea surface environment through the laser radar; (2) filtering the point cloud data; (3) segmenting the filtered point cloud data to obtain a point cloud cluster; (4) clustering the point cloud cluster to form a target, carrying out threshold screening on the target, and obtaining the target of the t-th frame sea surface environment; (5) repeating the steps (1) to (4) and processing a t+1th frame sea surface environment to obtain the target of the t+1th frame sea surface environment; (6) matching and associating the target of the t-th frame sea surface environmental with the target of the t+1th frame sea surface environmental to obtain a matched target pair; and (7) calculating a target moving state of the matched target pair and tracking the target moving state. By using the method, target detection and tracking accuracy is high, stability is good, and the accuracy and stability of target detection and tracking of an unmanned surface vehicle are increased.

The invention discloses a machine learning-based vehicle abnormal trajectory real-time recognition method and belongs to the field of vehicle trajectory anomaly recognition. The method includes the following steps that: collected data are cleaned, so that complete, non-repetitive, abnormal value-free training data are obtained; an unsupervised isolated forest method and training data are used to perform model training, so that an anomaly detection model is obtained; the anomaly detection model is put into a flow calculation engine for real-time prediction, and a prediction result is sent to avehicle owner; and the model is automatically updated and corrected according to the feedback information of the vehicle owner, and the updated model is put into the flow calculation engine for real-time prediction, and a prediction result is sent to the vehicle owner. According to the method of the invention, the vehicle information is periodically collected; the unsupervised isolated forest algorithm is adopted; real-time prediction analysis is performed on vehicle trajectories in the flow calculation engine; the probability value of the abnormal behavior of a vehicle is rendered; the modelis periodically adjusted according to the feedback data given by the vehicle user; the dynamic update of the model is realized; and the recognition accuracy of the model is improved.

The invention relates to monitoring method and system for microblogging public opinion. The system comprises a data acquiring die for crawling real-time microblog data and writing into a database, a hot microblog detecting module for sequencing the real-time microblog data within the preset time according to the preset algorithm and screening the hot microblogs within the specific time quantum, an abnormal microblog monitoring device for vectoring microblog texts, clustering the vectored microblog texts and selecting the microblogs with the highest hot level from various microblogs to be used as the monitored abnormal microblogs, and a data display module for displaying the hot microblogs and the abnormal microblogs with corresponding quantity according to the preset threshold. With the adoption of the method and system, the microblogs within a period of time can be quickly analyzed, the hot microblogs can be extracted, and the hot microblog topic can be tracked; the real-time microblog within a period of time can be monitored, and the abnormal microblog within the latest period of time can be detected; the data can be automatically cleaned by the system; the mature clustering method is used as the algorithm to ensure relatively high stability and accuracy.

The invention discloses an electrified cleaning insulating maintenance robot device for high-voltage power equipment. The electrified cleaning insulating maintenance robot device comprises a traveling mechanism, a lifting mechanism and a cleaning insulating maintenance mechanical arm. The lifting mechanism comprises a motor linkage driving circuit, a motor, a balance monitoring device and two or more pneumatic insulating lifting rods. The motor linkage driving circuit is in signal connection with the motor. The balance monitoring device is in signal connection with the motor linkage driving circuit. The pneumatic insulating lifting rods are made of epoxy resin or glass fiber steel and installed on the traveling mechanism. The cleaning insulating maintenance mechanical arm is arranged on the lifting mechanism. According to the electrified cleaning insulating maintenance robot device for the high-voltage power equipment, the supporting stability is improved, the driving force of the lifting mechanism is reduced, and the phenomenon of creepage short circuits is effectively avoided; the degree of balance of the lifting mechanism can be adjusted automatically, the probability of tilting and turning is lowered, and the safety performance is improved.

The invention discloses a xerophthalmia detection lighting and imaging system, comprising: a Placido disc that includes a plurality of concentric rings in black and white; a lighting optical system that is arranged on the rear surface of the Placido disc and that includes a plurality of groups of LED units arranged in circular form, wherein the lighting optical system is used to project the Placido disc to tear film; an imaging optical system that receives a Placido disc projection on the tear film and is along an imaging optical path, wherein the imaging optical system comprises an imaging objective and a CCD (charge coupled device) camera in sequence. The xerophthalmia detection lighting and imaging system has the advantages that a series of pictures of the eye surface tea film of a subject from intactness to break can be acquired in noninvasive manner, break time of the tea film can be acquired by post-processing through an image algorithm, and whether the break time of the tear film of the subject is abnormal or not can be judged; the system according to the technical scheme can ease the pain of a patient, and detection results are more stable and accurate.

The revolution automatic control gas measurement device comprises the light, wave plate, measuring gas chamber, filter plate, sensor, drive motor, processing unit, drive signal generation unit, pulse signal generation unit and drive voltage control module, with the pulse signal generation unit generating pulse signal with the frequency in direct proportion with the motor revolution or wave plate revolution, based on the frequency outputting control signal of the frequency to the drive signal generation unit, with the drive signal generating unit modulating PWM wave based on the control signal. Through timely monitor and modulation to the motor or wave plate revolution, it can remove factors due to the motor or wave plate, ensure the wave plate having consistent frequency to modulate the light source. Meanwhile, it can ensure the accuracy of the testing result.

The invention discloses a distributed filtering estimation method for improving gas path performance of a turbofan engine. The method comprises: firstly establishing a nonlinear component-level general model of a state above an idler of the turbofan engine, and designing a distributed extended Kalman filter to improve the local filtering algorithm of a sub-filter and improve the stability and accuracy of the filtering algorithm for the phenomenon of the packet loss of a sensor signal; then designing an adaptive algorithm by using a variance matching method in a main filter to improve the response speed to a sudden fault; and finally integrating the nonlinear calculation content comprising model linearization prior estimation and prior measurement in the sub-filter into the main filter, sothat the sub-filter only performs a linear calculation, and the main filter performs all the nonlinear calculations. According to the distributed filtering estimation method for improving the gas pathperformance of the turbofan engine, the problem of the decrease in the estimation accuracy caused by data the packet loss in an existing distributed gas path performance estimation of an aircraft engine in an existing envelope is solved, and the problem that the sub-filter has an insufficient calculation capability and cannot satisfy the real-time performance is solved.

The invention relates to a pretreatment method for detecting drugs in human hair. The pretreatment method comprises the following steps: S1, taking a hair sample, and cutting the hair sample into small sections with the length of 3-5 mm; S2, putting the hair sample cut in the step S1 into a grinding tube, adding a hair lysate, and crushing by using a hair crushing instrument; S3, taking out the hair sample crushed in the step S2, thereby completing pretreatment for detecting the drugs in the human hair. The hair sample can be directly applied to colloidal gold immunochromatography or fluorescence immunochromatography after being crushed. The crushing is implemented by performing high-speed oscillation in a low-temperature environment, so that the situations of local high temperature and non-uniform sample crushing are avoided; due to the addition of the hair lysate, small drug molecules in the hair of drug addicts can be well released. Compared with the conventional hair lysate, the pretreatment method has the advantages that the release rate of the small drug molecules is increased by 5-6 times, the cracking efficiency is greatly improved, and a novel treatment method is providedfor hair detection.

The invention discloses a planet wheel fault recognition method based on sensitive measure point data fusion, which includes: selecting characteristic parameters strong in applicability of planetary gear box fault recognition, training a sorting algorithm based on an adaptive neural-fuzzy inference system (ANFIS) by aid of historical data, and adopting the trained sorting algorithm to achieve automatic sorting recognition of planetary gear box faults on the basis of vibration data measured at a plurality of measure points. The planet wheel fault recognition method has the advantages of fusing information of the plurality of measure points, performing fault sorting by aid of the sorting algorithm based on the ANFIS and achieving accurate recognition of planetary gear box faults, solves the problem that different measure point information is inconsistent in fault sensitivity degree, and remarkably improves accuracy and stability of fault recognition.

The invention discloses a geomagnetic parking space detection method. The method includes the steps that S1, geomagnetic data is collected; S2, the geomagnetic variation Ts is calculated; S3, the zi-axis fluctuating quantity delta theta i is calculated; S4, whether the Ts is larger than the minimum noise threshold Tn or not is judged, and if yes, S5 is executed; S5, whether the delta theta i is larger than the z-axis fluctuation threshold value theta n or not is judged, and if yes, S6 is executed; S6, whether the Ts is larger than the vehicle reference quantity Th or not is judged, and if yes,S7 is executed; S7, whether a vehicle ruler is recorded or not is judged, if not, S8 is executed, and if yes, S9 is executed; S8, whether the Ts is smaller than the Th or not is judged, and if yes, S10 is executed; S9, whether the Ts is larger than the Tn or not is judged, and if yes, S11 is executed; S10, whether the z-axis change direction is opposite to the that when a vehicle exists, and if yes, the vehicle leaves; S11, whether the delta theta i is larger than the theta n or not is judged, and if yes, S12 is executed; S12, whether the Ts is smaller than the Th or not is judged, if yes, S10 is executed, and a vehicle-free ruler is updated and recorded. According to the method, on the basis of setting threshold value comparison, entering into and exiting from a parking space of the vehicle are judged by detecting the fluctuation angle of the z axis, so that the stability and accuracy of detection results can be improved.

The invention belongs to the field of bioengineering, and aims to provide a method for detecting fungaltoxin by using a quantum dot cation exchange signal amplification technology. The method includes the following steps: using carboxyl-modified water-soluble cadmium selenide/zinc sulfide quantum dots to mark monoclonal antibodies of fungaltoxin, and enabling carboxyl on the quantum dots and amidogen of the antibodies to form stable fungaltoxin monoclonal antibodies marked by quantum dots under the action of a coupling agent; then detecting fungaltoxin in competitive ELISA mode, and finally analyzing fluorescence signals after reaction by adopting the cation exchange signal amplification technology, so as to realize rapid qualitative detection and quantitative analysis on fungaltoxin. According to the method provided by the invention, the detection time is shortened while the detection sensitivity and stability are improved; the detection of fungaltoxin only needs 45 minutes; the detection sensitivity for the to-be-detected fungaltoxin can reach a picogram level; the quantum dots used in the invention are convenient to synthesize and low in cost; the method has the advantages of being efficient, fast, high in sensitivity, stable in detecting signals, and the like.

The invention provides a community discovery method based on node intimacy and label propagation. The method belongs to the field of complex network analysis. The main technical feature of the methodcomprises the steps of estimating intimacy among adjacent nodes in a network through utilization of local topology information, and establishing an intimacy matrix based on the intimacy; estimating importance of the nodes in the network based on the intimacy matrix, and ensuring that the nodes are updated based on the importance in a descending order; in an iterative updating process, updating a label of each node as the label with the greatest influence in neighbors, and determining the unique label through calculation of the intimacy of the labels if a plurality of labels with the greatest influence exist; and when an algorithm satisfies a stop condition, jumping out of a label propagation process, and subjecting the nodes wit the same label to the same community. The method is reasonable in design; a high quality community can be rapidly detected; the robustness of the label propagation method can be effectively improved; and the method can be widely applied to fields such as protein function prediction, epidemic surveillance and electronic commerce precise recommendation.

The invention relates to an iris location method. The method comprises the following steps of: step 1, detecting spots in an iris image; step 2, judging whether the iris image shakes; step 3, initially locating a pupil; step 4, locating the iris. The method has the following benefits that: the iris is located by a two-dimensional round Gabor filter and a weighted calculus detection operator, interference on an iris identification system caused by noise information is reduced, internal and external boundaries can be quickly and accurately located in the iris image, and the stability and the accuracy of an iris location algorithm are improved.

The application is applicable to the field of computer application technology, and provides a doctor recommendation method and device. The doctor recommendation method includes: obtaining patient information and doctor information; extracting patient features from the patient information and doctor features from the doctor information; constructing a fitness function according to the patient features and doctor features, and determining the feature weight corresponding to each doctor feature according to the fitness function; and determining the doctor information that most closely matches thepatient information according to the feature weight corresponding to the doctor features. An optimal weight combination is searched with the objective of optimizing the accuracy of the final recommendation result, then a method of weighting is used to combine the feature similarity calculation results in the collaborative filtering algorithm, and the doctor recommendation ranking result that bestmatches the patient is output. The multi-objective feature selection model improves the accuracy and stability of individual identification and classification, also improves the matching degree between doctors and patients, and achieves personalized and targeted doctor recommendation.