An assisted mining method for early medical information of respiratory infectious diseases

Abstract

The present invention relates to the technical field of medical care informatics, and in particular to an early medical information auxiliary mining method for respiratory infectious diseases, comprising: the present invention establishes a community by analyzing the feature vector and time information of the patient's medical record using cosine similarity. The frequency, slope and periodicity of the feature vocabulary are calculated to evaluate its mainstream degree in the community. In combination with the timeliness and mainstream degree, the reference of the community feature vector is determined, and data mining is performed accordingly. The present invention improves the accuracy of the data mining results.

Description

Technical Field

[0001] The invention relates to the technical field of medical care informatics, and in particular to an early medical information auxiliary mining method for respiratory infectious diseases. Background Art

[0002] Diseases have different modes of transmission, incubation periods, and clinical treatment methods depending on the pathogens and clinical manifestations. By mining early information, we can detect and warn of the outbreak and epidemic trend of respiratory infectious diseases in advance, which will help to take early intervention and control measures, thereby reducing the spread and impact of the disease. However, existing data mining algorithms rely on static data snapshots when modeling, and assume that data distribution and characteristics will not change significantly. In actual situations, pathogens will continue to evolve to adapt to the environment, and their early respiratory infectious disease characteristics may also change. In addition, existing data mining algorithms are often limited to basic data aggregation and statistical analysis when mining, and lack deeper correlation analysis between data. This may lead to a lack of sufficient accuracy in the process of medical diagnosis. Summary of the invention

[0003] The present invention provides an early medical information auxiliary mining method for respiratory infectious diseases to solve the existing problems.

[0004] The present invention provides an early medical information auxiliary mining method for respiratory infectious diseases using the following technical solutions:

[0005] An embodiment of the present invention provides an early medical information auxiliary mining method for respiratory infectious diseases, the method comprising the following steps:

[0006] Obtaining the registration time and feature vector of each patient's medical record information, wherein the feature vector includes multiple disease information;

[0007] All communities are obtained according to the cosine similarity of any two feature vectors; the communities contain all the disease information of a number of patients; according to the frequency of occurrence of each disease information in all the disease information contained in each community, characteristic vocabulary information is screened from all the disease information in each community to form a target feature vector for each community;

[0008] According to the registration time of each patient's medical record information, the timeliness of each community's target feature vector is obtained; according to each community's target feature vector and the preset threshold range, the occurrence frequency and the first slope of the feature vocabulary information in each quarter are obtained;

[0009] According to the preset threshold range and the frequency of occurrence of the characteristic vocabulary information in each quarter, whether the frequency spectrum of each characteristic vocabulary information is periodic is obtained; according to the frequency of occurrence of each characteristic vocabulary information in each quarter, the instantaneous slope of each characteristic vocabulary information at the current moment is obtained;

[0010] According to the preset threshold range, the frequency of occurrence of each characteristic vocabulary information in each quarter, the first slope, the instantaneous slope at the current moment, the frequency of occurrence in each community, and whether the spectrum graph has periodicity, the mainstream degree of each characteristic vocabulary information is obtained; according to the mainstream degree of all characteristic vocabulary information in each community, the mainstream degree of each community target feature vector is obtained; according to the timeliness of each community target feature vector and the mainstream degree of each community target feature vector, the reference of each community target feature vector is obtained;

[0011] Data mining is performed based on the reference of each community target feature vector and the mining results are output.

[0012] Furthermore, the specific steps of obtaining all communities according to the cosine similarity of any two feature vectors are as follows:

[0013] Among all the patients’ feature vectors, calculate the cosine similarity between any two patients’ feature vectors;

[0014] Taking the inverse of the cosine similarity of any two feature vectors as the distance, the feature vectors of all patients are clustered to obtain several clusters;

[0015] The clusters are sorted according to the number of eigenvectors contained in the clusters, and named the first community, the second community, and so on in order from large to small.

[0016] Furthermore, the method of selecting characteristic vocabulary information from all the disease information in each community to form a target feature vector for each community according to the frequency of occurrence of each disease information in all the disease information contained in each community includes the following specific steps:

[0017] Obtain all disease information contained in each community, count the frequency of occurrence of each disease information in each community, and record the disease information whose occurrence frequency is greater than a first preset threshold as characteristic vocabulary information;

[0018] All the characteristic vocabulary information of each community is formed into a vector and recorded as the target characteristic vector of each community.

[0019] Furthermore, the timeliness of each community target feature vector is obtained according to the registration time of each patient's medical record information, and the specific steps include the following:

[0020] The time when each patient's medical record information is registered is used as the time corresponding to each patient's feature vector;

[0021] Calculate the mean of the time interval between the corresponding time of all feature vectors in each community and the current time, and record it as the timeliness of the target feature vector of each community.

[0022] Furthermore, the acquisition of the occurrence frequency and the first slope of the characteristic vocabulary information in each quarter according to the target characteristic vector of each community and the preset threshold range includes the following specific steps:

[0023] Obtain the frequency of each characteristic word information of each community target feature vector in all the disease information of all patients in each quarter within one year as the frequency of each characteristic word information in each quarter;

[0024] According to the frequency of occurrence of each characteristic word information of each community target feature vector in each quarter, the slope of the frequency-time scatter plot fitting line of each characteristic word information is obtained, wherein the abscissa of the frequency-time scatter plot is the quarter and the ordinate is the frequency;

[0025] When the slope of any characteristic vocabulary information fitting straight line is not within a preset threshold range, the slope of any characteristic vocabulary information fitting straight line is recorded as a first slope.

[0026] Furthermore, the specific steps of obtaining whether the frequency spectrum of each characteristic vocabulary information has periodicity according to the preset threshold range and the occurrence frequency of the characteristic vocabulary information in each quarter include the following:

[0027] When the slope of the fitting straight line of each characteristic vocabulary information is within the preset threshold range, the frequency-time scatter plot curve of each characteristic vocabulary information is fitted and then Fourier transformed to obtain the spectrum diagram of each characteristic vocabulary information, with the horizontal axis being the frequency and the vertical axis being the amplitude, and the judgment result of whether the spectrum diagram has periodicity is obtained.

[0028] Furthermore, the instantaneous slope of each characteristic word information at the current moment is obtained according to the occurrence frequency of each characteristic word information in each quarter, including the following specific steps:

[0029] The difference between the frequency of the vth characteristic vocabulary information in the current quarter and the frequency of the vth characteristic vocabulary information in the quarter before the current quarter is recorded as the instantaneous slope of each characteristic vocabulary information at the current moment.

[0030] Furthermore, the mainstream degree of each characteristic word information is obtained according to the preset threshold range, the frequency of occurrence of each characteristic word information in each quarter, the first slope, the instantaneous slope at the current moment, the frequency of occurrence in each community and whether the spectrum graph has periodicity, including the following specific steps:

[0031] When the slope of the fitting line of each characteristic vocabulary information is not within the preset threshold range, the mainstream degree is the normalized value of the product of the first slope of each characteristic vocabulary information and the occurrence frequency of each characteristic vocabulary information in the quarter before the current quarter;

[0032] When the slope of the fitting line of each characteristic vocabulary information is within the preset threshold range and has periodicity, the mainstream degree is the normalized value of the product of the instantaneous slope of each characteristic vocabulary information at the current moment and the time interval from the current moment to the initial moment of the previous quarter of the quarter where the current moment is located;

[0033] When the slope of the fitting line of each characteristic vocabulary information is within the preset threshold range and has no periodicity, the mainstream degree is the average of the frequency of occurrence of each characteristic vocabulary information in all communities.

[0034] Furthermore, the step of obtaining the mainstream degree of the target feature vector of each community according to the mainstream degree of all feature vocabulary information in each community includes the following specific steps:

[0035] The average of the mainstream degrees of all feature vocabulary information in each community is recorded as the mainstream degree of the target feature vector of each community.

[0036] Furthermore, the step of obtaining the reference of each community target feature vector according to the timeliness of each community target feature vector and the mainstream degree of each community target feature vector includes the following specific steps:

[0037] The normalized value of the product of the timeliness of each community target feature vector and the mainstream degree of each community target feature vector is recorded as the reference of each community target feature vector.

[0038] The beneficial effect of the technical solution of the present invention is that the present invention can evaluate the prevalence of each sample and its characteristics in the current environment by analyzing the individuals in the data sample. This method can pay more attention to the current potential situation in the data mining process, so that the analysis results are more in line with the actual characteristics of the disease, thereby improving the accuracy and timeliness of the data mining results. In addition, in the process of data mining, the maximum number of repetitions of data mining can be set according to different situations, further improving the accuracy of the data mining results. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 The present invention is a flowchart of the steps of an early medical information auxiliary mining method for respiratory infectious diseases. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of an early medical information auxiliary mining method for respiratory infectious diseases proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The specific scheme of the early medical information auxiliary mining method for respiratory infectious diseases provided by the present invention is described in detail below with reference to the accompanying drawings.

[0044] See also Figure 1 , which shows a flowchart of a method for assisting in mining early medical information of respiratory infectious diseases provided by an embodiment of the present invention, the method comprising the following steps:

[0045] Step S001: Obtain the registration time and feature vector of each patient's medical record information, wherein the feature vector includes multiple symptom information.

[0046] Get the feature vector of each patient. A feature vector contains multiple disease information.

[0047] This embodiment obtains the medical record information of all patients, processes the medical record information using a natural language algorithm, extracts the symptom information representing the condition in the medical record information, and constructs it into a feature vector for each patient.

[0048] The registration time of each patient's medical record information is obtained, and the time is used as the time corresponding to each patient's feature vector.

[0049] It should be noted that in this implementation, medical record information within one year is obtained.

[0050] Step S002: Obtain all communities based on the cosine similarity of any two feature vectors; the communities contain all the disease information of several patients; based on the frequency of occurrence of each disease information in all the disease information contained in each community, filter out feature vocabulary information from all the disease information in each community to form a target feature vector for each community.

[0051] Early symptoms of respiratory infections may present with different symptom combinations and severity in different age groups or individuals. Early symptoms can be distinguished by dividing samples with similar symptom combinations into homogeneous communities, one of which is considered to represent an early symptom.

[0052] Specifically: Among the feature vectors of all patients, the cosine similarity of the feature vectors of any two patients is calculated.

[0053] The reciprocal of the cosine similarity of any two feature vectors is used as the distance to cluster the feature vectors of all patients to obtain several clusters. The clusters are sorted according to the number of feature vectors contained in the clusters, and are named the first community, the second community, and so on in order from large to small. In this embodiment, the clustering is K-Means, which is taken as an example.

[0054] All disease information contained in each community is obtained, the frequency of occurrence of each disease information in each community is counted, and the disease information whose frequency of occurrence is greater than a first preset threshold is recorded as characteristic vocabulary information.

[0055] All the characteristic vocabulary information of each community is formed into a vector and recorded as the target characteristic vector of each community.

[0056] In this embodiment, the first preset threshold is 0.8, which is taken as an example.

[0057] Step S003: According to the registration time of each patient's medical record information, the timeliness of each community's target feature vector is obtained; according to each community's target feature vector and a preset threshold range, the occurrence frequency and the first slope of the feature vocabulary information in each quarter are obtained.

[0058] With the influence of various external factors such as seasonal changes, pathogen evolution and mutation, respiratory infections may show different symptoms at different times. The purpose of data mining is to obtain various data that are more in line with the characteristics of the current disease. Therefore, the time corresponding to the data is particularly important in the process of data mining. By conducting time series analysis on each disease community, different community symptoms can be given different references.

[0059] The time interval between the corresponding time of all feature vectors in each community and the current time is calculated, and the mean of the time interval is further obtained, and the mean is recorded as the timeliness of each community target feature vector. It should be noted that the larger the mean, the worse the timeliness of the community target feature vector.

[0060] After the timeliness of the disease feature vectors corresponding to the community is obtained, a time series analysis is performed on the vocabulary features contained in the vector to determine the current mainstream degree of different disease feature vectors. It should be noted that the analysis process for vocabulary is based on the medical records of all patients.

[0061] The frequency of occurrence of each characteristic word information of each community target feature vector in each quarter is obtained, that is, the frequency of occurrence of the characteristic word information in all symptom information of all patients in each quarter of the year. The higher the frequency, the more mainstream the symptom is in this time period.

[0062] According to the frequency of occurrence of each characteristic word information of each community target feature vector in each quarter, a frequency-time scatter plot of each characteristic word information is constructed, with the horizontal axis representing the quarter and the vertical axis representing the frequency.

[0063] The frequency-time scatter plot of each characteristic word information is fitted with a straight line using the least square method, and the slope of the fitted straight line is obtained. The least square method is a well-known technique.

[0064] When the slope of any characteristic vocabulary information fitting straight line is not within a preset threshold range, the slope of the any characteristic vocabulary information fitting straight line is recorded as the first slope.

[0065] When the slope of the fitted straight line is greater than the maximum value of the preset threshold range, it indicates that the frequency of occurrence of the disease is increasing overall. When the slope of the fitted straight line is less than the minimum value of the preset threshold range, it indicates that the frequency of occurrence of the disease is decreasing overall.

[0066] In this embodiment, the preset threshold range is [-0.1, 0.1], which is taken as an example.

[0067] It should be further explained that the slope of the fitting line of some characteristic vocabulary information is close to 0, that is, the frequency of occurrence of the characteristic vocabulary information is consistent in different time periods, or its frequency of occurrence changes periodically. When the frequency of occurrence of symptom vocabulary is consistent in different time periods, its mainstream degree is always high; when the frequency of occurrence of symptom vocabulary shows periodic changes, it is also necessary to combine the recent frequency of occurrence of the symptom and its periodic characteristics for analysis.

[0068] Step S004: According to the preset threshold range and the frequency of occurrence of the characteristic vocabulary information in each quarter, obtain whether the frequency spectrum of each characteristic vocabulary information has periodicity; according to the frequency of occurrence of each characteristic vocabulary information in each quarter, obtain the instantaneous slope of each characteristic vocabulary information at the current moment.

[0069] When the slope of the fitting straight line of each characteristic vocabulary information is within the preset threshold range, the frequency-time scatter plot curve of each characteristic vocabulary information is fitted and then Fourier transformed to obtain the spectrum of each characteristic vocabulary information, with the horizontal axis being the frequency and the vertical axis being the amplitude, and a judgment result of whether the spectrum has periodicity is obtained. Among them, Fourier transform is a well-known technology.

[0070] It should be noted that in the spectrum graph, when the distance between the horizontal coordinates corresponding to each amplitude is equal, the spectrum graph is considered to be periodic, otherwise it is non-periodic. This is because if the frequency components are evenly distributed, the distance between the peaks is roughly equal, which may indicate that the signal contains a regular harmonic structure, which reflects that the signal is periodic, otherwise it is non-periodic.

[0071] The difference between the frequency of the vth characteristic vocabulary information in the current quarter and the frequency of the vth characteristic vocabulary information in the quarter before the current quarter is recorded as the instantaneous slope of each characteristic vocabulary information at the current moment.

[0072] It should be noted that the larger the time interval between the current moment and the moment before the current moment, the more mainstream the characteristic vocabulary information is in the recent period, according to this embodiment.

[0073] Step S005: Obtain the mainstream degree of each characteristic vocabulary information according to the preset threshold range, the frequency of occurrence of each characteristic vocabulary information in each quarter, the first slope, the instantaneous slope at the current moment, the frequency of occurrence in each community, and whether the spectrum diagram has periodicity; obtain the mainstream degree of each community target feature vector according to the mainstream degree of all characteristic vocabulary information in each community; obtain the reference of each community target feature vector according to the timeliness of each community target feature vector and the mainstream degree of each community target feature vector.

[0074] Calculate the degree of mainstreaming of each characteristic word information in the current moment and the previous quarter of the current moment.

[0075] When the slope of the fitting line of each characteristic vocabulary information is not within the preset threshold range, the mainstream degree is the normalized value of the product of the first slope of each characteristic vocabulary information and the occurrence frequency of each characteristic vocabulary information in the quarter before the current quarter.

[0076] When the slope of the fitting line of each characteristic vocabulary information is within the preset threshold range and has periodicity, the mainstream degree is the normalized value of the product of the instantaneous slope of each characteristic vocabulary information at the current moment and the time interval from the current moment to the initial moment of the previous quarter of the quarter where the current moment is located.

[0077] When the slope of the fitted straight line of each characteristic word information is within the preset threshold range and does not have periodicity, the mainstream degree is the average of the frequency of occurrence of each characteristic word information in all communities. It should be noted that communities where the word does not appear are not included in the calculation of the average.

[0078] The average of the mainstream degrees of all feature vocabulary information in each community is recorded as the mainstream degree of the target feature vector of each community.

[0079] The normalized value of the product of the timeliness of each community target feature vector and the mainstream degree of each community target feature vector is recorded as the reference of each community target feature vector. In this embodiment, the norm linear normalization function is used for normalization.

[0080] Step S006: Perform data mining based on the reference of each community target feature vector and output the mining results.

[0081] The reference of each community target feature vector, that is, certain combinations of symptoms in the current time period are more mainstream (easy to occur), and in the process of data mining, more attention is paid to the relevant information of this part of the features, so that the final data mining results can be more in line with the current environmental status, thereby improving the authenticity and accuracy of data mining.

[0082] Among all the patient feature vectors, the reference of each community target feature vector is used as the mining weight, and the FP-Growth data mining algorithm is used to mine all community target feature vectors and output the mining results.

[0083] It should be noted that after one round of data mining, it is still necessary to continuously collect early medical information on respiratory infections and set a maximum update time for the data mining results, such as once a quarter, to ensure the timeliness of the data mining results and to enable them to adapt to and obtain the latest early disease characteristics, thereby increasing the probability of early detection and achieving the purpose of early detection and early prevention.

[0084] So far, the present invention is completed.

[0085] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An early medical information auxiliary mining method for respiratory infectious diseases, characterized in that: The method comprises the following steps: Obtaining the registration time and feature vector of each patient's medical record information, wherein the feature vector includes multiple disease information; All communities are obtained according to the cosine similarity of any two feature vectors; the communities contain all the disease information of a number of patients; according to the frequency of occurrence of each disease information in all the disease information contained in each community, characteristic vocabulary information is screened from all the disease information in each community to form a target feature vector for each community; According to the registration time of each patient's medical record information, the timeliness of each community's target feature vector is obtained; according to each community's target feature vector and the preset threshold range, the occurrence frequency and the first slope of the feature vocabulary information in each quarter are obtained; According to the preset threshold range and the frequency of occurrence of the characteristic vocabulary information in each quarter, whether the frequency spectrum of each characteristic vocabulary information is periodic is obtained; according to the frequency of occurrence of each characteristic vocabulary information in each quarter, the instantaneous slope of each characteristic vocabulary information at the current moment is obtained; According to the preset threshold range, the frequency of occurrence of each characteristic vocabulary information in each quarter, the first slope, the instantaneous slope at the current moment, the frequency of occurrence in each community, and whether the spectrum graph has periodicity, the mainstream degree of each characteristic vocabulary information is obtained; according to the mainstream degree of all characteristic vocabulary information in each community, the mainstream degree of each community target feature vector is obtained; according to the timeliness of each community target feature vector and the mainstream degree of each community target feature vector, the reference of each community target feature vector is obtained; Perform data mining based on the reference of each community target feature vector and output the mining results; The method for obtaining the occurrence frequency and the first slope of the characteristic word information in each quarter is as follows: Obtain the frequency of each characteristic word information of each community target feature vector in all the disease information of all patients in each quarter within one year as the frequency of each characteristic word information in each quarter; According to the frequency of occurrence of each characteristic word information of each community target feature vector in each quarter, the slope of the frequency-time scatter plot fitting line of each characteristic word information is obtained, wherein the abscissa of the frequency-time scatter plot is the quarter and the ordinate is the frequency; When the slope of any one of the characteristic vocabulary information fitting lines is not within a preset threshold range, the slope of any one of the characteristic vocabulary information fitting lines is recorded as a first slope; The method for obtaining the instantaneous slope of each characteristic word information at the current moment is: The difference between the frequency of the vth characteristic vocabulary information in the current quarter and the frequency of the vth characteristic vocabulary information in the quarter before the current quarter is recorded as the instantaneous slope of each characteristic vocabulary information at the current moment.

2. According to claim 1, the early medical information auxiliary mining method for respiratory infectious diseases is characterized in that: The specific steps of obtaining all communities according to the cosine similarity of any two feature vectors are as follows: Among all the patients’ feature vectors, calculate the cosine similarity between any two patients’ feature vectors; Taking the inverse of the cosine similarity of any two feature vectors as the distance, the feature vectors of all patients are clustered to obtain several clusters; The clusters are sorted according to the number of eigenvectors contained in the clusters, and named the first community, the second community, and so on in order from large to small.

3. The method for assisting in mining early medical information of respiratory infectious diseases according to claim 1, characterized in that: The specific steps of selecting characteristic vocabulary information from all the disease information in each community according to the frequency of occurrence of each disease information in all the disease information contained in each community to form a target feature vector for each community are as follows: Among all the disease information contained in each community, disease information whose occurrence frequency is greater than a first preset threshold is recorded as characteristic vocabulary information; All the characteristic vocabulary information of each community is formed into a vector and recorded as the target characteristic vector of each community.

4. The method for assisting in mining early medical information of respiratory infectious diseases according to claim 1, characterized in that: The specific steps of obtaining the timeliness of each community target feature vector according to the registration time of each patient's medical record information are as follows: The time when each patient's medical record information is registered is used as the time corresponding to each patient's feature vector; Calculate the mean of the time interval between the corresponding time of all feature vectors in each community and the current time, and record it as the timeliness of the target feature vector of each community.

5. The method for assisting in mining early medical information of respiratory infectious diseases according to claim 1, characterized in that: The specific steps of obtaining whether the frequency spectrum of each characteristic word information has periodicity according to the preset threshold range and the occurrence frequency of the characteristic word information in each quarter include the following: When the slope of the fitting straight line of each characteristic vocabulary information is within the preset threshold range, the frequency-time scatter plot curve of each characteristic vocabulary information is fitted and then Fourier transformed to obtain the spectrum diagram of each characteristic vocabulary information, with the horizontal axis being the frequency and the vertical axis being the amplitude, and the judgment result of whether the spectrum diagram has periodicity is obtained.

6. The method for assisting in mining early medical information of respiratory infectious diseases according to claim 5, characterized in that: The method of obtaining the mainstream degree of each characteristic word information according to the preset threshold range, the occurrence frequency of each characteristic word information in each quarter, the first slope, the instantaneous slope at the current moment, the occurrence frequency in each community and whether the spectrum graph has periodicity includes the following specific steps: When the slope of the fitting line of each characteristic vocabulary information is not within the preset threshold range, the mainstream degree is the normalized value of the product of the first slope of each characteristic vocabulary information and the occurrence frequency of each characteristic vocabulary information in the quarter before the current quarter; When the slope of the fitting line of each characteristic vocabulary information is within the preset threshold range and has periodicity, the mainstream degree is the normalized value of the product of the instantaneous slope of each characteristic vocabulary information at the current moment and the time interval from the current moment to the initial moment of the previous quarter of the quarter where the current moment is located; When the slope of the fitting line of each characteristic vocabulary information is within the preset threshold range and has no periodicity, the mainstream degree is the average of the frequency of occurrence of each characteristic vocabulary information in all communities.

7. The method for assisting in mining early medical information of respiratory infectious diseases according to claim 1, characterized in that: The method of obtaining the mainstream degree of each community target feature vector according to the mainstream degree of all feature vocabulary information in each community includes the following specific steps: The average of the mainstream degrees of all feature vocabulary information in each community is recorded as the mainstream degree of the target feature vector of each community.

8. The method for assisting in mining early medical information of respiratory infectious diseases according to claim 1, characterized in that: The specific steps of obtaining the reference of each community target feature vector according to the timeliness of each community target feature vector and the mainstream degree of each community target feature vector are as follows: The normalized value of the product of the timeliness of each community target feature vector and the mainstream degree of each community target feature vector is recorded as the reference of each community target feature vector.

Images