A medical information recommendation method and platform based on big data analysis technology

Abstract

The present application discloses a method and platform for recommending medical information based on big data analysis technology. The method includes first collecting the current information of each potential doctor, establishing a dictionary, and generating a first potential doctor feature set; then calculating the similarity of each first potential doctor feature in pairs to obtain a first similarity set, and finally, when there is a feature value greater than a first preset threshold in the similarity of each first potential doctor feature, a second potential doctor feature set is generated; and used as a medical information recommendation target. The present invention realizes online digital marketing of medicines based on big data analysis technology, and provides different rehabilitation treatment plans and related drug recommendations for target doctors in the three stages of before, during and after consultation.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a medical information recommendation method and platform based on big data analysis technology. Background Art

[0002] Currently, the pharmaceutical industry is undergoing major business changes and evolutions around the world. With rising cost pressures, new policies and regulations, reforms in medical institutions, and the complexity of the overall medical market, some traditional business models, such as marketing models that rely mainly on face-to-face visits by sales representatives, may have problems with low efficiency. Summary of the invention

[0003] Based on this, the embodiments of the present application provide a medical information recommendation method and platform based on big data analysis technology. The present invention realizes online digital marketing of medicines based on big data analysis technology, and provides different rehabilitation treatment plans and related drug recommendations for target doctors in the three stages before, during and after consultation.

[0004] In a first aspect, a method for recommending medical information based on big data analysis technology is provided, the method comprising:

[0005] Collecting the current information of each potential doctor; the current information includes at least the types of diseases that the potential doctor is good at and the chat records, electronic prescriptions, drug orders, and examination reports of patients before, during, and after the consultation;

[0006] The current information of each potential doctor is used to establish a dictionary to generate a first potential doctor feature set, wherein the first potential doctor feature set includes a plurality of first potential doctor features, each of which represents a dictionary established by the current information of a potential doctor;

[0007] Calculate the similarity of each first potential doctor feature in pairs to obtain a first similarity set; wherein the first similarity set includes multiple first potential doctor feature similarities, and each similarity data in the first potential doctor feature similarities is used as a feature value;

[0008] When there is a feature value greater than a first preset threshold value in the similarities of the features of each first potential doctor, the first similarity doctor feature is used as the second potential doctor feature to generate a second potential doctor feature set; and the potential doctor corresponding to the second potential doctor feature set is used as a medical information recommendation target.

[0009] Optionally, the method further comprises:

[0010] It is determined whether there is a feature value greater than a first preset threshold value among the similarities of the features of the first potential doctors. When there is no feature value greater than the first preset threshold value among the similarities of the features of the first potential doctors, the similarities of the features of the first potential doctors corresponding to the next first similar doctor feature are compared with the first preset threshold value.

[0011] Optionally, the method further comprises:

[0012] The cosine of the angle of the feature vector of each doctor in the second potential doctor feature set is calculated in sequence according to the calculation formula of cosine similarity;

[0013] The second potential doctor feature corresponding to the feature vector angle cosine greater than or equal to 0.8 is extracted as the third potential doctor feature and a third potential doctor feature set is generated; wherein the third potential doctor feature set is used to characterize the doctor feature set with a similarity greater than 80% in the second potential doctor set.

[0014] Optionally, the method further comprises:

[0015] A text retrieval model is established for the third potential doctor feature set; wherein the text retrieval model is used to determine the representation of the text, the representation of the query, and the degree of matching between the query and the text; the text retrieval model includes a Boolean model, a vector space model, a probability theory model or a PageRank model.

[0016] Optionally, the method further comprises:

[0017] The calculated Euclidean distance value of the similarity between each third potential doctor and the relevant drug is greater than or equal to 0.8, generating a fourth potential doctor feature set; wherein the fourth potential doctor feature set is a doctor feature set used to characterize the similarity matching degree between the third potential doctor and the relevant drug information.

[0018] Optionally, the method further comprises:

[0019] Constructing an inverted file for the data in the fourth potential doctor feature set; extracting potential doctors ranked less than 10 in the inverted file to generate a fifth potential doctor feature set; wherein the fifth potential doctor feature set is used to represent the doctor feature set after extracting features from the top 10 doctors in the fourth potential doctor set using the file inverted technology;

[0020] Among them, the search method based on the inverted file includes:

[0021] Vocabulary query: split the query into separate words and search for these words in the vocabulary;

[0022] Query word occurrences: Get a list of occurrences related to all words in the query string;

[0023] Word occurrence operation: phrase query and approximate query are realized by operating the word occurrence list obtained in the previous step.

[0024] Optionally, constructing an inverted file for the data in the fourth potential doctor feature set further includes:

[0025] An inverted file is constructed for the data in the fourth potential doctor feature set; and the data of potential doctors whose ranking is greater than or equal to 10 in the inverted file are eliminated.

[0026] Optionally, the method further comprises:

[0027] Divide the generated fifth potential doctor feature set into several parts;

[0028] Build an index for each part and temporarily store it on disk, then release the memory space and traverse all the split parts until all documents are processed;

[0029] All indexes are multi-way merged to form an inverted index of the entire document library and generate the feature set of the sixth potential doctor.

[0030] In a second aspect, a medical information recommendation platform based on big data analysis technology is provided, the platform comprising:

[0031] A data collection module is used to collect the current information of each potential doctor; the current information includes at least the types of diseases that the potential doctor is good at and the chat records, electronic prescriptions, drug orders, and examination reports of patients before, during, and after the consultation;

[0032] A data mining module is used to establish a dictionary based on the current information of each potential doctor to generate a first potential doctor feature set, wherein the first potential doctor feature set includes a plurality of first potential doctor features, each of which represents a dictionary established by the current information of a potential doctor;

[0033] Calculate the similarity of each first potential doctor feature in pairs to obtain a first similarity set; wherein the first similarity set includes multiple first potential doctor feature similarities, and each similarity data in the first potential doctor feature similarities is used as a feature value;

[0034] When there is a feature value greater than a first preset threshold value in the similarities of the features of each first potential doctor, the first similarity doctor feature is used as the second potential doctor feature to generate a second potential doctor feature set; and the potential doctor corresponding to the second potential doctor feature set is used as a medical information recommendation target.

[0035] Optionally, the platform further includes:

[0036] A data cleaning module is used to clean the current information data of each potential doctor collected by the data collection module;

[0037] Data visualization module, used to display analysis results through charts, reports or data dashboards.

[0038] The beneficial effects brought by the technical solution provided by the embodiment of the present application include at least:

[0039] (1) By building a full-scenario medical service ecosystem, we can truly achieve the integration of data flow, capital flow, and service flow. This will help pharmaceutical companies reduce marketing costs, expand marketing channels, and achieve comprehensive digitalization of elements, processes, and services.

[0040] (2) Through this intelligent digital marketing platform, we can effectively tap into potential doctors and expand new target markets, thereby conducting intelligent and precise marketing for doctors.

[0041] (3) By combining big data, cloud computing, and artificial intelligence technologies, we can achieve digital identification of users, data analysis and feedback, and form a digital marketing closed loop, thereby helping medical companies achieve more efficient and accurate coverage, and more efficient and accurate transmission of professional information between pharmaceutical companies and doctors. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0043] Figure 1 A flowchart of the steps of a method for recommending medical information based on big data analysis technology provided in an embodiment of the present application;

[0044] Figure 2 A flowchart of a method for recommending medical information provided in an optional embodiment of the present application;

[0045] Figure 3 This is a flow chart of extracting the features of the second potential doctor using the cosine similarity calculation formula in the embodiment of the present application;

[0046] Figure 4 This is a schematic diagram of a process for extracting features of a third potential doctor using Euclidean distance in an embodiment of the present application;

[0047] Figure 5 A schematic diagram of an inverted file construction tree structure in a medical information recommendation method based on big data analysis technology provided in an embodiment of the present application;

[0048] Figure 6 A schematic diagram of a process for extracting the fifth potential doctor's features by using an inverted file to construct an index provided in an embodiment of the present application;

[0049] Figure 7 A schematic diagram of the application structure of a medical information recommendation platform based on big data analysis technology provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0051] In the description of the present invention, unless otherwise specified, "plurality" means two or more than two. The terms "first", "second", "third", "fourth", etc. in the specification and claims of the present invention and the above-mentioned drawings are intended to distinguish the objects referred to. For schemes with a sequential flow, this terminology does not have to be understood as describing a specific order or sequence. For schemes with system structures, this terminology does not distinguish between importance, positional relationships, etc.

[0052] In addition, the terms "comprises", "has" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that have been explicitly listed, but may also include other steps or units that are not explicitly listed but are inherent to these processes, methods, products or apparatuses, or steps or units that are added based on further optimization schemes conceived by the present invention.

[0053] To facilitate understanding of this embodiment, a medical information recommendation method based on big data analysis technology disclosed in an embodiment of the present application is first introduced in detail.

[0054] Please refer to Figure 1 , which shows a flowchart of a method for recommending medical information based on big data analysis technology provided by an embodiment of the present application. The method may include the following steps:

[0055] S1, collect current information of each potential doctor.

[0056] The current information includes at least the types of diseases that the potential doctor is good at and the chat records, electronic prescriptions, drug orders, and examination reports of previous patients before, during, and after the consultation.

[0057] S2, establishing a dictionary of the current information of each potential doctor to generate a first potential doctor feature set.

[0058] The first potential doctor feature set includes a plurality of first potential doctor features, and each first potential doctor feature represents a dictionary established by current information data of a potential doctor.

[0059] Collect the current information of the target doctor, build a dictionary, extract as much data information as possible and save it in the dictionary, generate the first potential doctor features and the first potential doctor feature set; structured data can be read directly, and unstructured data needs to be converted into structured data before use (for example: vector space model VSM (VectorSpace Modle), TF-IDF, etc. for text data).

[0060] A dictionary is a synonym dictionary or classification compilation. In a text retrieval system, a dictionary is a data compilation used to find related word information based on a word. In its simplest form, it contains two parts: a. important words related to a certain field; b. for each word in the dictionary, there are some words related to it.

[0061] The vector space model (VSM) is simple in concept. It simplifies the processing of text content into vector operations in the vector space. It also expresses semantic similarity with spatial similarity, which is intuitive and easy to understand. When documents are represented as vectors in the document space, the similarity between documents can be measured by calculating the similarity between vectors. The most commonly used similarity method in text processing is the cosine distance.

[0062] S3, calculating the similarity of each of the first potential doctor features in pairs to obtain a first similarity set.

[0063] The first similarity set includes a plurality of first potential doctor feature similarities, and the data of the first potential doctor feature similarities in the feature set are stored one by one, and each first potential doctor feature similarity is called a feature value.

[0064] The first potential doctor feature is vectorized using the vector space model. Generally, words are selected as feature items of the document vector. For example, the first potential doctor feature D1 = {depression, anxiety}, the first potential doctor feature D2 = {depression, anxiety, sleep disorder}, and the first potential doctor feature D3 = {depression, anxiety, bipolar disorder}. Now calculate the similarity between D1 and D2.

[0065] Step 1: Get the union of the two first potential doctor feature sets {depression, anxiety, sleep disorder};

[0066] Step 2: The vector representation of D1 is {1,1,0};

[0067] Step 3: The vector representation of D2 is {1,1,1};

[0068] Step 4: According to the vector cosine angle formula:

[0069]

[0070] The calculated similarity between D1 and D2 is 0.81, and the similarity between D2 and D3 is 0.66.

[0071] The first potential doctor features are combined in pairs and the first similarity between them is calculated.

[0072] Among them, the vector of the first potential doctor feature D1 is A, the vector of the first potential doctor feature D2 is B, the value range of cosine similarity is [-1,1], and the similarity between the same two vectors is 1. The value range of cosine distance is [0,2]. In the calculation result, the cosine eigenvalue of the angle between D1 and D2 is 81%, which is greater than the first preset threshold of 80%, so the two potential doctor features D1 and D2 are basically similar.

[0073] Thus, the processing flow of the first similarity doctor feature is obtained as follows:

[0074] (1) Combining the first potential doctor features in pairs, and extracting keywords between the two (every two first potential doctor features) in sequence;

[0075] (2) Combine the keywords between the two into a set and calculate the frequency of each feature for the words in this set;

[0076] (3) Generate word frequency vectors of each first potential doctor feature;

[0077] (4) Calculate the cosine similarity of the two vectors. The greater the eigenvalue is than the first preset threshold, the more similar the characteristics of the two potential doctors are.

[0078] S4, when there is a feature value greater than the first preset threshold value in the similarities of the features of each first potential doctor, the first similarity doctor feature is used as the second potential doctor feature to generate a second potential doctor feature set; and the potential doctor corresponding to the second potential doctor feature set is used as a medical information recommendation target.

[0079] After generating the first potential doctor feature set, the first potential doctor feature needs to be preprocessed to select suitable drugs for marketing. First, the similar features contained in the first potential doctor feature set are analyzed and decomposed into several similar feature sets. The similarities of each first potential doctor feature corresponding to the first similarity doctor feature are compared with the first preset threshold to determine whether there is a feature value greater than the first preset threshold in each first potential doctor feature similarity. If there is no feature value greater than the first preset threshold in each first potential doctor feature similarity, then the first potential doctor feature similarity corresponding to the next first similar doctor feature is compared with the first preset threshold; if there is a feature value greater than the first preset threshold in each first potential doctor feature similarity, then the first similarity doctor feature is used as the second potential doctor feature to generate the second potential doctor feature set. Among them, the first preset threshold is 80%.

[0080] Based on the above embodiments, Figure 2 FIG. 2 is a flow chart of the medical information recommendation method in the embodiment.

[0081] The following references Figure 3 ,The specific steps of extracting the second potential doctor features are described in detail using the cosine similarity calculation formula.

[0082] The definition formula of cosine similarity is:

[0083]

[0084] After normalization, ||A||2=1, ||B||2=1, ||A||2||B||2=1.

[0085] The calculated cosine distance is:

[0086]

[0087] For example: when this application counts user browsing behavior, the browsing vector of user A is (0, 1) and that of user B is (1, 0); at this time, the cosine distance between the two is large, while the Euclidean distance is small; this application analyzes the preferences of two users for different content and pays more attention to relative differences, so it is obvious that the cosine distance should be used.

[0088] According to the calculation formula of cosine similarity, the cosine of the angle between the feature vectors of each doctor in the feature set of the second potential doctor is calculated in turn, and it is determined whether the cosine of the angle between the feature vectors of each doctor in the feature set of the second potential doctor is 1, and the second potential doctor feature close to 1 is extracted as the third potential doctor feature and the third potential doctor feature set is generated. When it is close to 1, the more similar the two potential doctor features are, the smaller the cosine of the angle is, and the less related the two potential doctor features are; in this application, close to 1 specifically refers to greater than or equal to 0.8 and less than 1, wherein the third potential doctor feature set is used to characterize the doctor feature set with a similarity greater than 80% in the second potential doctor set. In the above example, A and B represent the feature vectors of two doctors in the same feature set.

[0089] The following references Figure 4 ,The specific steps of extracting the third potential doctor features using Euclidean distance are described in detail.

[0090] A text retrieval model is established for the third potential doctor feature set. The text retrieval model is used to strictly determine the representation of text, the representation of query, and the degree of matching between query and text. Text-based retrieval mainly refers to calculating the similarity between query and document based on the text content of the document. The representation and similarity calculation of query and document are collectively referred to as text retrieval models. There are several types: Boolean model, vector space model VSM (Vector Space Model), probability theory model, PageRank model.

[0091] When conducting precision marketing for target doctors, we hope to obtain drugs that are closely related to their characteristics, so we need to calculate the Euclidean distance between the characteristics of each third potential doctor and the similarity of the relevant drugs. Euclidean distance is a measure of the absolute distance between points in space, also known as Euclidean distance. It is directly related to the position coordinates of each point; while cosine distance measures the angle between space vectors, which is more reflected in the difference in direction rather than position.

[0092] Euclidean distance:

[0093]

[0094] Here, A represents the feature vector of the third potential doctor; B represents the feature vector of the relevant drug.

[0095] For example, when analyzing user activity, this application uses the number of logins (unit: times) and average usage time (unit: minutes) as features. The cosine distance will consider that the two users (1,10) and (1,100) are very close. However, it is obvious that the activity of these two users is very different. At this time, this application is more concerned with the absolute difference in values, and the Euclidean distance should be used.

[0096] In general, the Euclidean distance reflects the absolute difference in value, while the cosine distance reflects the relative difference in direction. The Euclidean distances of the calculated similarities between each third potential doctor and the relevant drug are extracted, and the values ​​close to 1 are extracted to generate the fourth potential doctor feature set. The fourth potential doctor feature set is a doctor feature set used to characterize the degree of similarity between the third potential doctor and the relevant drug information.

[0097] An inverted file is constructed for the data in the fourth potential doctor feature set to improve the query speed. The potential doctors ranked less than 10 in the inverted file are extracted to generate the fifth potential doctor feature set. The fifth potential doctor feature set is used to represent the doctor feature set after the features of the top 10 doctors in the fourth potential doctor set are extracted using the file inverted technology.

[0098] The search method based on inverted files is generally divided into the following three steps:

[0099] S1: Vocabulary query: split the query into separate words and search for these words in the vocabulary;

[0100] S2: Query word occurrence: obtain the occurrence list related to all words in the query string;

[0101] S3: Word occurrence operation: mainly through the operation of the word occurrence list obtained in the previous step to achieve phrase query and approximate query, etc.

[0102] Generally speaking, a database includes at least one forward document and one backward document. The forward document is a document collection that arranges all the records in the database according to the size of the record number, which constitutes the main content of the database. In the backward document, the characteristic identifier of the record is used as the basis for arrangement, and the record number containing this identifier is listed afterwards. Using the backward document can greatly improve the efficiency of retrieval.

[0103] The structure of the inverted file is relatively simple, for example Figure 5Take the example of "Today is Tuesday, we will have an ice day." shown in the figure. Assume that all words appearing in the document will be indexed, and the vocabulary list is stored in the index using a trie tree structure. The trie tree structure can sort the words. When recording the occurrence of a word, the number of characters from the index word to the start position of the document indicates the position of the index word (including spaces). An English letter is 1 character, and a Chinese character or Chinese punctuation is 2 characters. Of course, in actual applications, the occurrence of a word includes not only the position of the word in the document, but also the number of the document in which the word appears. This is only to illustrate the index construction principle, so the actual situation is simplified. For example: when the program is processing "Tuesday", a word occurrence list is added to the tree, and a record is added: "Tuesday: 10". When the user reads the English letter "t", the program obtains the English words containing "t" in the document through the inverted index query as "Today: 1" and "Tuesday: 10". The second keyword "o" or "u" of the word needs to be entered to accurately match the document number of the corresponding word.

[0104] The following references Figure 6 ,The specific steps of extracting the features of the fifth potential doctor using the inverted file to build an index are explained in detail.

[0105] If the text library is very large and the size of the inverted file exceeds the available memory size, the inverted file obviously cannot be constructed in one go. At this time, it is necessary to divide the fifth potential doctor feature set into several parts, build an index for each part and temporarily store it on disk, then release the memory space and process the other parts in turn until all documents are processed. Finally, all partial indexes are multi-way merged, that is, when the same word appears in multiple indexes, the corresponding word occurrence lists are merged. After merging, the inverted index of the entire document library is formed, and the final sixth potential doctor feature set is generated.

[0106] The present application also provides a medical information recommendation platform based on big data analysis technology, such as Figure 7 This is a schematic diagram of the application structure of the medical information recommendation platform, which specifically includes:

[0107] Data collection 1, collecting the current information of the potential doctor; the current information includes the types of diseases the target doctor is good at and the chat records, electronic prescriptions, drug orders, examination reports, etc. of the patients he / she has seen before, during and after the consultation.

[0108] Specifically, the data collection module is used to collect the current information of each potential doctor; the current information includes at least the types of diseases that the potential doctor is good at and the chat records, electronic prescriptions, drug orders, and examination reports of the patients he / she has seen before, during, and after the consultation;

[0109] Data cleaning 2. The collected raw data is often disorganized and needs to be cleaned. Data cleaning refers to processing the data to make it meet actual needs. In data cleaning, this application needs to remove useless information and deduplicate duplicate data.

[0110] Data mining 3. The data mining process is relatively complex and requires the use of various mining algorithms to find useful information from massive data. The present invention mainly includes predictive modeling, inverted classification, cluster analysis, etc.

[0111] Specifically, it is used to establish a dictionary with the current information of each potential doctor to generate a first potential doctor feature set, wherein the first potential doctor feature set includes multiple first potential doctor features, and each first potential doctor feature represents a dictionary established by the current information of a potential doctor; the similarity of each first potential doctor feature is calculated for each combination to obtain a first similarity set; wherein the first similarity set includes multiple first potential doctor feature similarities; when there is a feature value greater than a first preset threshold value in the similarities of each first potential doctor feature, the first similarity doctor feature is used as a second potential doctor feature to generate a second potential doctor feature set; and the potential doctor corresponding to the second potential doctor feature set is used as a medical information recommendation target.

[0112] Data visualization 4. Data visualization is the last link of the intelligent digital marketing platform. It can display the analysis results through charts, reports, data dashboards and other methods, and can support user-computer interaction. Compared with previous reports, statistical charts and other methods, data visualization has stronger interactivity and intuitiveness, and can provide users with a better experience.

[0113] For the specific definition of the medical information recommendation platform based on big data analysis technology, please refer to the definition of the medical information recommendation method based on big data analysis technology mentioned above, which will not be repeated here. Each module in the above-mentioned medical information recommendation platform based on big data analysis technology can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0114] In one embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned medical information recommendation method based on big data analysis technology are implemented.

[0115] The computer-readable storage medium provided in this embodiment has similar implementation principles and technical effects to those of the above method embodiments, and will not be described in detail here.

[0116] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent application shall be subject to the attached claims.

Claims

1. A method for recommending medical information based on big data analysis technology, characterized in that: The method comprises: Collecting the current information of each potential doctor; the current information includes at least the types of diseases that the potential doctor is good at and the chat records, electronic prescriptions, drug orders, and examination reports of patients before, during, and after the consultation; The current information of each potential doctor is used to establish a dictionary to generate a first potential doctor feature set, wherein the first potential doctor feature set includes a plurality of first potential doctor features, each of which represents a dictionary established by the current information of a potential doctor; Calculate the similarity of each first potential doctor feature in pairs to obtain multiple first similarity sets; wherein the first similarity set contains multiple first potential doctor feature similarities, and each similarity data in the first potential doctor feature similarities is used as a feature value; Determine whether there is a feature value greater than a first preset threshold value in the similarities of the features of each first potential doctor; if there is no feature value greater than the first preset threshold value in the similarities of the features of each first potential doctor, compare the similarities of the features of each first potential doctor corresponding to the next first similar doctor feature with the first preset threshold value; if there is a feature value greater than the first preset threshold value in the similarities of the features of each first potential doctor, use the first similarity doctor feature as the second potential doctor feature to generate a second potential doctor feature set; The method further comprises: The cosine of the angle of the feature vector of each doctor in the second potential doctor feature set is calculated in sequence according to the calculation formula of cosine similarity; Extract the second potential doctor feature corresponding to the feature vector angle cosine greater than or equal to 0.8 as the third potential doctor feature and generate a third potential doctor feature set; wherein the third potential doctor feature set is used to characterize the doctor feature set with a similarity greater than 80% in the second potential doctor set The method further comprises: Establishing a text retrieval model for the third potential doctor feature set; wherein the text retrieval model is used to determine the representation of the text, the representation of the query, and the degree of matching between the query and the text; the text retrieval model includes a Boolean model, a vector space model, a probability theory model or a PageRank model; The method further comprises: The calculated Euclidean distances of the similarities between each third potential doctor and the relevant drug have a distance value greater than or equal to 0.8, and a fourth potential doctor feature set is generated; wherein the fourth potential doctor feature set is a doctor feature set used to characterize the degree of similarity matching between the third potential doctor and the relevant drug information; The method further comprises: Constructing an inverted file for the data in the fourth potential doctor feature set; extracting potential doctors ranked less than 10 in the inverted file to generate a fifth potential doctor feature set; wherein the fifth potential doctor feature set is used to represent the doctor feature set after extracting features from the top 10 doctors in the fourth potential doctor set using the file inverted technology; Among them, the search method based on the inverted file includes: Vocabulary query: split the query into separate words and search for these words in the vocabulary; Query word occurrences: Get a list of occurrences related to all words in the query string; Word occurrence operation: phrase query and approximate query are realized by operating the word occurrence list obtained in the previous step; The inverted file constructed for the data in the fourth potential doctor feature set also includes: Construct an inverted file for the data in the fourth potential doctor feature set; remove the potential doctors whose ranking is greater than or equal to 10 in the inverted file; The method further comprises: Divide the generated fifth potential doctor feature set into several parts; Build an index for each part and temporarily store it on disk, then release the memory space and traverse all the split parts until all documents are processed; All indexes are multi-way merged to form an inverted index of the entire document library, a sixth potential doctor feature set is generated, and the potential doctors corresponding to the sixth potential doctor feature set are used as medical information recommendation targets.

2. A medical information recommendation platform based on big data analysis technology, characterized in that: The platform includes: A data collection module is used to collect the current information of each potential doctor; the current information includes at least the types of diseases that the potential doctor is good at and the chat records, electronic prescriptions, drug orders, and examination reports of patients before, during, and after the consultation; A data mining module is used to establish a dictionary based on the current information of each potential doctor to generate a first potential doctor feature set, wherein the first potential doctor feature set includes a plurality of first potential doctor features, each of which represents a dictionary established by the current information of a potential doctor; Calculate the similarity of each first potential doctor feature in pairs to obtain multiple first similarity sets; wherein the first similarity set contains multiple first potential doctor feature similarities, and each similarity data in the first potential doctor feature similarities is used as a feature value; Determine whether there is a feature value greater than a first preset threshold value in the similarities of the features of each first potential doctor; if there is no feature value greater than the first preset threshold value in the similarities of the features of each first potential doctor, compare the similarities of the features of each first potential doctor corresponding to the next first similar doctor feature with the first preset threshold value; if there is a feature value greater than the first preset threshold value in the similarities of the features of each first potential doctor, use the first similarity doctor feature as the second potential doctor feature to generate a second potential doctor feature set; The platform also includes: A data cleaning module is used to clean the current information data of each potential doctor collected by the data collection module; Data visualization module, used to display analysis results through charts, reports or data dashboards; The platform also includes: Determine whether there is a feature value greater than a first preset threshold value among the similarities of the first potential doctors' features; when there is no feature value greater than the first preset threshold value among the similarities of the first potential doctors' features, compare the similarities of the first potential doctors' features corresponding to the next first similar doctor feature with the first preset threshold value; The platform also includes: The cosine of the angle of the feature vector of each doctor in the second potential doctor feature set is calculated in sequence according to the calculation formula of cosine similarity; Extract the second potential doctor feature corresponding to the feature vector angle cosine greater than or equal to 0.8 as the third potential doctor feature and generate a third potential doctor feature set; wherein the third potential doctor feature set is used to characterize the doctor feature set with a similarity greater than 80% in the second potential doctor set The platform also includes: Establishing a text retrieval model for the third potential doctor feature set; wherein the text retrieval model is used to determine the representation of the text, the representation of the query, and the degree of matching between the query and the text; the text retrieval model includes a Boolean model, a vector space model, a probability theory model or a PageRank model; The platform also includes: The calculated Euclidean distances of the similarities between each third potential doctor and the relevant drug have a distance value greater than or equal to 0.8, and a fourth potential doctor feature set is generated; wherein the fourth potential doctor feature set is a doctor feature set used to characterize the degree of similarity matching between the third potential doctor and the relevant drug information; The platform also includes: Constructing an inverted file for the data in the fourth potential doctor feature set; extracting potential doctors ranked less than 10 in the inverted file to generate a fifth potential doctor feature set; wherein the fifth potential doctor feature set is used to represent the doctor feature set after extracting features from the top 10 doctors in the fourth potential doctor set using the file inverted technology; Among them, the search method based on the inverted file includes: Vocabulary query: split the query into separate words and search for these words in the vocabulary; Query word occurrences: Get a list of occurrences related to all words in the query string; Word occurrence operation: phrase query and approximate query are realized by operating the word occurrence list obtained in the previous step; The inverted file constructed for the data in the fourth potential doctor feature set also includes: Construct an inverted file for the data in the fourth potential doctor feature set; remove the potential doctors whose ranking is greater than or equal to 10 in the inverted file; The platform also includes: Divide the generated fifth potential doctor feature set into several parts; Build an index for each part and temporarily store it on disk, then release the memory space and traverse all the split parts until all documents are processed; All indexes are multi-way merged to form an inverted index of the entire document library, a sixth potential doctor feature set is generated, and the potential doctors corresponding to the sixth potential doctor feature set are used as medical information recommendation targets.

Images