CN114023447A

CN114023447A - Training method and device for rare patient number prediction model

Info

Publication number: CN114023447A
Application number: CN202111480742.5A
Authority: CN
Inventors: 张敏; 李佳玉; 刘奕群; 马少平; 苏航; 张抒扬; 金晔; 张磊
Original assignee: Tsinghua University; Peking Union Medical College Hospital Chinese Academy of Medical Sciences
Current assignee: Tsinghua University; Peking Union Medical College Hospital Chinese Academy of Medical Sciences
Priority date: 2021-12-06
Filing date: 2021-12-06
Publication date: 2022-02-08

Abstract

The disclosure relates to a rare patient number prediction model training method and a device, wherein the method comprises the following steps: preprocessing the text of the rare disease name to obtain a rare disease query keyword database; according to the rare disease query keyword database and the query history database, determining the query number of the plurality of rare diseases in a plurality of time periods in a plurality of regions; and training the rare patient number prediction model according to the number of inquired people and the number of confirmed people. According to the training method of the rare disease patient number prediction model, the text of the rare disease name can be preprocessed, omission of query data is reduced, and accuracy of the data is improved. And the number of people inquiring the rare disease information can be predicted based on the inquiry historical database of the search engine, and then the rare patient number prediction model is trained based on the number of people and the actual number of people to be diagnosed, so that the prediction precision of the model on the number of rare patients is improved.

Description

Training method and device for rare patient number prediction model

Technical Field

The disclosure relates to the technical field of computers, in particular to a training method and a training device for a rare patient number prediction model.

Background

Rare diseases, also known as "orphan disease", are a general term for a series of diseases with very low incidence rates. At present, more than 8000 kinds of rare diseases are known globally, genetic diseases are taken as main diseases, the number of people affected by the rare diseases globally exceeds 2.6 hundred million, about 2000 million people exist in China, and due to the characteristics that the rare diseases are difficult to discover and diagnose, diseased groups are difficult to locate and are less concerned. Therefore, in the research of the rare diseases, the early accurate screening and identification research of the rare diseases is promoted to have important significance, the early discovery, early diagnosis and early treatment of rare disease crowds can be gradually realized, and the mental and economic burden of patients and family members is reduced. In the process of screening and identifying the rare diseases, the method has important significance for researching the incidence trend of the rare diseases.

In the internet era, search engines have become an important way to acquire information, and search query data of users can reflect certain disease conditions and development trends. However, the existing methods for predicting the number and growth Trend of diseases based on search engines, such as the famous Google Trend (Google tend) for predicting the development of influenza, are focused on the research of infectious diseases or epidemic diseases, and the prediction method for rare diseases is still blank at present. Unlike influenza and other epidemic diseases, rare diseases have the characteristics of rare occurrence and rare occurrence, search keywords are difficult to determine, search query data are very sparse, the growth trend of cases is greatly different from influenza and other epidemic diseases, and difficulty is caused in predicting the number of rare patients.

Disclosure of Invention

The disclosure provides a training method and a device of a rare patient number prediction model.

According to an aspect of the present disclosure, there is provided a rare patient number prediction model training method, including: preprocessing texts of various rare disease names to obtain a rare disease query keyword database; determining the number of inquired people for the multiple rare diseases in multiple time periods in multiple regions according to the rare disease inquiry keyword database and the inquiry history database of the search engine; and training the rare patient quantity prediction model according to the number of inquired people of the rare diseases in the time periods of the regions and the number of diagnosed people of the rare diseases in the time periods of the regions, so as to obtain the trained rare patient quantity prediction model.

In one possible implementation, the text of the plurality of rare disease names is preprocessed, including at least one of: preprocessing symbol texts in the texts of the rare disease names; preprocessing the alphabetic text in the text of the rare disease name; preprocessing the digital text in the text of the rare disease name; preprocessing preset characters in the text of the rare disease name; preprocessing abbreviated texts in the texts of the rare disease names; pre-processing transliterated text in the text of the rare disease name.

In one possible implementation, preprocessing the text of multiple rare disease names to obtain a rare disease query keyword database includes: adding the text of the plurality of rare disease names and the preprocessed text to the rare disease query keyword database.

In one possible implementation manner, determining the number of inquired people for a plurality of rare diseases in a plurality of time periods in a plurality of regions according to the rare disease inquiry keyword database and the inquiry history database of the search engine comprises: determining the number ratio of the query times of various rare diseases in the rare disease query keyword database to the total query times in the corresponding time period of the corresponding region in a plurality of time periods of a plurality of regions in a query history database of the search engine; and determining the number of inquired people for the multiple rare diseases in multiple time periods of the multiple regions according to the frequency ratio and the population data in the corresponding time period of the corresponding region.

In a possible implementation manner, training the rare patient number prediction model according to the number of inquired people for the rare diseases in the time periods of the regions and the number of confirmed people for the rare diseases in the time periods of the regions to obtain the trained rare patient number prediction model, includes: respectively normalizing the inquired people number of the multiple rare diseases in multiple time periods of multiple regions and the diagnosed people number of the multiple rare diseases in multiple time periods of multiple regions to obtain inquired people number sample data and diagnosed people number sample data; determining a loss function of the rare patient quantity prediction model according to the sample data of the diagnosed number of patients and the predicted diagnosed number obtained by inputting the sample data of the inquired number of patients into the rare patient quantity prediction model; adjusting model parameters of the rare patient number prediction model on a training set according to the loss function and preset hyper-parameters; and under the condition that the training condition is met, obtaining the trained rare patient number prediction model.

In one possible implementation, obtaining the trained rare patient number prediction model when a training condition is satisfied includes: determining the evaluation index of the rare patient number prediction model according to the confirmed patient number sample data of the verification set and the predicted confirmed patient number obtained by inputting the inquired patient number sample data into the rare patient number prediction model; and in the preset number of training rounds, calculating an evaluation index after each training round, and stopping model training under the condition that the evaluation index is unchanged after a plurality of training rounds to obtain the trained rare patient number prediction model.

In one possible implementation, the method further includes: determining the number of inquired people aiming at the preset type of rare diseases in a preset time period in a preset area in a query historical database of the search engine; normalizing the number of inquired people of the rare diseases of a preset type in a preset time period of the preset area, inputting the trained rare patient number prediction model, and obtaining first prediction data of the number of cases of the rare diseases of the preset type in the preset time period of the preset area; and carrying out inverse normalization processing on the first prediction data to obtain the predicted confirmed diagnosis number of the preset type of rare diseases in a preset time period of a preset area.

According to an aspect of the present disclosure, there is provided a rare patient number prediction model training device, including: the text preprocessing module is used for preprocessing texts with various rare disease names to obtain a rare disease query keyword database; the query number determining module is used for determining the query number of the plurality of rare diseases in a plurality of time periods in a plurality of regions according to the rare disease query keyword database and the query history database of the search engine; and the training module is used for training the rare patient quantity prediction model according to the number of inquired people of the rare diseases in the time periods of the regions and the number of diagnosed people of the rare diseases in the time periods of the regions to obtain the trained rare patient quantity prediction model.

In one possible implementation, the text preprocessing module is further configured to at least one of: preprocessing symbol texts in the texts of the rare disease names; preprocessing the alphabetic text in the text of the rare disease name; preprocessing the digital text in the text of the rare disease name; preprocessing preset characters in the text of the rare disease name; preprocessing abbreviated texts in the texts of the rare disease names; pre-processing transliterated text in the text of the rare disease name.

In one possible implementation, the text preprocessing module is further configured to: adding the text of the plurality of rare disease names and the preprocessed text to the rare disease query keyword database.

In one possible implementation, the query people number determination module is further configured to: determining the number ratio of the query times of various rare diseases in the rare disease query keyword database to the total query times in the corresponding time period of the corresponding region in a plurality of time periods of a plurality of regions in a query history database of the search engine; and determining the number of inquired people for the multiple rare diseases in multiple time periods of the multiple regions according to the frequency ratio and the population data in the corresponding time period of the corresponding region.

In one possible implementation, the training module is further configured to: respectively normalizing the inquired people number of the multiple rare diseases in multiple time periods of multiple regions and the diagnosed people number of the multiple rare diseases in multiple time periods of multiple regions to obtain inquired people number sample data and diagnosed people number sample data; determining a loss function of the rare patient quantity prediction model according to the sample data of the diagnosed number of patients and the predicted diagnosed number obtained by inputting the sample data of the inquired number of patients into the rare patient quantity prediction model; adjusting model parameters of the rare patient number prediction model on a training set according to the loss function and preset hyper-parameters; and under the condition that the training condition is met, obtaining the trained rare patient number prediction model.

In one possible implementation, the training module is further configured to: determining the evaluation index of the rare patient number prediction model according to the confirmed patient number sample data of the verification set and the predicted confirmed patient number obtained by inputting the inquired patient number sample data into the rare patient number prediction model; and in the preset number of training rounds, calculating an evaluation index after each training round, and stopping model training under the condition that the evaluation index is unchanged after a plurality of training rounds to obtain the trained rare patient number prediction model.

In one possible implementation, the apparatus further includes: the prediction module is used for determining the number of inquired people aiming at the rare diseases of a preset type in a preset time period of a preset region in a query history database of the search engine; normalizing the number of inquired people of the rare diseases of a preset type in a preset time period of the preset area, inputting the trained rare patient number prediction model, and obtaining first prediction data of the number of cases of the rare diseases of the preset type in the preset time period of the preset area; and carrying out inverse normalization processing on the first prediction data to obtain the predicted confirmed diagnosis number of the preset type of rare diseases in a preset time period of a preset area.

According to an aspect of the present disclosure, there is provided an electronic device including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the memory-stored instructions to perform the above-described method.

According to an aspect of the present disclosure, there is provided a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the above-described method.

According to the training method of the rare disease patient number prediction model, the text of the rare disease name can be preprocessed according to the characteristics of rare occurrence and rare occurrence of rare diseases and the characteristic of less query data, omission of query data is reduced, and accuracy of the data is improved. And the number of people inquiring the rare disease information can be predicted based on the inquiry historical database of the search engine, and then the rare patient number prediction model is trained based on the number of inquired people and the actual number of diagnosed people, so that the prediction precision of the model on the number of rare patients is improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. Other features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments, which proceeds with reference to the accompanying drawings.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure.

FIG. 1 illustrates a flow diagram of a rare patient number prediction model training method according to an embodiment of the present disclosure;

FIG. 2 illustrates an application of the rare patient number prediction model training method according to an embodiment of the present disclosure;

FIG. 3 shows a block diagram of a rare patient number prediction model training apparatus according to an embodiment of the present disclosure;

FIG. 4 shows a block diagram of an electronic device in accordance with an embodiment of the disclosure;

fig. 5 shows a block diagram of an electronic device in accordance with an embodiment of the disclosure.

Detailed Description

Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. In the drawings, like reference numbers can indicate functionally identical or similar elements. While the various aspects of the embodiments are presented in drawings, the drawings are not necessarily drawn to scale unless specifically indicated.

The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration. Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments.

The term "and/or" herein is merely an association describing an associated object, meaning that three relationships may exist, e.g., a and/or B, may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the term "at least one" herein means any one of a plurality or any combination of at least two of a plurality, for example, including at least one of A, B, C, and may mean including any one or more elements selected from the group consisting of A, B and C.

Furthermore, in the following detailed description, numerous specific details are set forth in order to provide a better understanding of the present disclosure. It will be understood by those skilled in the art that the present disclosure may be practiced without some of these specific details. In some instances, methods, means, elements and circuits that are well known to those skilled in the art have not been described in detail so as not to obscure the present disclosure.

Fig. 1 shows a flow diagram of a method of rare patient number prediction model training according to an embodiment of the present disclosure, as shown in fig. 1, the method comprising:

in step S11, preprocessing the texts of the multiple rare disease names to obtain a rare disease query keyword database;

in step S12, determining the number of persons who can be queried for a plurality of rare diseases in a plurality of time periods in a plurality of regions according to the rare disease query keyword database and the query history database of the search engine;

in step S13, the rare patient number prediction model is trained according to the number of inquired people for the plurality of rare diseases in the plurality of time periods in the plurality of regions and the number of diagnosed people for the plurality of rare diseases in the plurality of time periods in the plurality of regions, so as to obtain the trained rare patient number prediction model.

In one possible implementation, with the development of the internet, the patients and their family, relatives and friends can inquire the information of the diseases through the internet, so that there is a certain correlation between the number of newly added patients in a time period (for example, within one year, within one month, etc.) and the number of people inquiring the information of the diseases in the time period, and similarly, there can also be a correlation between the number of people inquiring the information of rare diseases through the internet in a certain time period and the number of people newly added with rare patients in the time period.

In a possible implementation mode, aiming at the characteristics of rare occurrence and rare occurrence of the rare diseases, the characteristics of less query data and the like, the types of the rare diseases predicted by the rare patient quantity prediction model can be determined by utilizing the known rare disease catalogue. For example, in the first batch of rare diseases, 121 rare diseases are included, and the names of the rare diseases can be used as the basis for training the model. For example, a person who inquires about rare disease information (e.g., the patient himself, or the family, relatives, and friends of the patient, etc.) can use the name of the rare disease in the directory as an inquiry keyword and search for information about rare disease using a search engine.

In a possible implementation manner, as described above, the number of rare diseases is small, and when the query is performed on rare diseases, the number and the times of the query are small, so that in the process of selecting the query records of rare diseases in the database of the search engine, the query records may have omission or deviation for any text (for example, characters, letters, symbols, and the like) in the name of rare diseases, and in the case that the total amount of the query records is small, the data of the query records may have a large deviation due to any omission or deviation.

For example, there are billions of query records for a common disease, and even if an error occurs in a query keyword of the common disease, the data screened out from the query records is millions or even tens of millions less than the real data, the missing data still occupies a small proportion of the whole data, and the influence on processing (e.g., neural network training) using the query records may not be serious. However, the data volume of the query records with rare diseases is small overall, for example, only tens of thousands of query records exist, and due to omission or deviation of a certain keyword, thousands of query records are omitted, and the omitted data occupies a large proportion of the data as a whole, which may have a serious influence on the processing using the query records.

In another example, the rare disease names in the above-mentioned directory are often professional medical terms, and errors or omissions of some terms may also cause data distortion of the query log. For example, in the list of rare diseases, glycogen storage disease (type I, type ii) is included, and if the query keyword is wrongly written as diabetes (type I, type ii), the obtained query record is a query record for a common disease, i.e., diabetes, so that the data of the query record is distorted.

In a possible implementation manner, for the above problem, in step S11, the text of the name of the rare disease may be preprocessed, so that the screened query record for each rare disease is as accurate as possible, so as to improve the data accuracy and reduce the data omission. In an example, the name medical terms of each rare disease may be normalized and the name associated therewith determined, e.g., common acronym, alias, transliteration, foreign language abbreviation, etc., for the rare disease, i.e., the normalized medical terms and other appellations associated therewith are obtained by preprocessing the names of the various rare diseases. Step S11 may further include: adding text (e.g., canonical medical terms) of the plurality of rare disease names, and the preprocessed text to the rare disease query keyword database. That is, the database of query keywords is established to include not only the canonical medical terms of rare disease names, but also other names associated therewith obtained after preprocessing. The query keywords used by different people who query for the same rare disease may be different, the keyword query database established in the above manner may include multiple titles for each rare disease, and the query data for rare diseases in the query history database screened by the keyword query database may reduce omission and deviation of the query data.

In one possible implementation, the preprocessing may include a plurality of ways, and the step S11 may include at least one of the following: preprocessing symbol texts in the texts of the rare disease names; preprocessing the alphabetic text in the text of the rare disease name; preprocessing the digital text in the text of the rare disease name; preprocessing preset characters in the text of the rare disease name; preprocessing abbreviated texts in the texts of the rare disease names; pre-processing transliterated text in the text of the rare disease name.

In one possible implementation, the symbolic text in the text of the rare disease name may be pre-processed. If the text of the rare disease name contains punctuation marks (such as brackets, quotation marks and the like), the replaceable punctuation marks are punctuation marks under an English/Chinese input method, or the punctuation marks are removed, the replaceable punctuation marks can be replaced by spaces, and the text after the preprocessing and the text of the rare disease name are added to the rare disease query keyword database. In an example, an "osteogenesis imperfecta (brittle bone disease)" is a rare disease, a text of a name of the rare disease can be preprocessed to obtain a plurality of preprocessed texts such as the "osteogenesis imperfecta (brittle bone disease)", "osteogenesis imperfecta brittle bone disease", and the preprocessed text and the "osteogenesis imperfecta (brittle bone disease)" before preprocessing are added to a rare disease query keyword database to serve as query keywords for the disease.

In one possible implementation, the alphabetic text in the text of the rare disease name may be pre-processed. For example, if the text of the rare disease name contains greek letters, the greek letters are replaced with english translation and common chinese translation, respectively, such as "α" is replaced with "alpha" or "alpha", and vice versa, for example, if the text of the rare disease name contains english synonymous with other foreign language, such as "alpha", it is replaced with "α" or "alpha". In an example, the 'beta-ketothiolase deficiency' is a rare disease, a text of a name of the rare disease can be preprocessed to obtain a plurality of preprocessed texts such as 'beta-ketothiolase deficiency', and the like, and the preprocessed text and the 'beta-ketothiolase deficiency' before preprocessing are added to a rare disease query keyword database to serve as query keywords for the disease.

In one possible implementation, the numeric text in the text of the rare disease name may be pre-processed. For example, if the name of a rare disease contains one of Roman numerals, Arabic numerals or Chinese numerals, the other two numerical forms can be replaced. In the example, the 'type II glycogen storage disease' is a rare disease, the text of the name of the rare disease can be preprocessed to obtain a plurality of preprocessed texts such as 'type 2 glycogen storage disease', 'type II glycogen storage disease', and the preprocessed texts and the 'type II glycogen storage disease' before preprocessing are added to a rare disease query keyword database to serve as query keywords for the disease.

In one possible implementation, pre-set words in the text of the rare disease name are pre-processed. For example, if a "sign" or "symptom" is included in the rare disease name, the "sign" or "sign" may be substituted. In an example, the "alport syndrome" is a rare disease, the text of the name of the rare disease can be preprocessed to obtain the "alport syndrome", and both the preprocessed text and the "alport syndrome" before preprocessing are added to the rare disease query keyword database to serve as query keywords for the disease.

In one possible implementation, the abbreviated text in the text of the rare disease name may be pre-processed. For example, if the rare disease name includes an english abbreviation, a medical related word such as "disease" or "symptom" may be added after the english abbreviation, for example, "hhh" is a rare disease, the text of the rare disease name may be preprocessed to obtain a plurality of preprocessed texts such as "hhh disease" or "hhh symptom", and both the preprocessed text and the preprocessed "hhh" are added to the rare disease query keyword database as query keywords for the disease.

In one possible implementation, the transliterated text in the text of the rare disease name is preprocessed, for example, if the rare disease name contains a wrong-prone character, a missed-prone character, and a word in a different transliteration mode, the wrong-prone character, the missed-prone character, and the word in the different transliteration mode are replaced by the common wrong-prone character, the missed-prone character, or the word in the different transliteration mode. For example, "albert syndrome" is a rare disease, the text of the name of the rare disease can be preprocessed to obtain a plurality of preprocessed texts such as "alport syndrome", "albert syndrome", "alport syndrome", and the like, and both the preprocessed texts and the preprocessed "albert syndrome" are added to a rare disease query keyword database to serve as query keywords for the disease.

In one possible implementation, the preprocessing is not limited to the above, for example, a translation of Chinese and foreign language may be added, such as the English translation "fragilisissism" of "osteogenesis imperfecta (osteopathia)". Alternatively, transliteration text may be added, e.g., the transliteration of "alport syndrome" is "albert syndrome". Alternatively, a common alternative name, for example, "glass doll" or "glass person" other than "osteogenesis imperfecta (brittle bone disease)" may be added. The present disclosure is not limited as to the manner of pretreatment. Furthermore, the texts with the names of the rare diseases can be respectively preprocessed in multiple preprocessing modes to obtain multiple preprocessed texts, for example, the transliteration of the "alport syndrome" is the "albert syndrome", after specific characters are replaced, the "alport syndrome", the "albert syndrome", and the like can be obtained, and the combination mode of the preprocessing modes is not limited in the disclosure.

In the example, after various pretreatment of "osteogenesis imperfecta (brittle bone disease)", "osteogenesis imperfecta (brittle bone)", "osteogenesis imperfecta syndrome", "primary brittle bone disease", "periosteal dysplasia", "brittle bone disease", "osteogenesis imperfecta", "congenital dysplasia", "glass bone", "porcelain doll", "ceramic doll", "vitreous doll", "diagonalism ossium", "osteoprogenitrene immunoperotactoids", "periododysplasia", "iopathioticosis", and the like are performed, the text described above and the text before treatment are added to the rare disease query keyword database.

In the example, after preprocessing the "β -ketothiolase deficiency" in various ways, the "β -ketothiolase deficiency", "mitochondrial acetoacetyl-coa thiolase deficiency", "beta-ketothiolase deficiency", "mitochysene deficiency", "mitochroic acetoacyl-coa deficiency", "mitochoxystrobence deficiency", "mitochoxys, etc. may be obtained, and the text may be processed with the above-mentioned keywords and added to the database before preprocessing.

In the examples, "glycogen storage disease type I, glycogen storage disease type I", "glycogen storage disease type 1", "glycogen storage disease type I", "glycogen storage disease type one", "glycogen storage disease type 1", "glycogen storage disease type I", "glycogen storage disease type 1", "glycogen storage disease type ii glycogen storage disease", "type II glycogen storage disease", "type 2 glycogen storage disease", "type II glycogen storage disease", "type two glycogen storage disease", "type 2 glycogen storage disease", "type II glycogen accumulation disease", "type two glycogen accumulation disease", "type 2 glycogen accumulation disease", "type II glycogen accumulation disease", "type 2 glycogen accumulation disease", "type II glycogen accumulation disease", "type 2 glycogen accumulation disease", "glycogen accumulation disease type 2", "glycogen storage disease", "glycogen accumulation disease type (type i, type II)", "glycogen accumulation disease type, the method comprises the steps of preprocessing a plurality of texts such as glycogen storage disease, glycogen storage disease and glycogen storage disease, and adding the texts and the texts before processing to a rare disease query keyword database.

In the example, "Alport syndrome" is preprocessed in various ways, so that "Alport syndrome", "albert syndrome", "Alport syndrome", "albert syndrome", "albert syndrome", "alport syndrome", "hereditary nephropathies", "alport syndrome", "hereditary nephropathies", "eye-ear-kidney syndrome", "eye-ear syndrome", "genetic progressive nephritis", "genetic progressive renal syndrome", "genetic progressive renal, and adding the texts and the texts before processing into a rare disease query keyword database.

In an example, after preprocessing the "HHH syndrome" in various ways, various preprocessed texts such as "hyperornithine-hyperammonemia-homocitrullinuria syndrome", "hyperornithine", "hyperammonemia", "homocitrullinuria syndrome", "hyperornithine-hyperammonemia-homocitrullinemia syndrome", "ornithine transferase deficiency", "hyperornithinaemia homocitrullinia syndrome", "hyperonitis hydroammonemia homocitrullinia syndrome", "hyperonitis homocitrullinia syndrome", "hypernathiae syndrome", "HHH disease", and the like can be obtained, and the above texts and the text before processing are added to the rare text query database.

By the method, the texts of the names of the rare diseases can be preprocessed, the texts expressing the same rare disease in different expression modes are obtained, the rare disease query keyword database is reasonably expanded, deviation and omission of query data are reduced, and accuracy of the data is improved.

In one possible implementation, as described above, there may also be an association between the number of people who query for information on rare diseases using the internet during a certain period of time and the number of newly added rare patients during the period of time. Further, the rare diseases are different from epidemic diseases, the disease ways are more inherited than infectious diseases, and therefore, for the genetic diseases which are common in the same family, the distribution areas can also be used as reference factors for predicting the onset trend of the rare diseases. For example, a province is warm and humid in climate, and a family living in the province throughout the year has a higher probability of certain skeletal rare diseases than a province with dry and cold climate. Therefore, the incidence relation between the number of people who inquire the information of the rare diseases in a certain time period of a certain area by using the internet and the number of newly added rare patients in the time period of the certain area can be considered.

In a possible implementation manner, the number of inquired people for inquiring the information of the rare diseases in each time period in each region can be determined in an inquiry historical database of a search engine based on the rare disease inquiry keyword database, and the relation between the number of inquired people and the number of real diagnosed people in the corresponding time period of the corresponding region is utilized to solve, namely, the rare patient number prediction model is solved. And then the model can be used for predicting the number of the confirmed diagnoses of the newly-increased rare diseases in each region in other time periods based on the query records of the search engine.

In one possible implementation, in step S12, in the query history database of the search engine, based on the keywords in the rare disease query keyword database, the number of queries for various rare diseases can be determined. In an example, query records in a query history database of a search engine can be traversed for a plurality of query keywords for each rare disease to screen out query records for various rare diseases and count the number of queries for various rare diseases. Further, the query records may be classified, for example, by region and time period, to determine the number of queries for each rare disease in each time period in each region.

In one possible implementation manner, after the number of queries for various rare diseases in each time period of each region is determined, the number of queries for various rare diseases in each time period of each region can be determined based on the number of queries for various rare diseases in each time period of each region. In an example, the number of queries may be predicted based on a relationship between the number of queries and the number of queries, for example, statistics may be performed based on historical data, or a model such as a neural network may be used to determine a relationship between the number of queries and the number of queries, and thus the number of queries for each rare disease in a corresponding time period of a corresponding region may be predicted based on the number of queries for each rare disease in each time period of the corresponding region and the relationship.

In a possible implementation manner, the number of inquired people aiming at various rare diseases in each time period of each region can be predicted according to the population data of each region in each time period. Step S12 may include: determining the number ratio of the query times of various rare diseases in the rare disease query keyword database to the total query times in the corresponding time period of the corresponding region in a plurality of time periods of a plurality of regions in a query history database of the search engine; and determining the number of inquired people for the multiple rare diseases in multiple time periods of the multiple regions according to the frequency ratio and the population data in the corresponding time period of the corresponding region.

In one possible implementation, it may be assumed that the ratio of the number of people in a certain time period for a region to query rare disease information using a search engine to the total number of people in the time period for the region is the same as the ratio of the number of queries in the time period for the region to query rare disease information using a search engine to the total number of all query records in the time period for the region to use a search engine. The query times of the rare disease information in each time period of each region by using the search engine and the total number of query records (namely, the total query times) of the query records in each time period of each region by using the search engine can be obtained by a query history database of the search engine, and the time ratio of the query times to the total query records can be obtained immediately. The total population for each time period for each region may also be obtained by querying the population data, e.g., by querying the statistics for each region. After the data are obtained, the number of inquired people for various rare diseases in each time period of each area can be solved based on the frequency ratio and the population data of the corresponding area in the corresponding time period.

In an example, the number of queries for a plurality of rare diseases over a plurality of time periods for a plurality of regions can be determined according to the following equation (1):

wherein t represents a time period, l represents an area, d represents a rare disease type, x (t, l, d) represents the number of inquired people for the rare disease d, and q represents the time period t of the area l_d(t, l) represents the number of queries over a time period t of region l for rare disease d, q_all(t, l) represents the total number of query records (i.e., total number of queries) for the search engine over time period t for region l. p (t, l) represents the total population over the time period t of the region l.

In this way, the number of queries for rare diseases can be predicted based on the ratio of the number of queries for rare diseases to the total number of queries, and a data basis is provided for training a rare patient number prediction model.

In one possible implementation manner, in step S13, the number of newly-added diagnosed rare diseases in each time period in each region may be obtained, for example, the number of newly-added diagnosed rare diseases in each year or each month may be obtained from the medical and health department in each region. The number of confirmed persons is the actual number, the actual number and the number of inquired persons can be used for fitting, the regression model is used for fitting the actual number and the number of inquired persons, regression coefficients are obtained, and then the regression model, namely the rare patient number prediction model is obtained.

In a possible implementation manner, a training manner can be used to adjust parameters of the rare patient number prediction model, so as to obtain the trained rare patient number prediction model. Step S13 may include: respectively normalizing the inquired people number of the multiple rare diseases in multiple time periods of multiple regions and the diagnosed people number of the multiple rare diseases in multiple time periods of multiple regions to obtain inquired people number sample data and diagnosed people number sample data; determining a loss function of the rare patient quantity prediction model according to the sample data of the diagnosed number of patients and the predicted diagnosed number obtained by inputting the sample data of the inquired number of patients into the rare patient quantity prediction model; adjusting model parameters of the rare patient number prediction model on a training set according to the loss function and preset hyper-parameters; and under the condition that the training condition is met, obtaining the trained rare patient number prediction model.

In a possible implementation manner, the data may be refined first, for example, for a case that the number of diagnoses of a certain rare disease in a certain time period in a certain area is greater than 0, but the number of queries for the rare disease in the time period in the certain area is not recorded in the query, the number of queries for the rare disease in the time period in the certain area may be equal to 0. The similar process can be performed, for example, the number of inquired persons in a certain area for a certain rare disease is greater than 0 in a certain time period, but the number of confirmed cases in the certain area is not increased in the certain time period, and can also be equal to 0. Under the condition that the number of inquired people and the number of newly-added confirmed people are not recorded, the number of inquired people and the number of newly-added confirmed people can be both equal to 0. The data may also be denoised, for example, to remove significant data errors (e.g., some extremes).

In a possible implementation manner, normalization processing in rare diseases can be respectively carried out on the preprocessed inquired people number and confirmed people number, and normalized inquired people number sample data and confirmed people number sample data are obtained. In an example, the normalization process may be performed by the following equations (2) and (3):

wherein, t_iDenotes the ith time period, l_iDenotes the ith area, d_iIndicating the ith rare disease (i is a positive integer). T is a set of multiple time periods, L is a set of multiple regions, x (T)_i,l_i,d_i) To query the population sample data, y (t)_i,l_i,d_i) To confirm the diagnosis of the population sample data, x' (t)_i,l_i,d_i) Is normalized query population sample data, y' (t)_i,l_i,d_i) The sample data of the number of the diagnosed people after normalization.

In one possible implementation, the rare patient number prediction model may be set in the form of, for example, a rare patient number prediction model represented by the following equation (4):

wherein the content of the first and second substances,

the model parameters are all in the form of vectors,

and

representing the slope and intercept affected by the rare disease class, with dimensions equal to the total number of rare disease classes.

And

representing the slope and intercept affected by the region, whose dimensions are equal to the total number of regions. a is_tIndicating the change in the number of newly diagnosed persons over time. [ k ] A]The expression is taken to be the kth dimension in the vector, and k is any positive integer less than or equal to the vector dimension.

In one possible implementation, after the form of the rare patient number prediction model is set, the model can be trained based on the inquired number sample data and the confirmed number sample data. The method can obtain the sample data of the number of inquired people for various rare diseases in a plurality of time periods in a plurality of regions, divide the data into a training set, a verification set and a test set, train by using the data in the training set, verify by using the data in the verification set and test by using the data in the test set. The system can input the inquired number sample data of any rare disease in any time period in any region into the rare patient number prediction model, the model can output the predicted confirmed number, and further the loss function can be determined based on the difference between the output predicted confirmed number and the sample data of the confirmed number. In an example, the loss function may be determined according to equation (5) below:

where MSE is a loss function, y'_pred(t, l, d) is the predicted number of confirmed diagnoses output by the model. L | represents the size of the set L, i.e., the number of regions, | D | represents the size of the set D, i.e., the total number of rare disease types, | T | represents the size of the set T, i.e., the total number of slots.

In one possible implementation, the model parameters of the rare patient number prediction model may be adjusted by a loss function based on a preset hyper-parameter, for example, by a gradient descent method in a direction in which the loss function decreases. During the adjustment, the hyper-parameters may be preset, for example, learning rate, batch size, regularization norm, etc. may be set. In an example, the initial learning rate is set to 0.005 and decreases to 0.9 times the current learning rate every 50 rounds as the training round increases, the batch size is set to 16, and the regularization l2 norm is set to 1 e-6. The training may be performed according to these pre-set hyper-parameters and the model parameters adjusted to improve the model performance, i.e. to make the loss function gradually smaller in each training.

In one possible implementation, the query population sample data may be divided into a plurality of batches, for example, the query population sample data for each training round includes 1000 pieces, the query population sample data may be divided into a plurality of batches, each batch may include 20 pieces, and a batch of training samples is input into the rare patient number prediction model for training, so that 50 batches of data can be input for each training round.

In one possible implementation, training conditions can be set, and training is completed when the training conditions are met, so that a trained rare patient number prediction model is obtained. In an example, the training conditions may include a number of rounds condition, e.g., training is completed after the number of training rounds reaches a predetermined number. For another example, the training condition may include an error condition, e.g., training is completed when the loss function converges, or the loss function is less than or equal to a preset threshold. The present disclosure does not limit the training conditions.

In a possible implementation manner, other evaluation indexes can be used as the basis for completing the training. Obtaining the trained rare patient number prediction model under the condition that the training condition is met, wherein the obtaining comprises the following steps: determining the evaluation index of the rare patient number prediction model according to the confirmed patient number sample data of the verification set and the predicted confirmed patient number obtained by inputting the inquired patient number sample data into the rare patient number prediction model; and in the preset number of training rounds, calculating an evaluation index after each training round, and stopping model training under the condition that the evaluation index is unchanged after a plurality of training rounds to obtain the trained rare patient number prediction model.

In a possible implementation manner, after each round of training is completed, verification can be performed in a verification set, the verification set can also comprise a plurality of inquired people number sample data and confirmed people number sample data, the inquired people number sample data in the verification set can be input into a rare patient number prediction model, and the predicted confirmed people number is output. And determining an evaluation index based on a difference between the output predicted confirmed person number and the actual confirmed person number sample data. The evaluation index can be determined by the following formula (6):

wherein MRD is the evaluation index, i.e., the average relative difference.

In a possible implementation manner, in multiple rounds of verification, if the evaluation index is not reduced any more, namely is kept unchanged, the error in verification is considered to be small and is difficult to reduce, the training is stopped when the training target is reached, and the trained rare patient number prediction model is obtained.

In one possible implementation, after obtaining the trained rare patient number prediction model, the model can be used to predict the newly-increased diagnosed number of rare diseases in a certain region in a certain time period based on the number of the queries for the rare diseases in the certain region (for example, the time period not belonging to the set T, such as the latest certain time period).

In one possible implementation, the method further includes: determining the number of inquired people aiming at the preset type of rare diseases in a preset time period in a preset area in a query historical database of the search engine; inputting the trained rare disease patient number prediction model aiming at the number of inquired people with rare diseases of a preset type in a preset time period of the preset area, and obtaining first prediction data of the number of cases with rare diseases of the preset type in the preset time period of the preset area; and carrying out inverse normalization processing on the first prediction data to obtain the predicted confirmed diagnosis number of the preset type of rare diseases in a preset time period of a preset area.

In an example, the number of queries for a preset type of rare disease within a preset time period of a preset region may be determined in a query history database of a search engine, and the number of queries may be determined based on equation (1). Further, the number of inquired people can be normalized through a formula (2), and the trained prediction model of the number of the rare disease patients is input to obtain first prediction data of the number of cases of the preset type of rare diseases in a preset time period of a preset region. The first prediction data is data of normalized caliber, and inverse normalization processing can be carried out on the first prediction data to obtain the predicted confirmed diagnosis number of the preset type of rare diseases in the preset time period of the preset area.

In an example, the inverse normalization process may be performed by the following equation (7):

wherein, y'_pred(t, l, d) is the first prediction data, y_pred(t, l, d) is the predicted number of confirmed patients.

According to the rare disease patient number prediction model training method disclosed by the embodiment of the disclosure, the text of the name of the rare disease can be preprocessed according to the characteristics of rare occurrence and rare occurrence of the rare disease and the characteristic of less query data, the rare disease query keyword database is reasonably expanded, the omission of query data is reduced, and the accuracy of the data is improved. And the query number aiming at the rare diseases is predicted based on the ratio of the query times aiming at the rare diseases to the total query times, a data basis is provided for training the rare patient number prediction model, the rare patient number prediction model is trained based on the number and the actual diagnosed number, and the prediction precision of the model on the number of the rare patients is improved.

Fig. 2 is a schematic application diagram of the training method of the rare patient quantity prediction model according to the embodiment of the disclosure, as shown in fig. 2, 121 rare diseases are included in the catalog of the first batch of rare diseases, and the prediction capability of the rare patient quantity prediction model for the quantity of newly added cases of the 121 rare diseases can be trained. In an example, the number of new cases of various rare diseases among 31 provinces in 2016-2019 can also be queried.

In one possible implementation, the text of the rare disease names in the directory can be preprocessed to obtain a rare disease query keyword database, where for each rare disease, multiple titles for that rare disease can be included, and these titles can constitute a list of query keywords for that rare disease. Aiming at the keywords in the rare disease query keyword database, screening is carried out in a query history database of a search engine, and according to the formula (1), the number of related query people of 31 provinces aiming at multiple rare diseases in the year is 829403 in total, and 31 provinces of 121 diseases in 4 years in 2016-2019 are covered.

In one possible implementation, the data is first preprocessed during training, and in the example, the value of the data that is not recorded can be set to 0. And unreasonable extreme values are removed, and data of several rare diseases such as ' homocysteinemia ' and ' Parkinson's disease (juvenile type, early onset type) ' are unreasonable, so that the unreasonable extreme values can be removed.

In one possible implementation, 2016-. Wherein the validation set and test set retain the same data relating to regional and rare disease categories as in the training set. For example, there is no data related to "osteogenesis imperfecta (osteopathia)" in the training set, and if there is data related to "osteogenesis imperfecta (osteopathia)" in the validation set or the test set, the data needs to be deleted from the validation set or the test set, otherwise, the result of validation or test may be inaccurate.

In a possible implementation, after the preprocessing, 7085 pieces of data can be included, that is, the data amount of x (t, l, d) is 7085, and the data amount of the corresponding real number of confirmed persons is 7085. The data volume of the training set is 3589, the data volume of the verification set is 1741, and the data volume of the test set is 1755.

In one possible implementation, x (t, l, d) in the training set can be input into the rare patient number prediction model in the form of equation (4), the output predicted diagnosed number is obtained, and the loss function is determined based on equation (5), and the model parameters are adjusted. After each training round, verification can be performed in the verification set, for example, x (t, l, d) in the verification set can be input into the rare patient number prediction model to obtain the output predicted diagnosed number, and the evaluation index is determined based on formula (6), and if the evaluation index is not reduced in the verification of more than 50 training rounds, training can be completed to obtain the trained rare patient number prediction model.

In one possible implementation, the trained rare patient number prediction model may be tested in a test set. The test mode may include three, one: according to the data in the test set and the training set, the data of each rare disease in the training set is averaged, and the average relative difference (which can be determined based on formula (6)) of the prediction mode is 1.6749, wherein the prediction value is used as the number of the cases of the rare disease newly added in 2019. II, secondly: the average number of cases of the rare diseases in each region in the training set is used as a predicted value of the number of cases of the rare diseases newly added in each region in 2019, and the average relative difference of the prediction modes is 0.6329. Thirdly, the method comprises the following steps: the trained rare patient number prediction model is used for predicting the number of newly-increased cases of various rare diseases in each region in 2019, the average relative difference of the prediction modes is 0.4108, and the prediction modes are smaller than the two modes, namely the prediction accuracy of the trained rare patient number prediction model is higher.

In one possible implementation, the distribution of the relative differences of the three test patterns may also be compared. The proportion of the data amount in which the relative difference in the first mode is less than 0.1 is 7.69%, the proportion of the data amount in which the relative difference in the second mode is less than 0.1 is 14.81%, and the proportion of the data amount in which the relative difference in the third mode is less than 0.1 is 37.32%. The first mode has a ratio of the data amount in the interval [0.1, 0.2) of 7.12% relative difference, the second mode has a ratio of the data amount in the interval [0.1, 0.2) of 6.84% relative difference, and the third mode has a ratio of the data amount in the interval [0.1, 0.2) of 20.97% relative difference. The first mode has a ratio of 7.52% in the data amount in the interval [0.2, 0.3), the second mode has a ratio of 8.43% in the data amount in the interval [0.2, 0.3), and the third mode has a ratio of 10.60% in the data amount in the interval [0.2, 0.3). The first mode has a ratio of the data amount in the interval [0.3, 0.5) of 14.88% relative difference, the second mode has a ratio of the data amount in the interval [0.3, 0.5) of 14.25% relative difference, and the third mode has a ratio of the data amount in the interval [0.3, 0.5) of 9.86% relative difference. The proportion of the amount of data with a relative difference of more than 0.5 in the first mode is 62.79%, the proportion of the amount of data with a relative difference of more than 0.5 in the second mode is 55.67%, and the proportion of the amount of data with a relative difference of more than 0.5 in the third mode is 21.25%. The method comprises the following steps of calculating the average relative difference of the rare patient quantity prediction model, calculating the variance of the rare patient quantity prediction model, and calculating the average relative difference of the rare patient quantity prediction model and the variance of the rare patient quantity prediction model.

Fig. 3 shows a block diagram of a rare patient number prediction model training apparatus according to an embodiment of the present disclosure, as shown in fig. 3, the apparatus comprising: the text preprocessing module 11 is used for preprocessing texts with various rare disease names to obtain a rare disease query keyword database; the query number determining module 12 is used for determining the query number of the plurality of rare diseases in a plurality of time periods in a plurality of regions according to the rare disease query keyword database and the query history database of the search engine; and the training module 13 is used for training the rare patient quantity prediction model according to the number of inquired people of the rare diseases in the time periods of the regions and the number of diagnosed people of the rare diseases in the time periods of the regions, so as to obtain the trained rare patient quantity prediction model.

It is understood that the above-mentioned method embodiments of the present disclosure can be combined with each other to form a combined embodiment without departing from the logic of the principle, which is limited by the space, and the detailed description of the present disclosure is omitted. Those skilled in the art will appreciate that in the above methods of the specific embodiments, the specific order of execution of the steps should be determined by their function and possibly their inherent logic.

In addition, the disclosure also provides a training device of the rare patient number prediction model, an electronic device, a computer-readable storage medium, and a program, which can be used to implement any one of the training methods of the rare patient number prediction model provided by the disclosure, and the corresponding technical solutions and descriptions and corresponding descriptions in the method section are omitted for brevity.

In some embodiments, functions of or modules included in the apparatus provided in the embodiments of the present disclosure may be used to execute the method described in the above method embodiments, and specific implementation thereof may refer to the description of the above method embodiments, and for brevity, will not be described again here.

Embodiments also provide a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the above-described method. The computer readable storage medium may be a non-volatile computer readable storage medium.

An embodiment of the present disclosure further provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the memory-stored instructions to perform the above-described method.

The disclosed embodiments also provide a computer program product comprising computer readable code that, when run on a device, a processor in the device executes instructions for implementing the rare patient population prediction model training method provided by any of the above embodiments.

The disclosed embodiments also provide another computer program product for storing computer readable instructions that, when executed, cause a computer to perform the operations of the rare patient population prediction model training method provided by any of the above embodiments.

The electronic device may be provided as a terminal, server, or other form of device.

Fig. 4 illustrates a block diagram of an electronic device 800 in accordance with an embodiment of the disclosure. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, or the like terminal.

Referring to fig. 4, electronic device 800 may include one or more of the following components: processing component 802, memory 804, power component 806, multimedia component 808, audio component 810, input/output (I/O) interface 812, sensor component 814, and communication component 816.

The processing component 802 generally controls overall operation of the electronic device 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing components 802 may include one or more processors 820 to execute instructions to perform all or a portion of the steps of the methods described above. Further, the processing component 802 can include one or more modules that facilitate interaction between the processing component 802 and other components. For example, the processing component 802 can include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.

The memory 804 is configured to store various types of data to support operations at the electronic device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phonebook data, messages, pictures, videos, and so forth. The memory 804 may be implemented by any type or combination of volatile or non-volatile memory devices such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disks.

The power supply component 806 provides power to the various components of the electronic device 800. The power components 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the electronic device 800.

The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and a user. In some embodiments, the screen may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive an input signal from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense an edge of a touch or slide action, but also detect a duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front facing camera and/or a rear facing camera. The front camera and/or the rear camera may receive external multimedia data when the electronic device 800 is in an operation mode, such as a shooting mode or a video mode. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

The audio component 810 is configured to output and/or input audio signals. For example, the audio component 810 includes a Microphone (MIC) configured to receive external audio signals when the electronic device 800 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may further be stored in the memory 804 or transmitted via the communication component 816. In some embodiments, audio component 810 also includes a speaker for outputting audio signals.

The I/O interface 812 provides an interface between the processing component 802 and peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to: a home button, a volume button, a start button, and a lock button.

The sensor assembly 814 includes one or more sensors for providing various aspects of state assessment for the electronic device 800. For example, the sensor assembly 814 may detect an open/closed state of the electronic device 800, the relative positioning of components, such as a display and keypad of the electronic device 800, the sensor assembly 814 may also detect a change in the position of the electronic device 800 or a component of the electronic device 800, the presence or absence of user contact with the electronic device 800, orientation or acceleration/deceleration of the electronic device 800, and a change in the temperature of the electronic device 800. Sensor assembly 814 may include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 may access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 further includes a Near Field Communication (NFC) module to facilitate short-range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

In an exemplary embodiment, the electronic device 800 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, micro-controllers, microprocessors or other electronic components for performing the above-described methods.

In an exemplary embodiment, a non-transitory computer-readable storage medium, such as the memory 804, is also provided that includes computer program instructions executable by the processor 820 of the electronic device 800 to perform the above-described methods.

Fig. 5 illustrates a block diagram of an electronic device 1900 in accordance with an embodiment of the disclosure. For example, the electronic device 1900 may be provided as a server. Referring to fig. 5, electronic device 1900 includes a processing component 1922 further including one or more processors and memory resources, represented by memory 1932, for storing instructions, e.g., applications, executable by processing component 1922. The application programs stored in memory 1932 may include one or more modules that each correspond to a set of instructions. Further, the processing component 1922 is configured to execute instructions to perform the above-described method.

The electronic device 1900 may also include a power component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input/output (I/O) interface 1958. The electronic device 1900 may operate based on an operating system, such as Windows Server, stored in memory 1932^TM，Mac OS X^TM，Unix^TM,Linux^TM，FreeBSD^TMOr the like.

In an exemplary embodiment, a non-transitory computer readable storage medium, such as the memory 1932, is also provided that includes computer program instructions executable by the processing component 1922 of the electronic device 1900 to perform the above-described methods.

The present disclosure may be systems, methods, and/or computer program products. The computer program product may include a computer-readable storage medium having computer-readable program instructions embodied thereon for causing a processor to implement various aspects of the present disclosure.

The computer readable storage medium may be a tangible device that can hold and store the instructions for use by the instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic memory device, a magnetic memory device, an optical memory device, an electromagnetic memory device, a semiconductor memory device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a Static Random Access Memory (SRAM), a portable compact disc read-only memory (CD-ROM), a Digital Versatile Disc (DVD), a memory stick, a floppy disk, a mechanical coding device, such as punch cards or in-groove projection structures having instructions stored thereon, and any suitable combination of the foregoing. Computer-readable storage media as used herein is not to be construed as transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission medium (e.g., optical pulses through a fiber optic cable), or electrical signals transmitted through electrical wires.

The computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to a respective computing/processing device, or to an external computer or external storage device via a network, such as the internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers and/or edge servers. The network adapter card or network interface in each computing/processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in the respective computing/processing device.

The computer program instructions for carrying out operations of the present disclosure may be assembler instructions, Instruction Set Architecture (ISA) instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider). In some embodiments, the electronic circuitry that can execute the computer-readable program instructions implements aspects of the present disclosure by utilizing the state information of the computer-readable program instructions to personalize the electronic circuitry, such as a programmable logic circuit, a Field Programmable Gate Array (FPGA), or a Programmable Logic Array (PLA).

Various aspects of the present disclosure are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the disclosure. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer-readable program instructions.

These computer-readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer-readable program instructions may also be stored in a computer-readable storage medium that can direct a computer, programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer-readable medium storing the instructions comprises an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks.

The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer, other programmable apparatus or other devices implement the functions/acts specified in the flowchart and/or block diagram block or blocks.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The computer program product may be embodied in hardware, software or a combination thereof. In an alternative embodiment, the computer program product is embodied in a computer storage medium, and in another alternative embodiment, the computer program product is embodied in a Software product, such as a Software Development Kit (SDK), or the like.

Having described embodiments of the present disclosure, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen in order to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for training a rare patient number prediction model, comprising:

preprocessing texts of various rare disease names to obtain a rare disease query keyword database;

determining the number of inquired people for the multiple rare diseases in multiple time periods in multiple regions according to the rare disease inquiry keyword database and the inquiry history database of the search engine;

and training the rare patient quantity prediction model according to the number of inquired people of the rare diseases in the time periods of the regions and the number of diagnosed people of the rare diseases in the time periods of the regions, so as to obtain the trained rare patient quantity prediction model.

2. The method of claim 1, wherein preprocessing the text of the plurality of rare disease names comprises at least one of:

preprocessing symbol texts in the texts of the rare disease names;

preprocessing the alphabetic text in the text of the rare disease name;

preprocessing the digital text in the text of the rare disease name;

preprocessing preset characters in the text of the rare disease name;

preprocessing abbreviated texts in the texts of the rare disease names;

pre-processing transliterated text in the text of the rare disease name.

3. The method of claim 1, wherein preprocessing the text of the plurality of rare disease names to obtain a rare disease query keyword database comprises:

adding the text of the plurality of rare disease names and the preprocessed text to the rare disease query keyword database.

4. The method of claim 1, wherein determining the number of queried people for a plurality of rare diseases over a plurality of time periods in a plurality of regions from the database of rare disease query terms and a query history database of a search engine comprises:

determining the number ratio of the query times of various rare diseases in the rare disease query keyword database to the total query times in the corresponding time period of the corresponding region in a plurality of time periods of a plurality of regions in a query history database of the search engine;

and determining the number of inquired people for the multiple rare diseases in multiple time periods of the multiple regions according to the frequency ratio and the population data in the corresponding time period of the corresponding region.

5. The method of claim 1, wherein training the rare patient number prediction model based on the number of inquired rare patients in the plurality of time periods in the plurality of regions and the number of diagnosed rare diseases in the plurality of time periods in the plurality of regions to obtain the trained rare patient number prediction model comprises:

respectively normalizing the inquired people number of the multiple rare diseases in multiple time periods of multiple regions and the diagnosed people number of the multiple rare diseases in multiple time periods of multiple regions to obtain inquired people number sample data and diagnosed people number sample data;

determining a loss function of the rare patient quantity prediction model according to the sample data of the diagnosed number of patients and the predicted diagnosed number obtained by inputting the sample data of the inquired number of patients into the rare patient quantity prediction model;

adjusting model parameters of the rare patient number prediction model on a training set according to the loss function and preset hyper-parameters;

and under the condition that the training condition is met, obtaining the trained rare patient number prediction model.

6. The method of claim 5, wherein obtaining the trained rare patient population prediction model if training conditions are met comprises:

determining the evaluation index of the rare patient number prediction model according to the confirmed patient number sample data of the verification set and the predicted confirmed patient number obtained by inputting the inquired patient number sample data into the rare patient number prediction model;

and in the preset number of training rounds, calculating an evaluation index after each training round, and stopping model training under the condition that the evaluation index is unchanged after a plurality of training rounds to obtain the trained rare patient number prediction model.

7. The method of claim 1, further comprising:

determining the number of inquired people aiming at the preset type of rare diseases in a preset time period in a preset area in a query historical database of the search engine;

normalizing the number of inquired people of the rare diseases of a preset type in a preset time period of the preset area, inputting the trained rare patient number prediction model, and obtaining first prediction data of the number of cases of the rare diseases of the preset type in the preset time period of the preset area;

and carrying out inverse normalization processing on the first prediction data to obtain the predicted confirmed diagnosis number of the preset type of rare diseases in a preset time period of a preset area.

8. A training apparatus for a rare patient number prediction model, comprising:

the text preprocessing module is used for preprocessing texts with various rare disease names to obtain a rare disease query keyword database;

the query number determining module is used for determining the query number of the plurality of rare diseases in a plurality of time periods in a plurality of regions according to the rare disease query keyword database and the query history database of the search engine;

and the training module is used for training the rare patient quantity prediction model according to the number of inquired people of the rare diseases in the time periods of the regions and the number of diagnosed people of the rare diseases in the time periods of the regions to obtain the trained rare patient quantity prediction model.

9. An electronic device, comprising:

a processor;

a memory for storing processor-executable instructions;

wherein the processor is configured to invoke the memory-stored instructions to perform the method of any of claims 1 to 7.

10. A computer readable storage medium having computer program instructions stored thereon, which when executed by a processor implement the method of any one of claims 1 to 7.