Production process traceability method and system for icodextrin peritoneal dialysis fluid

Abstract

The present invention relates to the field of production process management, and in particular to a production process tracing method and system for icodextrin peritoneal dialysis fluid. The method first obtains the production files of the department to be tested of the icodextrin peritoneal dialysis fluid and the generation time of each production file, obtains the production density of each phrase according to the difference in word vectors between each phrase and the number of occurrences of each phrase in the production file, screens out key production files according to the difference in word vectors between each production file and the difference in generation time and the production density of phrases in the production file, and then clusters the sentences in the key production files to obtain multiple sentence clusters, and traces the production process of the department to be tested according to the difference in word vectors between each phrase in the sentences of the sentence clusters and the production density of the phrases. The present invention can timely and accurately trace the problems in the production process of icodextrin peritoneal dialysis fluid.

Description

Technical Field

[0001] The present invention relates to the field of production process management, and in particular to a production process tracing method and system for icodextrin peritoneal dialysis fluid. Background Art

[0002] Icodextrin peritoneal dialysis fluid is a drug used for peritoneal dialysis treatment. Icodextrin is a glucose polymer that can provide stable osmotic pressure, helping to maintain the stability of the osmotic pressure of the peritoneal dialysis fluid. The production process of icodextrin peritoneal dialysis fluid involves multiple operation links. Any error in each operation link will affect its quality. Therefore, the production process needs to be traced to ensure its production quality.

[0003] In the related art, various documents generated in the production process of icodextrin peritoneal dialysis fluid are usually analyzed to trace the problems that occur in the production process. However, since the production process of icodextrin peritoneal dialysis fluid involves multiple departments, and the number of production documents generated by each department is large and complex, the information bias of production documents of different departments is different, resulting in the existing method being unable to accurately trace problems in the production process of icodextrin peritoneal dialysis fluid. Summary of the invention

[0004] In order to solve the technical problem that the existing methods cannot accurately trace the production process of icodextrin peritoneal dialysis solution when problems occur, the purpose of the present invention is to provide a production process tracing method and system for icodextrin peritoneal dialysis solution, and the technical scheme adopted is as follows:

[0005] The present invention provides a production process tracing method for icodextrin peritoneal dialysis solution, the method comprising:

[0006] Obtain the production files of the department to be tested of icodextrin peritoneal dialysis fluid within a preset time period and the generation time of each production file; perform natural language processing on all production files to obtain multiple sentences and phrases of the production files and the word vector of each phrase;

[0007] According to the difference of the word vectors between each word group and the number of occurrences of each word group in all production files, the production density of each word group is obtained; according to the difference of the word groups between each production file and the difference of the generation time, and the production density of the word groups in each production file, the importance of each production file is obtained; based on the importance, the key production files of the department to be tested are screened out from all production files;

[0008] According to the word vectors and the production closeness of each phrase in each sentence of the key production file, a reference vector of each sentence of the key production file is obtained; according to the difference of the reference vectors of each sentence of the key production file, a similarity parameter of each sentence of the key production file is obtained; based on the similarity parameter, the sentences in the key production file are clustered to obtain a plurality of sentence clusters; according to the difference of the word vectors between each phrase in the sentence of each sentence cluster and the production closeness of each phrase, an abnormal coefficient of each phrase in the sentence of each sentence cluster is obtained;

[0009] Based on the abnormal coefficient of each phrase in the key production documents, the production process of the department to be tested is traced.

[0010] Further, obtaining the production density of each phrase includes:

[0011] Taking any phrase as a target phrase, negatively correlating the Euclidean distances of the word vectors between the target phrase and each other phrase except the target phrase, and obtaining the initial phrase similarity between the target phrase and each other phrase;

[0012] Taking the accumulated value of the initial phrase similarities between the target phrase and all other phrases as the comprehensive phrase similarity of the target phrase;

[0013] The number of occurrences of the target phrase in all production documents and the comprehensive phrase similarity are integrated and normalized to obtain the production density of the target phrase.

[0014] Furthermore, the importance of obtaining each production file includes:

[0015] Taking the average of the production closeness of all phrases in each production file as the overall closeness of each production file;

[0016] Based on the overall compactness, all production files are clustered to obtain a plurality of first file clusters, and according to the difference in the generation time of the production files in each first file cluster, a periodicity parameter of each first file cluster is obtained;

[0017] The set of all phrases in each production file is taken as the phrase set of each production file, any production file is taken as the target production file, the Jaccard similarity coefficient of the phrase set between the target production file and each other production file except the target production file is taken as the initial content similarity between the target production file and each other production file; the cumulative value of the initial content similarities between the target production file and all other production files is taken as the comprehensive content similarity of the target production file;

[0018] The overall closeness of the target production file, the comprehensive content similarity and the periodic parameter of the first file cluster where the target production file is located are integrated and normalized to obtain the importance of the target production file.

[0019] Further, the obtaining of the periodic parameters of each first file cluster includes:

[0020] For any first file cluster, the production files in the first file cluster are sorted according to the generation time to obtain a production file sequence of the first file cluster;

[0021] Taking two adjacent production files in the production file sequence as a production file group, and taking the absolute value of the difference between the generation time of the two production files in each production file group as the time interval of each production file group;

[0022] After analyzing the discrete degrees of the time intervals of all production file groups and performing negative correlation mapping, the periodicity parameters of the first file clustering cluster are obtained.

[0023] Furthermore, based on the importance, the key production documents of the department to be tested are screened out from all production documents, including:

[0024] Based on the importance, clustering all production files to obtain multiple second file clusters, and taking the average of the importance of all production files in each second file cluster as the cluster center of each second file cluster;

[0025] The production files in the second file cluster corresponding to the maximum value of the cluster center are used as key production files of the department to be tested.

[0026] Furthermore, the step of obtaining a reference vector for each statement of a key production file includes:

[0027] Taking any sentence in the key production file as the target sentence, using the production density of each phrase in the target sentence, weighting the word vector of each phrase in the target sentence, and obtaining an adjusted word vector of each phrase in the target sentence;

[0028] The average value of the adjusted word vectors of all phrases in the target sentence is used as the reference vector of the target sentence.

[0029] Furthermore, the obtaining of similarity parameters of each sentence of the key production file includes:

[0030] In the key production file, the average value of the cosine similarities of the reference vectors between the target sentence and all other sentences except the target sentence is used as the similarity parameter of the target sentence.

[0031] Furthermore, the step of obtaining the abnormal coefficient of each phrase in each sentence cluster includes:

[0032] Taking any sentence clustering cluster as the target sentence clustering cluster, taking any phrase in the sentence of the target sentence clustering cluster as the phrase to be tested, and taking the accumulated value of the Euclidean distance between the word vectors of the phrase to be tested and all other phrases in the sentence of the target sentence clustering cluster except the phrase to be tested as the initial abnormal parameter of the phrase to be tested;

[0033] The initial abnormality parameter is weighted and normalized by using the production density of the phrase to be tested to obtain the abnormality coefficient of the phrase to be tested.

[0034] Furthermore, the tracing of the production process of the department to be tested based on the abnormal coefficient of each phrase in the key production file includes:

[0035] If the abnormal coefficient of each phrase in the key production file is greater than a preset abnormal threshold, an abnormal warning is issued.

[0036] The present invention also proposes a production process tracing system for icodextrin peritoneal dialysis fluid, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the steps of a production process tracing method for icodextrin peritoneal dialysis fluid when executing the computer program.

[0037] The present invention has the following beneficial effects:

[0038] The present invention takes into account that the existing methods cannot accurately trace the production process of icodextrin peritoneal dialysis fluid when problems occur. Therefore, the production files of the department to be tested of the icodextrin peritoneal dialysis fluid within a preset time period and the generation time of each production file are first obtained. Considering that different phrases in the production files have different correlations with the production process of the icodextrin peritoneal dialysis fluid, the degree of correlation between each phrase and the preparation and production of the icodextrin peritoneal dialysis fluid is first reflected by the production closeness, so as to facilitate the subsequent analysis of the importance of the production files based on the production closeness. At the same time, in the subsequent calculation, the production closeness can be used as a weight to improve the accuracy of the calculation. Since there are still some redundant files in the production files with a low correlation with the production process of the icodextrin peritoneal dialysis fluid, and the internal correlation between the important production files is not large, the correlation between the important production files is relatively large. The contents of the documents are relatively similar, and the time intervals of their generation are relatively fixed. Therefore, the key production documents of the department to be tested can be accurately screened out by the importance obtained, and the interference of redundant documents can be removed. Considering that if the differences of the entire files are directly compared, it may be difficult to trace the problem due to the overall similarity of the files, the reference vector of each sentence in the key production file is first obtained, and then the similarity parameter is used to reflect the similarity between each sentence in the key production file and other sentences, and the sentences with similarity parameters are clustered. By comparing the differences in the word vectors between the phrases in the sentences of the sentence clustering cluster, and reflecting the possibility of abnormalities in the operation process represented by each phrase through the abnormality coefficient, the problems in the production process of icodextrin peritoneal dialysis fluid are traced back in a timely and accurate manner based on the abnormality coefficient. BRIEF DESCRIPTION OF THE DRAWINGS

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

[0040] Figure 1 A flow chart of a production process tracing method for icodextrin peritoneal dialysis fluid provided by one embodiment of the present invention;

[0041] Figure 2 A flow chart of a method for obtaining the importance of each production file provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the production process tracing method and system of an icodextrin peritoneal dialysis solution proposed according to the present invention, its specific implementation, structure, characteristics and effects as follows. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics of one or more embodiments can be combined in any suitable form.

[0043] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0044] The specific scheme of the production process tracing method and system of icodextrin peritoneal dialysis fluid provided by the present invention is described in detail below with reference to the accompanying drawings.

[0045] See also Figure 1 , which shows a flow chart of a production process tracing method of icodextrin peritoneal dialysis solution provided by one embodiment of the present invention, the method comprising:

[0046] Step S1: Obtain the production files of the department to be tested of icodextrin peritoneal dialysis fluid within a preset time period and the generation time of each production file; perform natural language processing on all production files to obtain multiple sentences and phrases of the production files and the word vector of each phrase.

[0047] The entire production process of icodextrin peritoneal dialysis fluid usually involves multiple departments, such as a raw material department, a storage department, and various departments involved in the specific production process. During operation, each department will record detailed information of the specific operation process in the form of a text file, such as a production record file, a test result file, and an equipment maintenance record file. Due to the different responsibilities of different departments, the information bias of the files generated by each department is different. That is to say, the information in the file generated by each department has the characteristics of the responsibilities of the department. Therefore, the embodiment of the present invention first selects any department in the production process of icodextrin peritoneal dialysis fluid as the department to be tested, and obtains multiple production files generated by the department to be tested within a preset time period during the operation process and the generation time of each production file, wherein the preset time period is set to one week, and the specific value of the preset time period can also be set by the implementer according to the specific implementation scenario, which is not limited here.

[0048] Since production files are usually recorded in text form, and the information in the production files is usually reflected in sentences and phrases, the embodiment of the present invention performs natural language processing on all production files, divides the production files into sentences and words, and removes common but meaningless stop words or modal particles, etc., to obtain multiple sentences and phrases in the production files. At the same time, the GloVe algorithm or Word2Vec algorithm in natural language processing is used to quantify each phrase to obtain the word vector of each phrase, which is convenient for subsequent calculation and analysis. Natural language processing is a technical means well known to those skilled in the art and will not be elaborated here.

[0049] Step S2: According to the difference in word vectors between each phrase and the number of occurrences of each phrase in all production files, the production density of each phrase is obtained; according to the difference in phrases and the difference in generation time between production files, and the production density of phrases in each production file, the importance of each production file is obtained; based on the importance, multiple key production files of the department to be tested are screened out from all production files.

[0050] Since different phrases in the production documents have different correlations with the production process of icodextrin peritoneal dialysis fluid, and phrases with a high correlation with the production and preparation process of icodextrin peritoneal dialysis fluid are similar to other phrases and appear frequently in the production documents, the differences in word vectors between phrases and the number of occurrences of each phrase in all production documents can be analyzed. The obtained production closeness reflects the degree of correlation between each phrase and the production of icodextrin peritoneal dialysis fluid. The greater the production closeness, the stronger the correlation between the phrase and the production process of icodextrin peritoneal dialysis fluid, which is convenient for subsequent analysis of the importance of production documents based on production closeness. At the same time, in subsequent calculations, production closeness can be used as a weight to improve the accuracy of calculations.

[0051] Preferably, in one embodiment of the present invention, the method for obtaining the production density of each phrase specifically includes:

[0052] Take any phrase as the target phrase, perform negative correlation mapping on the Euclidean distance between the target phrase and every other phrase except the target phrase, obtain the initial phrase similarity between the target phrase and every other phrase, and take the cumulative value of the initial phrase similarity between the target phrase and all other phrases as the comprehensive phrase similarity of the target phrase. The greater the comprehensive phrase similarity, the more similar the target phrase is to other phrases.

[0053] In other embodiments of the present invention, Manhattan distance may also be used as a distance measure between the target phrase and other phrases, which is not limited here.

[0054] Then, the number of occurrences of the target phrase in all production documents and the comprehensive phrase similarity are integrated and normalized to obtain the production density of the target phrase.

[0055] In the embodiment of the present invention, the combination of the two can be achieved by calculating the sum or product of the number of occurrences of the target phrase in all production documents and the similarity of the comprehensive phrase, which is not limited here.

[0056] In one embodiment of the present invention, the normalization processing can be specifically, for example, maximum and minimum value normalization processing, and the normalization in subsequent steps can all adopt maximum and minimum value normalization processing. In other embodiments of the present invention, other normalization methods can be selected according to the specific range of numerical values, which will not be described in detail.

[0057] As an example, in one embodiment of the present invention, the expression of the production closeness of the target phrase may be specifically, for example, as follows:

[0058]

[0059] in, Indicates the production tightness of the target phrase; represents the number of occurrences of the target phrase in all production documents; Indicates the target phrase and the first The Euclidean distance between word vectors of other word groups; Indicates the number of phrases other than the target phrase; Indicates the target phrase and The similarity of the initial phrase to other phrases; represents the comprehensive phrase similarity of the target phrase; represents the normalization function; Indicates the preset first adjustment parameter, which is used to prevent the denominator from being 0. The value range is In one embodiment of the present invention, Set to 0.01, The specific value of can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0060] It should be noted that in other embodiments of the present invention, negative correlation mapping may be achieved through other basic mathematical operations, which will not be elaborated herein.

[0061] The production density of each phrase can be obtained by the same method as above.

[0062] Since there are still some redundant files in the production documents that have a low correlation with the production process of icodextrin peritoneal dialysis fluid, in order to be able to trace and troubleshoot problems in the production process in a timely and accurate manner in the future, these redundant files need to be removed and important files with a high correlation with the production process of icodextrin peritoneal dialysis fluid need to be retained. The contents of important production files are relatively similar, and the intervals between their generation times are relatively fixed. Therefore, the differences in phrases and generation times between production files can be analyzed. At the same time, the greater the production closeness of a phrase in a production file, the more important the production file is. Therefore, the production closeness of the phrases in each production file can be combined to reflect the importance of each production file by obtaining the importance level. The greater the importance level, the more important the production file is in the production process of icodextrin peritoneal dialysis fluid. Subsequently, key production files can be screened out based on the importance level to improve the efficiency of subsequent problem tracing.

[0063] Preferably, in one embodiment of the present invention, the method for obtaining the importance of each production file specifically includes:

[0064] See also Figure 2 , which shows a flow chart of a method for obtaining the importance of each production file provided by an embodiment of the present invention.

[0065] Step S201: taking the average of the production closeness of all phrases in each production file as the overall closeness of each production file.

[0066] From the above analysis, it can be seen that the greater the production closeness of the phrases in each production file, the more important the production file is. Therefore, the average value of the production closeness of all phrases in each production file is taken as the overall closeness of each production file. The importance of each production file can be accurately calculated and analyzed based on the overall closeness.

[0067] Step S202: based on the overall compactness, cluster all production files to obtain a plurality of first file clusters, and obtain the periodicity parameter of each first file cluster according to the difference in the generation time of the production files in each first file cluster.

[0068] Since the intervals of the generation time of important production files are relatively fixed, that is, the generation time of important production files has a certain periodicity, and the overall closeness can preliminarily reflect the importance of each production file, all production files can be clustered based on the overall closeness to obtain multiple first file clustering clusters, so that production files with similar importance can be divided into the same clustering cluster, where the clustering algorithm can use the existing K-means algorithm or other clustering algorithms. In the clustering process, the Euclidean distance or Manhattan distance between the overall closeness of each production file can be used as the distance measure of each production file.

[0069] Furthermore, the differences in the generation time of the production files in each first file clustering cluster can be analyzed, and the obtained periodicity parameters can reflect the degree of periodicity of each production file in each first file clustering cluster. The larger the periodicity parameter, the more periodic the generation of each production file in the first file clustering cluster has a certain periodicity, which further indicates that the production files in the first file clustering cluster are more important.

[0070] Preferably, in one embodiment of the present invention, the method for obtaining the periodic parameters of each first file cluster specifically includes:

[0071] For any first file clustering cluster, the production files in the first file clustering cluster are sorted according to the generation time to obtain the production file sequence of the first file clustering cluster. In an embodiment of the present invention, the production files can be sorted in the order of production time from large to small or from small to large, which is not limited here.

[0072] Two adjacent production files in the production file sequence are taken as a production file group, and the absolute value of the difference between the generation time of the two production files in each production file group is taken as the time interval of each production file group. Then, the discrete degree of the time intervals of all production file groups can be analyzed and negative correlation mapping can be performed to obtain the periodic parameters of the first file clustering cluster. It should be noted that if there is only one production file in the first file clustering cluster, it means that the production file in the first file clustering cluster is basically irrelevant to the production process. At this time, the periodic parameter of the first clustering cluster can be directly set to the value 0 to facilitate the smooth progress of subsequent calculations.

[0073] In the embodiment of the present invention, the analysis of the degree of dispersion of the time intervals of all production file groups can be achieved by calculating the variance or standard deviation of the time intervals of all production file groups, which is not limited here.

[0074] As an example, in one embodiment of the present invention, the expression of the periodicity parameter of each first file cluster may be specifically, for example, as follows:

[0075]

[0076] in, Indicates The periodicity parameters of the first file clusters; Indicates The variance of the time intervals of all production file groups of the first file cluster; Indicates the preset second adjustment parameter, which is used to prevent the denominator from being 0. The value range is In one embodiment of the present invention, Set to 0.01, The specific value of can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0077] It should be noted that in other embodiments of the present invention, negative correlation mapping may be achieved through other basic mathematical operations, which will not be elaborated herein.

[0078] Step S203: take the set consisting of all phrases in each production file as the phrase set of each production file, take any production file as the target production file, take the Jaccard similarity coefficient of the phrase set between the target production file and each other production file except the target production file as the initial content similarity between the target production file and each other production file; take the cumulative value of the initial content similarities between the target production file and all other production files as the comprehensive content similarity of the target production file.

[0079] Since the contents between important production files have a certain degree of standardization, that is, the contents between important production files are relatively similar, and the more consistent the phrases in each production file, the more similar the contents between the production files are, so the set of all phrases in each production file can be used as the phrase set of each production file, any production file can be used as the target production file, and the Jaccard similarity coefficient of the phrase set between the target production file and each other production file except the target production file is used as the initial content similarity between the target production file and each other production file. The greater the initial content similarity, the more similar the target production file is to the content of each other production file, and then the cumulative value of the initial content similarity between the target production file and all other production files is used as the comprehensive content similarity of the target production file, wherein the Jaccard similarity coefficient is a technical means well known to those skilled in the art and will not be elaborated here.

[0080] As an example, in one embodiment of the present invention, the expression of the comprehensive content similarity of the target production file may be specifically, for example, as follows:

[0081]

[0082] in, Indicates the comprehensive content similarity of the target production file; Indicates the target production file and the The Jaccard similarity coefficient of the phrase set between the target production file and the other production files, that is, The initial content similarity of other production files; Indicates the number of production files other than the target production file.

[0083] Step S204: The overall closeness of the target production file, the comprehensive content similarity, and the periodic parameters of the first file cluster where the target production file is located are integrated and normalized to obtain the importance of the target production file.

[0084] Then, the overall closeness of the target production file, the comprehensive content similarity and the periodic parameters of the first file cluster where the target production file is located can be integrated and normalized to obtain the importance of the target production file.

[0085] In an embodiment of the present invention, the integration of the three can be achieved by calculating the overall closeness of the target production file, the comprehensive content similarity and the sum or product of the periodic parameters of the first file cluster where the target production file is located, which is not limited here.

[0086] As an example, in one embodiment of the present invention, the expression of the importance of the target production file may be specifically, for example, as follows:

[0087]

[0088] in, Indicates the importance of the target production document; Indicates the overall compactness of the target production file; Indicates the comprehensive content similarity of the target production file; Indicates the periodicity parameter of the first file cluster where the target production file is located; Represents the normalization function.

[0089] The importance of each production document can be obtained through the same method as above, and then based on the importance, multiple key production documents of the department to be tested can be screened out from all production documents, thereby removing redundant documents and improving the accuracy of subsequent production process traceability.

[0090] Preferably, in one embodiment of the present invention, the method for obtaining multiple key production files of the department to be tested specifically includes:

[0091] Based on the importance, all production files are clustered to obtain multiple second file clusters, wherein the clustering algorithm can use the existing K-means algorithm or other clustering algorithms, and in the clustering process, the Euclidean distance or Manhattan distance between the importance of each production file can be used as the distance measure of each production file.

[0092] Then, the average value of the importance of all production files in each second file cluster is used as the cluster center of each second file cluster. The larger the cluster center is, the more important the production files in the second file cluster are. Therefore, the production files in the second file cluster corresponding to the maximum value of the cluster center can be used as the key production files of the department to be tested.

[0093] In other embodiments of the present invention, a threshold may be set to select production files with an importance greater than the threshold as key production files of the department to be tested, which is not limited here.

[0094] At this point, the key production documents of the department to be tested have been obtained.

[0095] Step S3: According to the word vectors and production closeness of each phrase in each sentence of the key production file, obtain the reference vector of each sentence of the key production file; according to the difference of the reference vectors of each sentence of the key production file, obtain the similarity parameter of each sentence of the key production file; based on the similarity parameter, cluster the sentences in the key production file to obtain multiple sentence clustering clusters; according to the difference of the word vectors between the phrases in the sentences of each sentence clustering cluster and the production closeness of each phrase, obtain the abnormal coefficient of each phrase in the sentences of each sentence clustering cluster.

[0096] Considering that if the differences between key production files are directly compared, it may be difficult to trace the problems in the production process due to the overall similarity of the files, and the sentences in the key production files are shorter, the differences between the sentences will be more obvious, and it is easier to find the problems in the production process. Therefore, in the subsequent steps, the embodiment of the present invention analyzes the sentences in the key production files. First, according to the word vectors and production density of each phrase in each sentence of the key production files, the reference vector of each sentence in the key production files is obtained, and then the differences between the sentences can be analyzed based on the reference vector.

[0097] Preferably, in one embodiment of the present invention, the method for obtaining the reference vector of each sentence of the key production file specifically includes:

[0098] Any sentence in the key production document is taken as the target sentence, wherein the target sentence is a sentence randomly selected from all the key production documents. Since the greater the production closeness of a phrase, the stronger the correlation between the phrase and the production process of icodextrin peritoneal dialysis fluid, the production closeness of the phrase can be used as a weight, and the production closeness of each phrase in the target sentence is used to weight the word vector of each phrase in the target sentence to obtain the adjusted word vector of each phrase in the target sentence, and then the average value of the adjusted word vectors of all phrases in the target sentence is used as the reference vector of the target sentence.

[0099] As an example, in one embodiment of the present invention, the expression of the reference vector of the target sentence may be specifically, for example:

[0100]

[0101] in, A reference vector representing the target sentence; Indicates the first The production tightness of each phrase; Indicates the first The word vectors of the phrases; Indicates the number of phrases in the target sentence; Indicates the first Adjusted word vectors for each phrase.

[0102] The same method as above can be used to obtain the reference vector of each sentence in the key production file, and then the similarity parameter of each sentence in the key production file can be obtained by comparing the differences in the reference vectors of each sentence in the key production file. The larger the similarity parameter, the more similar each sentence in the key production file is to other sentences. Subsequently, based on the similarity parameter, similar sentences in the key production file can be divided into the same category, which facilitates the subsequent difference analysis of phrases in similar sentences and accurately traces the problems in the production process of icodextrin peritoneal dialysis fluid.

[0103] Preferably, in one embodiment of the present invention, the method for obtaining the similarity parameter of each sentence of the key production file specifically includes:

[0104] In the key production file, the average value of the cosine similarity of the reference vectors between the target sentence and all other sentences except the target sentence is used as the similarity parameter of the target sentence, wherein the cosine similarity is a technical means well known to those skilled in the art and will not be elaborated here.

[0105] As an example, in one embodiment of the present invention, the expression of the similarity parameter of the target sentence may be specifically, for example:

[0106]

[0107] in, Represents the similarity parameter of the target sentence; Indicates the target sentence and the cosine similarity of reference vectors between other sentences; Indicates the number of statements other than the target statement in the key production file.

[0108] The similarity parameter of each sentence in the key production file can be obtained by the same method as above, and then all the sentences in the key production file can be clustered based on the similarity parameter of each sentence to obtain multiple sentence clustering clusters, so that the sentences with similar similarity parameters in the key production file can be divided into the same clustering cluster, wherein the clustering algorithm can use the existing K-means algorithm or DBSCAN algorithm, which is not limited here. In the clustering process, the Euclidean distance or Manhattan distance between the similarity parameters of each sentence in the key production file can be used as the distance measure of each sentence.

[0109] The sentences in each sentence cluster are relatively similar and come from different key production documents. The greater the difference between the phrases in the sentence cluster, the greater the possibility that the relevant operation links of the production process represented by the phrases in the sentence cluster will have problems. Therefore, the differences in the word vectors between the phrases in the sentences of each sentence cluster can be analyzed. At the same time, combined with the production density of the phrases, the possibility of abnormalities in the relevant operation links represented by each phrase can be reflected by the obtained abnormal coefficient. Subsequently, based on the abnormal coefficient of the phrase in the key production documents, the abnormal problems in the production process of icodextrin peritoneal dialysis fluid can be traced back in a timely and accurate manner.

[0110] Preferably, in one embodiment of the present invention, the method for obtaining the abnormal coefficient of each phrase in the sentence of each sentence cluster specifically includes:

[0111] Any sentence clustering cluster is taken as the target sentence clustering cluster, any phrase in the sentence of the target sentence clustering cluster is taken as the phrase to be tested, and the accumulated value of the Euclidean distance between the word vectors of the phrase to be tested and all other phrases in the sentence of the target sentence clustering cluster except the phrase to be tested is taken as the initial abnormal parameter of the phrase to be tested. At the same time, for some phrases that are closely related to the production process of icodextrin peritoneal dialysis fluid, more attention should be paid to whether there are abnormal problems in the related operation links they represent. Therefore, the production density of the phrase to be tested can be used to weight the initial abnormal parameters and normalize them to obtain the abnormal coefficient of the phrase to be tested.

[0112] As an example, in one embodiment of the present invention, the expression of the abnormality coefficient of the phrase to be tested may be specifically, for example, as follows:

[0113]

[0114] in, The abnormality coefficient of the test phrase in the sentence representing the target sentence cluster; Indicates the production tightness of the phrase to be tested; The first sentence in the sentence other than the tested phrase in the cluster of the tested phrase and the target sentence The Euclidean distance between word vectors of other word groups; The number of phrases other than the phrase to be tested in the sentences representing the cluster of the target sentences; represents the initial abnormal parameters of the phrase to be tested; Represents the normalization function.

[0115] The same method as above can be used to obtain the abnormal coefficient of the phrase in each sentence in each sentence cluster, thereby obtaining the abnormal coefficient of each phrase in the key production file.

[0116] Step S4: Based on the abnormal coefficient of each phrase in the key production file, the production process of the department to be tested is traced.

[0117] The phrases in the key production documents represent the relevant operation links in the production process of icodextrin peritoneal dialysis fluid in the department to be tested. The larger the abnormal coefficient of the phrase in the key production document, the more problems there are in the related operations represented by the phrase. Therefore, based on the abnormal coefficient of each phrase in the key production document, the production process of the department to be tested can be traced, so that the problem can be traced back in a timely and accurate manner.

[0118] Preferably, in one embodiment of the present invention, the method for tracing the production process of the department to be tested specifically includes:

[0119] If the abnormal coefficient of each phrase in the key production file is greater than the preset abnormal threshold, an abnormal warning will be issued to remind the staff to promptly check the problems in the related operations represented by the phrases that have issued the abnormal warning. The preset abnormal threshold is set to 0.7. The specific value of the preset abnormal threshold can also be set by the implementer according to the specific implementation scenario, and is not limited here.

[0120] An embodiment of the present invention provides a production process traceability system for icodextrin peritoneal dialysis fluid, the system comprising a memory, a processor and a computer program, wherein the memory is used to store the corresponding computer program, the processor is used to run the corresponding computer program, and when the computer program runs in the processor, the method described in steps S1 to S4 can be implemented.

[0121] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0122] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A production process tracing method for icodextrin peritoneal dialysis solution, characterized in that: The method comprises: Obtain the production files of the department to be tested of icodextrin peritoneal dialysis fluid within a preset time period and the generation time of each production file; perform natural language processing on all production files to obtain multiple sentences and phrases of the production files and the word vector of each phrase; According to the difference of the word vectors between each word group and the number of occurrences of each word group in all production files, the production density of each word group is obtained; according to the difference of the word groups between each production file and the difference of the generation time, and the production density of the word groups in each production file, the importance of each production file is obtained; based on the importance, the key production files of the department to be tested are screened out from all production files; According to the word vectors and the production closeness of each phrase in each sentence of the key production file, a reference vector of each sentence of the key production file is obtained; according to the difference of the reference vectors of each sentence of the key production file, a similarity parameter of each sentence of the key production file is obtained; based on the similarity parameter, the sentences in the key production file are clustered to obtain a plurality of sentence clusters; according to the difference of the word vectors between each phrase in the sentence of each sentence cluster and the production closeness of each phrase, an abnormal coefficient of each phrase in the sentence of each sentence cluster is obtained; Based on the abnormal coefficient of each phrase in the key production documents, the production process of the department to be tested is traced.

2. The method for tracing the production process of an icodextrin peritoneal dialysis solution according to claim 1, characterized in that: The obtaining of the production density of each phrase comprises: Taking any phrase as a target phrase, negatively correlating the Euclidean distances of the word vectors between the target phrase and each other phrase except the target phrase, and obtaining the initial phrase similarity between the target phrase and each other phrase; Taking the accumulated value of the initial phrase similarities between the target phrase and all other phrases as the comprehensive phrase similarity of the target phrase; The number of occurrences of the target phrase in all production documents and the comprehensive phrase similarity are integrated and normalized to obtain the production density of the target phrase.

3. The method for tracing the production process of an icodextrin peritoneal dialysis solution according to claim 1, characterized in that: The importance of obtaining each production document includes: Taking the average of the production closeness of all phrases in each production file as the overall closeness of each production file; Based on the overall compactness, all production files are clustered to obtain a plurality of first file clusters, and according to the difference in the generation time of the production files in each first file cluster, a periodicity parameter of each first file cluster is obtained; The set of all phrases in each production file is taken as the phrase set of each production file, any production file is taken as the target production file, the Jaccard similarity coefficient of the phrase set between the target production file and each other production file except the target production file is taken as the initial content similarity between the target production file and each other production file; the cumulative value of the initial content similarities between the target production file and all other production files is taken as the comprehensive content similarity of the target production file; The overall closeness of the target production file, the comprehensive content similarity and the periodic parameter of the first file cluster where the target production file is located are integrated and normalized to obtain the importance of the target production file.

4. The method for tracing the production process of an icodextrin peritoneal dialysis solution according to claim 3, characterized in that: The step of obtaining the periodic parameters of each first file cluster includes: For any first file cluster, the production files in the first file cluster are sorted according to the generation time to obtain a production file sequence of the first file cluster; Taking two adjacent production files in the production file sequence as a production file group, and taking the absolute value of the difference between the generation times of the two production files in each production file group as the time interval of each production file group; After analyzing the discrete degrees of the time intervals of all production file groups and performing negative correlation mapping, the periodicity parameters of the first file clustering cluster are obtained.

5. The method for tracing the production process of an icodextrin peritoneal dialysis solution according to claim 1, characterized in that: Based on the importance, the key production documents of the department to be tested are screened from all production documents, including: Based on the importance, clustering all production files to obtain multiple second file clusters, and taking the average of the importance of all production files in each second file cluster as the cluster center of each second file cluster; The production files in the second file cluster corresponding to the maximum value of the cluster center are used as key production files of the department to be tested.

6. The method for tracing the production process of icodextrin peritoneal dialysis solution according to claim 1, characterized in that: The step of obtaining a reference vector for each sentence of a key production file includes: Taking any sentence in the key production file as the target sentence, using the production density of each phrase in the target sentence, weighting the word vector of each phrase in the target sentence, and obtaining an adjusted word vector of each phrase in the target sentence; The average value of the adjusted word vectors of all phrases in the target sentence is used as the reference vector of the target sentence.

7. The method for tracing the production process of icodextrin peritoneal dialysis solution according to claim 6, characterized in that: The similarity parameters of each sentence of the key production file are obtained including: In the key production file, the average value of the cosine similarities of the reference vectors between the target sentence and all other sentences except the target sentence is used as the similarity parameter of the target sentence.

8. The method for tracing the production process of icodextrin peritoneal dialysis solution according to claim 1, characterized in that: The abnormal coefficient of each phrase in each sentence cluster of each sentence is obtained, which includes: Taking any sentence clustering cluster as the target sentence clustering cluster, taking any phrase in the sentence of the target sentence clustering cluster as the phrase to be tested, and taking the accumulated value of the Euclidean distance between the word vectors of the phrase to be tested and all other phrases in the sentence of the target sentence clustering cluster except the phrase to be tested as the initial abnormal parameter of the phrase to be tested; The initial abnormality parameter is weighted and normalized by using the production density of the phrase to be tested to obtain the abnormality coefficient of the phrase to be tested.

9. The method for tracing the production process of an icodextrin peritoneal dialysis solution according to claim 1, characterized in that: The tracing of the production process of the department to be tested based on the abnormal coefficient of each phrase in the key production file includes: If the abnormal coefficient of each phrase in the key production file is greater than a preset abnormal threshold, an abnormal warning is issued.

10. A production process tracing system for icodextrin peritoneal dialysis fluid, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images