System and method for providing test time estimation suggestions based on product configuration information

Abstract

The present application discloses a system and method for providing test time estimation suggestion based on product configuration information, which pre-establishes a data database, a product configuration term database and a product embedding vector database, performs word analysis on new product configuration information to obtain analysis words and queries a corresponding one-bit effective code, sequentially introduces the one-bit effective code of the analysis words into a trained word2vec neural network model to calculate a plurality of word embedding vectors, sums all the word embedding vectors to calculate a new product embedding vector, compares the new product embedding vector with approximate product embedding vectors, and uses the model test time and series test time of the existing product corresponding to the approximate product embedding vectors as the estimation suggestion of the model test time and series test time of the new product, thereby achieving the technical effect of providing test time estimation suggestion based on product configuration information.

Description

TECHNICAL FIELD

[0001] The present application relates to a test time estimation suggestion system and method thereof, and in particular to a product configuration information based neural network calculation to provide test time estimation suggestion. BACKGROUND

[0002] Currently, in the production of manufacturing enterprises, the product testing link is an indispensable production link for checking and ensuring product quality. The prediction of test time for the testing link is not only important for production scheduling and equipment management, but also an important part of providing transparent production.

[0003] Because the product configuration information is not clearly specified, different manufacturing enterprises will have significantly different product configuration information for the same product (for example, the names used are different, the parameter display methods are different, the information quantity in the product configuration information is different, etc.). With the existing statistical method, the utilization of product configuration information is insufficient, so it is difficult to effectively use product configuration information to predict the test time of new products.

[0004] In summary, it can be seen that there has been a problem of inaccurate test time estimation for products in the prior art for a long time, so it is necessary to propose an improved technical means to solve this problem. SUMMARY

[0005] In view of the problem of inaccurate test time estimation for products in the prior art, the present application discloses a product configuration information based test time estimation suggestion system and method thereof, wherein:

[0006] The product configuration information based test time estimation suggestion system disclosed by the present application is suitable for an analysis and calculation device, which comprises a data database, a product configuration term database, a product embedding vector database, a receiving module, a neural network-like module, and a time estimation module.

[0007] The data database is used to store the existing product series, the existing product model, the model test time and the series test time of the existing product; the product configuration word database is used to store the product configuration words and one-hot code obtained by analyzing the product configuration information of the existing product; the product embedding vector database is used to store the product embedding vectors of the existing product, and the product embedding vectors are clustered and layered to generate a hierarchical clustering tree with at least three layers; the receiving module is used to receive the new product configuration information of the new product; the neural network module is used to analyze the words of the new product configuration information to analyze a plurality of analysis words, query the one-hot code corresponding to the analysis words from the product configuration word database, sequentially import the one-hot code of the analysis words into the trained word2vec neural network model to calculate a plurality of word embedding vectors, and add all the word embedding vectors to calculate the new product embedding vector of the new product configuration information; and the time estimation module is used to compare the new product embedding vector from the hierarchical clustering tree to the clustering layer, and then compare the product embedding vector most similar to the new product embedding vector from the clustering layer, and the model test time and the series test time of the existing product corresponding to the product embedding vector are used as the estimated suggestion of the model test time and the series test time of the new product.

[0008] The product configuration information-based test time estimation method disclosed by the present application is suitable for an analysis and calculation device, which comprises the following steps:

[0009] First, the analysis computing device pre-establishes a data database to store an existing product series, an existing product model, a model test time, and a series test time of an existing product; then, the analysis computing device pre-establishes a product configuration word database to store product configuration words and a one-hot encoding obtained by analyzing product configuration information of the existing product; then, the analysis computing device pre-establishes a product embedding vector database to store product embedding vectors of the existing product, and performs clustering and layering on the product embedding vectors to generate a hierarchical clustering tree with at least three layers; then, the analysis computing device receives new product configuration information of a new product; then, the analysis computing device analyzes the new product configuration information to obtain a plurality of analysis words; then, the analysis computing device queries a one-hot encoding corresponding to the analysis words from the product configuration word database; then, the analysis computing device sequentially inputs the one-hot encoding of the analysis words into a trained word2vec neural network model to calculate a plurality of word embedding vectors, and adds all the word embedding vectors to calculate a new product embedding vector of the new product configuration information; then, the analysis computing device compares the new product embedding vector from the hierarchical clustering tree to obtain a product embedding vector most similar to the new product embedding vector; finally, the analysis computing device takes the model test time and the series test time of the existing product corresponding to the product embedding vector as an estimation suggestion of the model test time and the series test time of the new product.

[0010] The system and method disclosed in the present application differ from the prior art in that a data database, a product configuration word database, and a product embedding vector database are pre-established, analysis words are obtained by performing word analysis on new product configuration information, and a one-hot encoding corresponding to the analysis words is queried, the one-hot encoding of the analysis words is sequentially input into a trained word2vec neural network model to calculate a plurality of word embedding vectors, all the word embedding vectors are added to calculate a new product embedding vector of the new product configuration information, the new product embedding vector is compared to obtain a product embedding vector most similar to the new product embedding vector, and the model test time and the series test time of the existing product corresponding to the product embedding vector are taken as an estimation suggestion of the model test time and the series test time of the new product.

[0011] Through the above technical means, the present application can achieve the technical effect of providing test time estimation suggestions based on product configuration information. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 A system block diagram of a system for providing test time estimation suggestions based on product configuration information is shown.

[0013] Figure 2A data database schematic diagram of the present application based on product configuration information to provide test time estimation suggestion is shown.

[0014] Figure 3 A product configuration term database schematic diagram of the present application based on product configuration information to provide test time estimation suggestion is shown.

[0015] Figure 4 A product embedding vector database schematic diagram of the present application based on product configuration information to provide test time estimation suggestion is shown.

[0016] Figure 5A And Figure 5B A method flow chart of the present application based on product configuration information to provide test time estimation suggestion method is shown.

[0017] Explanation of reference numerals:

[0018] 10 analysis computing device

[0019] 11 data database

[0020] 12 product configuration term database

[0021] 13 product embedding vector database

[0022] 14 receiving module

[0023] 15 neural network-like module

[0024] 16 time estimation module DETAILED DESCRIPTION

[0025] The embodiments of the present application will be described in detail below with reference to the drawings and examples, so that the realization process of how the present application applies technical means to solve technical problems and achieve technical effects can be fully understood and implemented.

[0026] First, the present application based on product configuration information to provide test time estimation suggestion system is disclosed, and please refer to Figure 1 shown, Figure 1 A system block diagram of the present application based on product configuration information to provide test time estimation suggestion system is shown.

[0027] The present application based on product configuration information to provide test time estimation suggestion system is disclosed, which is suitable for analysis computing device 10, which includes: data database 11, product configuration term database 12, product embedding vector database 13, receiving module 14, neural network-like module 15 and time estimation module 16, analysis computing device 10 is, for example: general computer, notebook computer, server, etc., here is only for example, and does not limit the application range of the present application.

[0028] The data database 11 is used to store the existing product series, the existing product model, the model test time and the series test time of the existing product. The existing product series, the existing product model, the model test time and the series test time of the existing product stored in the data database 11 are analyzed and calculated by the analysis and calculation device 10. The historical data of the existing product is excluded from the discrete data in the form of a box plot. The historical data of the existing product after the discrete data is excluded is statistically analyzed. The average value of the test time is calculated according to the existing product series and the existing product model respectively to serve as the model test time and the series test time. The schematic diagram of the data database 11 is shown in FIG. 2. Figure 2 Figure 2 The schematic diagram of the data database of the present application based on the product configuration information to provide the test time estimation suggestion is shown.

[0029] The product configuration word database 12 is used to store the product configuration words and the one-hot code obtained by analyzing the product configuration information of the existing product. Generally, according to different products, the product configuration information is composed of a plurality of words with inconsistent lengths. Specifically, a part of the product configuration information is, for example, "power supply 800W redundant 2sets mm 1u". The analysis and calculation device 10 analyzes the product configuration information "power supply 800W redundant 2sets mm 1u" as "power", "supply", "800W", "redundant", "2sets", "mm" and "1u". "Power", "supply", "800W", "redundant", "2sets", "mm" and "1u" are product configuration words. "Power", "supply", "800W", "redundant", "2sets", "mm" and "1u" are respectively set as unique and non-repetitive one-hot codes "45", "128", "1357", "9", "88", "2468" and "666". This is only an example and does not limit the application range of the present application. The schematic diagram of the product configuration word database 12 is shown in FIG. 3. Figure 3 Figure 3 The schematic diagram of the product configuration word database of the present application based on the product configuration information to provide the test time estimation suggestion is shown. In the product configuration word database 12, the existing product model is used as an example. The existing product series can also be changed. The existing product model and the existing product series can also be listed at the same time.

[0030] ​​The product embedding vector database 13 is used to store product embedding vectors of existing products and to cluster and hierarchize the product embedding vectors to generate a hierarchical clustering tree with at least three layers. The analysis computing device 10 analyzes the product configuration information by word analysis to analyze a plurality of sets of central words and at least two environmental words by the word2vec neural network module 15. The product configuration word database 12 is queried for one-bit effective codes corresponding to the central words and the at least two environmental words. The one-bit effective codes corresponding to the at least two environmental words are inputted and the one-bit effective codes corresponding to the central words are labeled as labels for multiple times to train the word2vec neural network model to obtain a plurality of training embedding vectors. The product embedding vector of the product configuration information is calculated by summing all the training embedding vectors. The training method and process of the word2vec neural network model are described in the prior art, and will not be described herein. The product embedding vector database 13 is shown in FIG. 7. Figure 4 Figure 4 The product embedding vector database 13 is used to store product embedding vectors of existing products and to cluster and hierarchize the product embedding vectors to generate a hierarchical clustering tree with at least three layers. The analysis computing device 10 analyzes the product configuration information by word analysis to analyze a plurality of sets of central words and at least two environmental words by the word2vec neural network module 15. The product configuration word database 12 is queried for one-bit effective codes corresponding to the central words and the at least two environmental words. The one-bit effective codes corresponding to the at least two environmental words are inputted and the one-bit effective codes corresponding to the central words are labeled as labels for multiple times to train the word2vec neural network model to obtain a plurality of training embedding vectors. The product embedding vector of the product configuration information is calculated by summing all the training embedding vectors. The training method and process of the word2vec neural network model are described in the prior art, and will not be described herein. The product embedding vector database 13 is shown in FIG. 7.

[0031] It is worth noting that the product configuration information is analyzed by word analysis to analyze a plurality of sets of central words and at least two environmental words. The set number of words is sequentially selected. The middle of the selected word number is the central word, and the remaining words are the at least two environmental words. The set number is an odd number.

[0032] Specifically, if the product configuration information is "power supply 800W redundant 2sets mm 1u" and the set number is 5, the sequentially selected words are "power supply 800W redundant 2sets", the central word is "800W", and the environmental words are "power", "supply", "redundant", and "2sets", respectively. The selected words are "supply 800W redundant 2sets mm", the central word is "redundant", and the environmental words are "supply", "800W", "2sets", and "mm", respectively. This is only an example and does not limit the application range of the present application.

[0033] ​When the receiving module 14 receives the new product configuration information of the new product, the neural network module 15 analyzes the new product configuration information to analyze a plurality of analysis words, queries a one-hot encoding corresponding to the analysis words from the product configuration word database 12, sequentially inputs the one-hot encoding of the analysis words into the trained word2vec neural network model to calculate a plurality of word embedding vectors, and adds all the word embedding vectors to calculate a new product embedding vector of the new product configuration information. The calculation content of the word2vec neural network model is described in the prior art, and the present application will not be described in detail.

[0034] Notably, if the new product configuration information received by the receiving module 14 has words that are different from the product configuration words in the product configuration word database 12, the analysis computing device 10 will establish product configuration words that are different from the product configuration words and set a unique and non-repeating one-hot encoding to update the product configuration word database 12.

[0035] The time estimation module 16 compares the new product embedding vector from the hierarchical clustering tree to a clustering hierarchy, and then compares the product embedding vector most similar to the new product embedding vector from the clustering hierarchy to the model test time and series test time of the existing product corresponding to the product embedding vector to estimate the model test time and series test time of the new product.

[0036] Further, the time estimation module 16 compares the new product embedding vector from the hierarchical clustering tree to a clustering hierarchy, and then selects the product embedding vector from the clustering hierarchy according to a data selection threshold to estimate the model test time and series test time of the new product. The data selection threshold can be pre-set in the analysis computing device 10 or input by external input, which is only an example and does not limit the application range of the present application.

[0037] Specifically, assuming that the number of product embedding vectors in the clustering hierarchy is 10 and the data selection threshold is set to 5, the time estimation module 16 will select the 5 product embedding vectors most similar to the new product embedding vector from the product embedding vectors in the clustering hierarchy. The model test time and series test time of the existing product corresponding to the selected product embedding vector are used as the estimated model test time and series test time of the new product.

[0038] It is worth noting that the analysis computing device 10 will update the product configuration word database 12 regularly, and when the product configuration word database 12 is updated, the analysis computing device 10 can retrain the word2vec neural network model through the neural network module 15 to calculate the product embedding vectors of the product configuration information, and cluster and layer the product embedding vectors to update the product embedding vector database 13. Through the regular updating of the product configuration word database 12 and the product embedding vector database 13, the estimation suggestion of the test time provided by the time estimation module 16 can ensure timeliness and accuracy.

[0039] The present application proposes the configuration semantics of product configuration information in big data analysis, uses a neural network-based deep learning model to learn the configuration semantics, converts the product configuration information into unique embedding vectors through the embedding layer of the model, and enables the product configuration information of each product to be approximately compared and calculated through the embedding vectors. Further, through the hierarchical clustering tree established by clustering and layering the embedding vectors, the approximate comparison speed of the product configuration information can be effectively improved, and the accuracy of the model test time and series test time estimation suggestion of the new product can be improved.

[0040] Next, the operation system and method of the first embodiment of the present application will be described below with the first embodiment, and please refer to Figure 5A and Figure 5B shown, Figure 5A and Figure 5B show the method flowchart of the test time estimation method based on product configuration information.

[0041] Firstly, the analysis computing device pre-establishes a data database to store an existing product series, an existing product model, a model test time and a series test time of an existing product (step 101); then, the analysis computing device pre-establishes a product configuration word database to store product configuration words obtained by analyzing product configuration information of the existing product and a one-hot encoding (step 102); then, the analysis computing device pre-establishes a product embedding vector database to store product embedding vectors of the existing product, and performs hierarchical clustering on the product embedding vectors to generate a hierarchical clustering tree with at least three layers (step 103); then, the analysis computing device receives new product configuration information of a new product (step 104); then, the analysis computing device performs word analysis on the new product configuration information to analyze a plurality of analysis words (step 105); then, the analysis computing device queries a one-hot encoding corresponding to the analysis words from the product configuration word database (step 106); then, the analysis computing device sequentially imports the one-hot encoding of the analysis words into a trained word2vec type neural network model to calculate a plurality of word embedding vectors, and adds all the word embedding vectors to calculate a new product embedding vector of the new product configuration information (step 107); then, the analysis computing device compares the new product embedding vector from the hierarchical clustering tree to obtain a clustering layer (step 108); then, the analysis computing device compares the new product embedding vector from the clustering layer to obtain a product embedding vector most similar to the new product embedding vector (step 109); finally, the analysis computing device takes the model test time and the series test time of the existing product corresponding to the product embedding vector as an estimation suggestion of the model test time and the series test time of the new product (step 110).

[0042] In summary, the difference between the present application and the prior art is that a data database, a product configuration word database and a product embedding vector database are pre-established, analysis words are obtained by performing word analysis on new product configuration information, a one-hot encoding corresponding to the analysis words is queried, the one-hot encoding of the analysis words is sequentially imported into a trained word2vec type neural network model to calculate a plurality of word embedding vectors, all the word embedding vectors are added to calculate a new product embedding vector of the new product configuration information, the new product embedding vector is compared to obtain a similar product embedding vector, and the model test time and the series test time of an existing product corresponding to the similar product embedding vector are taken as an estimation suggestion of the model test time and the series test time of the new product.

[0043] This technical means can solve the problem of inaccurate product test time estimation in the prior art, and achieve the technical effect of providing test time estimation suggestions based on product configuration information.

[0044] Although the disclosed embodiments of the present application are as described above, the above description is not intended to limit the scope of patent protection of the present application directly. Any person skilled in the art can make some changes in the form and details of the implementation without departing from the spirit and scope of the present application disclosed. The scope of patent protection of the present application shall be subject to the scope defined by the appended claims.

Claims

1. A system for providing test time estimation suggestion based on product configuration information, adapted for an analysis computing device, comprising: a data database for storing existing product series, existing product models, model test time and series test time of existing products; a product configuration term database for storing product configuration terms and corresponding valid codes obtained by analyzing product configuration information of the existing products; a product embedding vector database for storing product embedding vectors of the existing products, and clustering and layering the product embedding vectors to generate a hierarchical clustering tree with at least three layers; a receiving module for receiving new product configuration information of a new product; a neural network-like module for analyzing the new product configuration information to obtain a plurality of analysis terms, querying the product configuration term database to obtain the valid codes corresponding to the analysis terms, sequentially inputting the valid codes of the analysis terms into a trained word2vec neural network-like model to calculate a plurality of term embedding vectors, and summing all the term embedding vectors to calculate a new product embedding vector of the new product configuration information; and a test time estimation module for comparing the new product embedding vector with the hierarchical clustering tree to obtain a clustering layer, and comparing the new product embedding vector with the product embedding vectors in the clustering layer to obtain a product embedding vector most similar to the new product embedding vector, and using the model test time and the series test time of the existing product corresponding to the product embedding vector as the model test time and the series test time estimation suggestion of the new product. 2.The system for providing test time estimation suggestion based on product configuration information according to claim 1, wherein the analysis computing device is to statistically analyze historical data of the existing products excluding discrete data, and calculate average values of test time according to the existing product series and the existing product models respectively to obtain the model test time and the series test time. 3.The system for providing test time estimation suggestion based on product configuration information according to claim 1, wherein the analysis computing device is to analyze the product configuration information of the existing products by term analysis, and assign unique and non-repeated valid codes to each product configuration term to pre-establish the product configuration term database. ​ ​ ​ ​ ​ ​ 4. The system for providing test time estimation suggestions based on product configuration information according to claim 1, wherein the product embedding vector of the existing product is obtained by the analysis and calculation device performing word analysis on the product configuration information through the neural network module to analyze multiple sets of central words and at least two environmental words, querying the effective codes corresponding to the central words and the at least two environmental words from the product configuration word database, performing multiple training embedding vectors by taking the effective codes corresponding to the at least two environmental words as input and taking the effective codes corresponding to the central words as labels for training a word2vec neural network model, and summing all the training embedding vectors to calculate the product embedding vector of the product configuration information.

5. The system for providing test time estimation suggestions based on product configuration information according to claim 4, wherein the neural network module analyzes the product configuration information to analyze multiple sets of central words and at least two environmental words, sequentially selects a set number of words, the middle of the selected number of words is the central word, and the remaining words are the at least two environmental words, wherein the set number is an odd number.

6. A method for providing test time estimation suggestions based on product configuration information, applicable to analytical computing devices, comprising the following steps: The analysis and computing device pre-establishes a data database to store existing product series, existing product models, model test times, and series test times of existing products; The analysis and computing device pre-establishes a product configuration term database to store product configuration terms and valid codes obtained from analyzing the product configuration information of the existing products. The analysis and computing device pre-establishes a product embedding vector database to store the product embedding vectors of the existing products, and clusters the product embedding vectors to generate a hierarchical clustering tree of at least three levels. The analysis and computing device receives the new product configuration information of the new product; The analysis and computing device performs word analysis on the new product configuration information to generate multiple analytical words; The analysis and calculation device retrieves the valid code corresponding to the analyzed word from the product configuration term database; The analysis and calculation device sequentially imports the effective encoding of the analyzed words into a trained word2vec neural network model to calculate multiple word embedding vectors, and sums all the word embedding vectors to calculate the new product embedding vector of the new product configuration information. The analysis and computing device compares the new product embedding vector from the hierarchical clustering tree to determine the clustering hierarchy; The analysis and calculation device compares the product embedding vector from the cluster hierarchy to find the one that most closely approximates the new product embedding vector; and The analysis and calculation device uses the model test time and series test time of the existing product corresponding to the product embedding vector as a prediction suggestion for the model test time and series test time of the new product.

7. The method of claim 6, wherein the step of the analysis computing device pre-establishing the data database to store the existing product series, the existing product model, the model test time and the series test time of the existing product is to statistically analyze the historical data of the existing product excluding discrete data to calculate the average value of test time according to the existing product series and the existing product model respectively as the model test time and the series test time.

8. The method of claim 6, wherein the step of the analysis computing device pre-establishing the product configuration term database to store the product configuration terms and the valid codes obtained by analyzing the product configuration information of the existing product is to analyze the product configuration information of the existing product by term analysis, and then set each product configuration term with a unique and non-repeated valid code to pre-establish the product configuration term database.

9. The method of claim 6, wherein the step of the analysis computing device pre-establishing the product embedding vector database to store the product embedding vector of the existing product is to analyze the product configuration information by term analysis to analyze a plurality of central terms and at least two context terms, query the valid codes corresponding to the central terms and the at least two context terms from the product configuration term database, and perform multiple times of inputting the valid codes corresponding to the at least two context terms and setting the valid codes corresponding to the central terms as labels to train a word2vec type neural network model to obtain a plurality of training embedding vectors, and then sum all the training embedding vectors to calculate the product embedding vector of the product configuration information.

10. The method of claim 9, wherein the step of the analysis computing device analyzing the product configuration information by term analysis to analyze a plurality of central terms and at least two context terms is to sequentially select a set number of terms, the middle of the selected term number is the central term, and the remaining terms are the at least two context terms, wherein the set number is an odd number.

Images