TWI604322B - Solution search system and solution search system operation method - Google Patents

Description

Solution search system and solution search system operation method

The present invention relates to a solution search system, and more particularly to a solution search system that utilizes massive data and data exploration techniques.

In addition to being closely related to R&D technology, the success of a product requires a large number of tests to ensure product stability, especially for high-stability, high-reliability technology products such as industrial instruments, mobile devices, workstations, and individuals. For products such as computers or servers, the standards for quality control testing are more stringent. When a product is detected, it must be done by copying the problem, collecting and analyzing the relevant information, identifying the cause of the problem, suggesting possible solutions, and verifying the proposed solution... to ensure that the detected problem is Properly resolved, these processes may not only be very time consuming, but may even lead to missed opportunities for the product to enter the market, and implementation must rely on the personal experience and professionalism of the engineer; that is, whether the engineer's experience and professionalism are sufficient will be greatly It affects the time required to propose a solution, and it may also affect whether the proposed solution can completely solve the problem, resulting in the quality of the solution is difficult to grasp. In addition, because personal experience is not easy to pass on, even if the problem to be solved is the same or similar, different engineers may have to repeat the above process to get a solution. This practice is not only inefficient, but also ensures that the engineer can find the most appropriate. s solution.

In addition, for the same type of products, the proportion of the same or similar problems is very high. Although there have been cases of recording or archiving solutions in the past, due to the wide variety of problems, the amount of information involved is quite large, plus Engineers may not be able to describe the problem in a way that is inconsistent, so it is difficult to systematically store it. As a result, engineers are still not easy to find relevant solutions, and it is difficult to achieve the purpose of sharing experience with engineers.

An embodiment of the present invention provides a solution search system including a computing server, a huge database, a database server, a modeling server, and a hub server.

When receiving a plurality of resolved problem description files, the computing server compares the common standard word comparison table with the text of the problem generative field in each of the resolved problem description files to generate the main problem file of the resolved problem description file. The classification code is based on the sub-standard word comparison table corresponding to the main classification code and the text in the resolved problem description file to generate a secondary classification code of the solved problem description file, and at least according to the solved problem, the main classification of the file is described. The code and the secondary classification code generate a solution code for the resolved problem description file, and a model input file for generating a resolved problem description file based on the resolved problem description file and data exploration algorithm.

The database server stores the solution in the resolved problem description file to a huge database based on the resolved solution description file.

The modeling server builds a prediction model based on the data exploration algorithm and the model input file and solution code of the problem description file.

When receiving a plurality of resolved problem description files, the hub server transmits a plurality of resolved problem description files to the computing server. The hub server transmits the model input file and solution code corresponding to the solved problem description file generated by the computing server to the modeling server.

Another embodiment of the present invention provides a method for operating a solution search system including a computing server, a huge database, a database server, and a hub server. The operation method of the solution search system includes when the hub server receives a plurality of solved problem description files, the hub server transmits a plurality of solved problem description files to the operation server, and the operation server compares the common standard words. The table is compared with the text of the problem causal field in each of the solved problem description files to generate the main classification code of the solved problem description file, and the calculation server according to the sub-standard word comparison table corresponding to the main classification code and the solved The text in the problem description file is compared with the secondary classification code for generating the resolved problem description file, and the computing server generates the resolved problem description file by at least the main classification code and the secondary classification code of the file according to the solved problem. The solution code, the computing server according to the solved problem description file and the data exploration algorithm generates a model input file of the solved problem description file, and the database server describes the solved problem of the file solution code according to the solved problem. The solution in the description file is saved to The volume database, the hub server transmits the model input file and solution code corresponding to the solved problem description file generated by the computing server to the modeling server, and the modeling server is based on the data exploration algorithm and has been solved. The problem description file model input file and solution code establish a prediction model. The invention is related to a high-efficiency solution search system, which can effectively solve the defects of the core of the current Internet of Things architecture that cannot accurately hit the problem.

1 is a schematic diagram of a solution search system 100 in accordance with an embodiment of the present invention. The solution search system 100 includes a computing server 110, a model server 120, a huge database 130, a database server 140, a hub server 150, an associated database 160, and a modeling server 170.

In some embodiments of the present invention, the user may organize the problem encountered when operating a product system into a problem description file, and upload the problem description file to the solution search system 100, in the solution search system 100. The model server 120 can use its internal predictive model to analyze the problem description file and identify possible solutions to the problems encountered by the product.

The problem description file can use text form to record information related to product system issues, which can include system problem descriptions, observed phenomena and results, subsystems related to product system problems, and problems. That is, it can explain how the problem can be copied, but is not limited to the above information.

However, before the user utilizes the solution search system 100 to search for a solution, the solution search system 100 also allows the user to build a predictive model based on prior experience with previous product issues. In some embodiments of the present invention, when the user U1 wants to establish a prediction model through the solution search system 100, the user U1 must first provide a plurality of solved first problem description files A1, each of which has been solved. A problem description file A1 contains the relevant fields for recording and system problems, including the description of system problems and phenomena, the subsystem to which the system problem belongs, the process of the problem, etc. It also contains information related to problem solving, such as the cause description field of the system problem, the subsystem field where the problem occurs, the solution...etc., etc., in addition to the suggested method for solving the problem, the solution also includes the corresponding solution. Program code. In this way, the solution search system 100 can use the corresponding data exploration algorithm to find the relationship between each solved first problem description file A1 and its solution, and establish a prediction model, and the data exploration used here. The algorithm is not specifically limited, and any algorithm that can be used to calculate the degree of association can be used for this calculation. In some embodiments of the present invention, the solution search system 100 classifies each of the resolved first problem description files A1 through the solution code of each of the resolved first problem description files A1.

Although the first solution D1 of the first problem description file A1 may already contain the solution code set by the user U1, the computing server 110 may further describe the file A1 and solve the problem according to the first problem. The information in the relevant fields, such as the description field and the subsystem field in question, adjust the first solution code C1 included in the first solution D1 of the first problem description file A1. For example, the solution code can be generated according to the content of the cause description field to generate a solution code, so that all solutions with the same cause description can have the same solution code.

In some embodiments of the present invention, if the solution code of the problem description file is preset, the solution code includes a plurality of subcodes, such as bios.mrc, where bios indicates the resolved problem description file and the basic input/output system (basic) Input/output system) is related, and mrc indicates that the problem description file is related to the memory reference code in the basic input/output system, and the operation server 110 can solve the information according to other fields. The solution code bios.mrc in the problem description file is extended to bios.mrc.i2c, indicating the resolved problem description file system and the internal reference circuit (I2C) of the memory reference code in the basic input/output system. Related.

When the number of subcodes included in the solution code is greater, it means that the solution search system 100 classifies the problem more finely, so the number of solved problem description files with the same solution code will be less, and each solution The solution to which the code corresponds is usually less. Conversely, when the number of subcodes included in the solution code is smaller, it means that the problem is classified more coarsely, so each solution code usually has more solutions. However, whether the solution code corresponds to too many or too few solutions may reduce the prediction accuracy of the solution search system 100. For example, if there are too many solutions for the solution code, then the different types of problem description files are likely to still be set with the same solution code, and the user must be among many possible solutions. Trying one by one makes it possible to find a suitable solution. However, if the solution code corresponds to too few solutions, even if the problem description files of similar types may be set with different solution codes, even if there is already a solution to the problem in the solution search system 100 In the case of the program, it may not be available to the user for reference.

In order to maintain the prediction accuracy of the solution search system 100, in the embodiment of FIG. 1, when the computing server 110 receives the plurality of solved first problem description files A1, the computing server 110 may first adopt the general standard. The word comparison table is compared with the text of the problem generative field in the first problem description file A1 of each of the solved first problem descriptions to generate the primary classification code MC of the first problem description file A1 that has been solved. The common standard word comparison table may contain system-related vocabulary such as "firmware", "basic input/output system (bios)", "substrate management controller (bmc)", "power", ... Etc., the computing server 110 may first compare the first problem that has been solved to describe the vocabulary in the problem generative field of the file A1 that appears to be the same or similar to the common standard word comparison table, and the same or similar vocabulary The main classification code MC is generated in a standard representation.

For example, if the problem description field in the first problem description file A1 states: "The scratchpad setting of the baseboard controller in the basic input/output system is incorrect....", since it appears in this description The "basic input/output system" and "substrate controller" are the same or similar to the "basic input/output system (bios)" and "substrate management controller (bmc)" described in the common standard word comparison table. The calculation server 110 can set the main classification code MC to "bios.bmc". In addition, in order to avoid the first problem description, the order of the explanatory text in the problem-causing field in the file A1 is different, and the substantially identical content is obtained, for example, "bios.bmc" and "bmc.bios". In some embodiments of the present invention, the common standard word comparison table may further define the priority weight of each standard word, so the computing server 110 generates the main classification code MC in a specific order.

After the main classification code MC is generated, the operation server 110 further compares the text in the file A1 with the first problem description file corresponding to the first sub-standard word comparison table corresponding to the main classification code MC to generate the first problem solved. Describe the first classification code SC1 of the file A1. For example, if the primary classification code MC of the first problem description file A1 contains the code bmc, indicating that the first problem description file A1 may be related to the baseboard management controller, the computing server 110 may be related to the baseboard management controller. A sub-standard word comparison table continues to be compared with the text in the first problem description file A1. In this case, the first sub-standard word comparison table utilized by the computing server 110 will contain the words associated with the baseboard management controller. word. Then, the computing server 110 can generate the first solution code C1 of the solved first problem description file A1 according to the first classification code MC of the file A1 and the first secondary classification code SC1 according to the first problem that has been solved. In this way, each of the first problem description files A1 can be classified efficiently and accurately.

In order to avoid a situation in which the prediction accuracy of the solution search system 100 is reduced because the solution code corresponds to too many solutions, in some embodiments of the present invention, when the solution corresponds to the first solution code C1 When the number of the solutions is greater than the upper limit, for example, greater than 30 solutions, the computing server 110 may further perform a second sub-standard word comparison table corresponding to the first secondary classification code SC1 in the first solution code C1. The first problem solved describes the text collation in the file A1 to generate the second sub-classification code SC2 of the first problem description file A1 that has been resolved. That is, the computing server 110 can utilize the similarity principle described above to generate the first sub-category code SC1, that is, the computing server 110 can utilize the second sub-standard word comparison table associated with the first sub-category code SC1. The first problem describes the file A1 to be more precise and detailed classification, and then according to the first problem solved, the main classification code MC of the file A1, the first classification code SC1 and the second classification code SC2 are updated. The first problem describes the first solution code C1 of the file A1. In this way, it is possible to classify each of the first problem description files A1 in a more detailed manner, and also to avoid a decrease in prediction accuracy. In some embodiments of the present invention, as long as the sub-standard word comparison table in the solution search system 100 has sufficient correspondence, the computing server 110 may be when the number of solutions corresponding to the solution code is greater than the upper limit. The classification is continued according to the above method until the number of solutions corresponding to the solution code is not greater than the upper limit.

In contrast, in order to avoid a decrease in the prediction accuracy of the solution search system 100 because the solution code corresponding to the solution code is too small, in some embodiments of the present invention, when corresponding to the first solution code C1 The number of solutions is less than the lower limit, for example less than 10 solutions, and the number of solutions corresponding to one of the first solution codes C1 recently solved the classification code and the number of solutions corresponding to the first solution code C1 When the sum is not greater than the upper limit, for example, no more than 30 solutions, the computing server 110 may update the first solution code C1 to the nearest solution classification code, or the first solution code C1 and Its recently resolved classification code is updated together to the same new resolution classification code.

For example, if the first solution code C1 is "bios.bmc.i2c.mrc.spd", the solution code closest to the first solution code C1 should have the closest classification to the first solution code C1. For example, it may be "bios.bmc.i2c.mrc", "bios.bmc.i2c.spd", "bios.bmc.mrc.spd", "bios.bmc.i2c", etc., so the computing server 110 It is possible to check the number of solutions corresponding to the solution codes similar to the first solution code C1 mentioned above, and the number of solutions corresponding to the solution code closest to the first solution code C1 and the first solution The sum of the number of solutions corresponding to the code C1 is not greater than the upper limit value, and the two can be combined at the same time, and the solution code of one of them can be considered when merging. For example, the code C1 is bios.mrc.i2c.spd and its most recent solution code is bios.i2c, and the number of solutions corresponding to the two is 10, because it is less than the upper limit of 30, so The solution corresponding to the bios.mrc.i2c.spd code is added to the code bios.i2c. In addition, a new solution code can be created, for example, a new solution code for both of them, for example, if code C1 is firmware.bios.mrc and its most recent solution code is firmware. Bios.i2c, the number of solutions corresponding to the two is 20, because it is less than the upper limit of 30, so the solution corresponding to the two codes can be merged, so the two codes have the same keyword as firmware. With bios, the new code is firmware.bios).

Through the above-mentioned solution search system 100, each solution code can be checked to avoid the situation that the solution code corresponds to too many or too few solutions, thereby achieving the effect of ensuring the prediction accuracy.

After generating the first solution code C1 of each solved first problem description file A1, the computing server also generates a data exploration algorithm according to the first problem described in the description file A1 and the prediction model to be established. The first problem solved describes the first model input file B1 of the file A1.

In an embodiment of the present invention, the computing server 110 may describe a text description attribute file of the file A1 according to the first problem. The attribute description file may be composed of a plurality of keywords, each of which is composed of a pair of attribute name and attribute value, and is one of the present inventions. In the embodiment, it can be described in the text format of json. When the first problem description file A1 uses non-fixed format text bar information related to system problems, the computing server 110 may also use a regular expression to identify the keyword name and the value of the obtained keyword. Furthermore, the computing server 110 can use the standard word comparison table to compare the text of the first problem description file A1 to generate a keyword description file, and the method is the same as the foregoing according to the common standard word comparison table to generate the first problem description file A1. The classification code MC is similar. However, the first problem description file A1 of the first model input file B1 is used to test and train the prediction model, so when establishing the keyword description file, the computing server 110 will describe the file A1 against the first problem and solve the problem indirectly. Relevant information, such as fields other than the field that resolves the cause of the problem or the solution.

Figure 5 is a partial view of a standard word comparison table according to an embodiment of the present invention. Syntactic vocabulary and vocabulary can be standardized through the standard word comparison table, so that the meaning of the keyword description file can be expressed more efficiently and correctly. In addition, in order to avoid the confusion of the semantics of the keyword description file, the values of all keywords can be represented in lowercase.

After completing the keyword description file, the computing server 110 may preferentially select keywords with higher weights or user preset preferences from the keyword description file as predictors to generate predictor files, and then according to the predictive factor file. And the predictive model generates a first model input file B1. For example, the computing server 110 can adjust the predictor file according to the characteristics of the predictive model. For example, in the CBayes model, regardless of the dependency of the numbers, the numbers in the predictor file can be Partially deleted to generate the first model input file B1. However, different prediction models have different requirements for the format of the input file, and the present invention is not limited to the above embodiment.

After the computing server 110 generates the first model input file B1 and the first solution code C1 corresponding to each of the solved first problem description files A1, the hub server 150 describes each of the solved first problem description files. The first model input file B1 corresponding to A1 is transmitted to the modeling server 170, and the modeling server 170 can describe the first model input of the file A1 according to each solved first problem. The file B1 and the first solution code C1, and the data exploration algorithm corresponding to the modeling server 170, such as Bayes, CBayes or SGD, are used to build the prediction model. After the modeling server 170 completes the establishment of the prediction model, the solution search system 100 can utilize the prediction model established by the modeling server 170 to search for possible solutions to the system problem.

In addition, the hub server 150 transmits the first solution code C1 corresponding to each of the solved first problem description files A1 to the database server 140, and the database server 140 is based on the first solution solved. The first solution code C1 of the problem description file A1 stores the first solution D1 in the first problem description file A1 that has been resolved to the huge database 130. That is, the database server 140 stores the first solution D1 of the first problem description file A1 having the same first solution code C1 in the same field in the huge database 130. Therefore, if another user uploads a problem description file with the same solution code, the solution search system 100 can output each first solution D1 in the field for other users to select.

For example, FIG. 2 is a schematic diagram of another usage scenario of the solution search system 100 of the present invention. In FIG. 2, when the user U2 uploads the unresolved second problem description file A2 to the solution search system 100, the hub server 150 transmits the unresolved second problem description file A2 to the computing server 110. . When the computing server 110 receives the unresolved second problem description file A2, the computing server 110 generates the second problem description file A2 according to the second problem description file A2 and the data exploration algorithm used by the model server 120. The second model enters the file B2. Then, the hub server 150 can transmit the second model input file B2 of the second problem description file A2 to the model server 120, and the model server 120 can analyze the second model by using the prediction model established by the previous modeling server 170. The input file B2 is closer to which of the first problem description files in the plurality of solved first problem description files that have been previously received, and further generates the second solution code C2 of the second problem description file A2. After the hub server 150 transmits the second solution code C2 to the database server 140, the database server 140 can read the second solution D2 from the huge database 130 according to the second solution code C2, and finally the hub The server 150 then outputs the second solution D2 read by the database server 140 from the huge database 130 for reference by the user U2.

In an embodiment of the present invention, the database server 140 and the huge database 130 can be a database server and a huge database supporting systems such as Hadoop Distribute File System (HDFS), Hadoop Map/Reduce, and Hive. Or, it can support other database systems suitable for processing huge amounts of data to meet the needs of the solution search system 100 for quickly processing and storing large amounts of data. In addition, the association database 160 can be, for example, an associated database such as MySql, PostgreSql, etc., which can store small and/or temporary data required by the hub server 150 for calculation.

Through the above-described solution searching system 100 of the embodiment of the present invention, engineers can share their experiences in solving system problems in the past, and can easily find possible solutions to reduce the time for solving product problems and improve the quality of the solution. . In addition, by automatically establishing solution code and continuously classifying or merging the solution code, the solution search system 100 can further avoid solution search due to too many or too few solutions corresponding to the unsolved code. The problem of reduced system accuracy.

FIG. 3 is a flow diagram of an operation method 300 of the solution search system 100 in accordance with an embodiment of the present invention. The operation method 300 of the solution search system includes steps S310 to S380:

S310: When the hub server 150 receives the plurality of solved first problem description files A1, the hub server 150 transmits a plurality of solved first problem description files A1 to the computing server 110;

S320: The computing server 110 compares the common standard word comparison table with the text of the problem causal field in each of the solved first problem description files A1 to generate the main classification code MC of the solved first problem description file A1;

S330: The computing server 110 compares the first sub-standard word comparison table corresponding to the main classification code MC with the text of the first problem description file A1 that has been solved to generate the first problem description file A1 of the first problem. To classify the code SC;

S340: The computing server 110 generates the first solution code C1 of the first problem description file A1 that has been solved according to at least the primary classification code MC of the file A1 and the first secondary classification code SC.

S350. The computing server 110 generates the first model input file B1 of the first problem description file A1 according to the first problem description file A1 and the data exploration algorithm.

S360: the database server 140 stores the first solution D1 in the first problem description file A1 that has been resolved to the huge database 130 according to the first solution code C1 of the first problem description file A1 that has been solved;

S370: The hub server 150 transmits the first model input file B1 and the first solution code C1 corresponding to the solved first problem description file A1 generated by the computing server 110 to the modeling server 170;

S380: The modeling server 170 establishes a prediction model according to the data exploration algorithm and the first model input file B1 and the first solution code C1 of the first problem description file A1 that have been solved.

In addition, in order to avoid a situation in which the prediction accuracy of the solution search system 100 is reduced due to too many solutions corresponding to the solution code, in some embodiments of the present invention, the method 300 may also correspond to the first When the number of solutions of the solution code C1 is greater than the upper limit, for example, greater than 30 solutions, the operation server 110 may be caused to be the second corresponding to the first secondary classification code SC1 in the first solution code C1. The sub-standard word comparison table is compared with the text in the first problem description file A1 that has been resolved to generate the second sub-classification code SC2 of the first problem description file A1 that has been resolved. That is, the computing server 110 can utilize the similarity principle described above to generate the first sub-category code SC1, that is, the computing server 110 can utilize the second sub-standard word comparison table associated with the first sub-category code SC1. The first problem describes the file A1 to be more precise and detailed classification, and then according to the first problem solved, the main classification code MC of the file A1, the first classification code SC1 and the second classification code SC2 are updated. The first problem describes the first solution code C1 of the file A1. In this way, it is possible to classify each of the first problem description files A1 in a more detailed manner, and also to avoid a decrease in prediction accuracy. In some embodiments of the present invention, as long as the sub-standard word comparison table in the solution search system 100 has sufficient correspondence, the computing server 110 may be when the number of solutions corresponding to the solution code is greater than the upper limit. The classification is continued according to the above method until the number of solutions corresponding to the solution code is not greater than the upper limit.

In contrast, in some embodiments of the present invention, the method 300 may also correspond to the first solution, in order to avoid that the solution accuracy of the solution search system 100 is reduced because the solution code corresponding to the solution code is too small. The number of solutions of the solution code C1 is less than the lower limit value, for example, less than 10 solutions, and the number of solutions corresponding to one of the first solution codes C1 that recently solved the classification code corresponds to the first solution code C1. When the sum of the number of solutions is not greater than the upper limit, for example, no more than 30 solutions, the computing server 110 may update the first solution code C1 to the nearest solution classification code, or the first solution Code C1 and its most recently resolved classification code are updated together to the same new resolution classification code.

FIG. 4 illustrates a flow chart of an operation method 400 of the solution search system 100 in accordance with another embodiment of the present invention. The method 400 of operation of the solution search system 100 includes steps S410 through S470:

S410: When the hub server 150 receives the unresolved second problem description file A2, the unresolved second problem description file A2 is transmitted to the computing server 110;

S420: When the computing server 110 receives the unresolved second problem description file A2, according to the unsolved second problem description file A2 and the data exploration algorithm generate an unresolved second problem description file A2 second model input File B2;

S430: The hub server 150 transmits the second model input file B2 generated by the computing server 110 to the model server 120;

S440: The model server 120 generates a second solution code C2 according to the second model input file B2 and the prediction model;

S450: The hub server 150 transmits the second solution code C2 generated by the model server 120 to the database server 140;

S460: The database server 140 reads the second solution D2 from the huge database 130 according to the second solution code C2;

S470: The hub server 150 outputs the second solution D2 that the database server 140 reads from the huge database 130.

Through the operation methods 300 and 400 of the solution searching system 100 of the above embodiment of the present invention, the algorithm of the huge database and the data exploration can be utilized to enable the engineers to share their experiences in solving system problems in the past, and can easily find the possibility. Solutions to reduce time to resolve product issues and improve solution quality. In addition, the operation methods 300 and 400 can also check each solution code to avoid the situation that the solution code corresponds to too many or too few solutions, thereby achieving the effect of ensuring the prediction accuracy.

In summary, the solution search system and the solution search system operation method of the embodiments of the present invention can utilize a huge database and a data exploration algorithm to assist users in sharing their past experience in solving problems, while using When discovering system problems, it is easy to find possible solutions to reduce the time to resolve product issues. In this way, it is possible to avoid the problems in the prior art because the related solutions are not easy to search, and the efficiency and quality of solving the system problems are difficult to control. In addition, since the solution search system and the solution search system operation method of the embodiment of the present invention can also check each solution code to avoid having too many or too few solutions corresponding to the solution code. Therefore, the effect of ensuring the accuracy of the prediction can be further achieved. The above are only the preferred embodiments of the present invention, and all changes and modifications made to the scope of the present invention should be within the scope of the present invention.

100‧‧‧Solution Search System

110‧‧‧ Computing Server

120‧‧‧Model Server

130‧‧‧ huge database

140‧‧‧Database Server

150‧‧‧Central Server

160‧‧‧Related database

170‧‧‧Modeling Server

A1‧‧‧First problem description file

B1‧‧‧ first model input file

C1‧‧‧First Solution Code

D1‧‧‧ first solution

MC‧‧‧ main classification code

SC1‧‧‧ first classification code

SC2‧‧‧Second classification code

U1, U2‧‧‧ users

A2‧‧‧Second problem description file

B2‧‧‧Second model input file

C2‧‧‧Second solution code

D2‧‧‧Second solution

300, 400‧‧‧ method

Steps S310 to S380, S410 to S470‧‧

FIG. 1 is a schematic diagram of a solution search system according to an embodiment of the present invention. Figure 2 is a schematic diagram of the use scenario of the solution search system of Figure 1. FIG. 3 is a flow chart of an operation method of the solution searching system according to an embodiment of the present invention. 4 is a flow chart of an operation method of a solution search system according to another embodiment of the present invention. Figure 5 is a partial view of a standard word comparison table according to an embodiment of the present invention.

300‧‧‧ method

S310 to S380‧‧‧ steps

Claims (10)

A solution search system comprising: an operation server for comparing a common standard word comparison table with each of the solved first problem description files when receiving a plurality of solved first problem description files a textual comparison of the cause field of the problem to generate a primary classification code of the first problem description file of the solved problem, according to the first sub-standard word comparison table corresponding to the main classification code and the first problem solved The text in the file is compared with a first sub-category code for generating the first problem description file of the solved problem, and at least the main classification code of the file and the first sub-category code are generated according to the first problem of the solved first problem. The first problem code of the first problem is described as a first solution code, and a first model input of the solved first problem description file is generated according to the first problem description file and a data exploration algorithm. a file; a huge database for storing the solution; a database server coupled to the huge database for solving the problem The first solution code of the problem description file stores a first solution in the solved first problem description file to the huge database; a modeling server for performing a data exploration algorithm And the first model input file of the first problem description file and the first solution code to establish a prediction model; and a hub server coupled to the computing server, the database server, and the a modeling server, configured to: when receiving the solved first problem description file, transmit the solved first problem description file to the computing server; and generate the generated operation server The first problem solved by the file describes the first model input file corresponding to the file, and the first solution code is transmitted to the modeling server. The solution search system of claim 1, wherein: the computing server is further configured to: when receiving an unresolved second problem description file, generate the file according to the second problem description file and the data exploration algorithm The second problem describes a second model input file of the file; the solution search system further includes a model server coupled to the hub server for generating a second image according to the second model input file and the prediction model a solution code; the database server is further configured to read a second solution from the huge database according to the second solution code; and the hub server is further configured to: when the hub server receives the When the second problem is not solved, the unresolved second problem description file is transmitted to the computing server; and the second model input file generated by the computing server is transmitted to the model server; Transmitting the second solution code generated by the model server to the database server; and outputting the second solution that the database server reads from the huge database Program. The solution search system of claim 1, wherein the computing server is further configured to: when the number of the first solutions corresponding to a first solution code is greater than an upper limit, according to the first solution A second sub-standard word comparison table corresponding to a first-time classification code in the code is compared with the text in the solved first problem description file to generate a second of the solved first problem description file. a secondary classification code; and updating the first problem description file of the resolved first problem code according to the primary classification code of the first problem description file, the first secondary classification code, and the second secondary classification code One Solution code. The solution search system of claim 1, wherein the computing server is further configured to: when the number of the first solutions corresponding to a first solution code is less than a lower limit, and the one of the first solution codes Updating the first solution code to the most recent when the sum of the number of the first solutions corresponding to the classification code and the number of the first solutions corresponding to the first solution code is not more than an upper limit value Solve the classification code. The solution search system of claim 1, wherein the computing server is further configured to: when the number of the first solutions corresponding to a first solution code is less than a lower limit, and the one of the first solution codes When the sum of the number of the first solutions corresponding to the classification code and the number of the first solutions corresponding to the first solution code is not more than an upper limit, the first solution code and the most recently resolved The classification code is updated to a new resolution classification code. A method for operating a solution search system, the solution search system comprising a computing server, a huge database, a database server, a modeling server and a hub server, the method comprising: when the hub When the server receives the plurality of solved first problem description files, the hub server transmits the solved first problem description files to the computing server; the computing server compares a common standard word with the Each of the solved first problem descriptions a textual comparison of a problem generative field in the file to generate a primary classification code of the solved first problem description file; the computing server is corresponding to one of the main classification codes The first sub-standard word comparison table is compared with the text in the first problem description file of the solved first problem to generate the first problem solved. a first sub-category code of the file; the computing server generates the solved first problem description file according to at least the main classification code of the first problem description file and the first sub-category code a first solution code; the computing server generates a first model input file of the solved first problem description file according to the solved first problem description file and a data exploration algorithm; the database server The first solution code according to the solved first problem description file stores a first solution in the solved first problem description file to the huge database; the hub server uses the operation servo The first problem input file corresponding to the first model input file corresponding to the archive, the first solution code is transmitted to the modeling server; and the modeling server is based on a data exploration algorithm and The first problem input file of the solved first problem description file and the first solution code establish a prediction model. The method for operating a solution search system according to claim 6, wherein the solution search system further comprises a model server, the method further comprising: when the hub server receives an unresolved second problem description file Transmitting the unresolved second problem description file to the computing server; when the computing server receives the unresolved second problem description file, describing the file and the data exploration according to the unresolved second problem The algorithm generates a second model input file of the unresolved second problem description file; the hub server transmits the second model input file generated by the computing server to the model server; The model server generates a second solution code according to the second model input file and the prediction model; the hub server transmits the second solution code generated by the model server to the database server; The database server reads a second solution from the huge database according to the second solution code; and the hub server outputs the second solution that the database server reads from the huge database . The method for operating the solution search system of claim 6, further comprising: when the number of the first solutions corresponding to a first solution code is greater than an upper limit: the computing server is configured according to the first a second sub-standard word comparison table corresponding to a first-time classification code in the solution code is compared with the text in the solved first problem description file to generate a first file of the solved first problem description file. a secondary classification code; and the operation server updates the first problem solved according to the primary classification code of the first problem description file, the first secondary classification code, and the second secondary classification code Describe the first solution code for the archive. The method for operating the solution search system of claim 6, further comprising: when the number of the first solutions corresponding to a first solution code is less than a lower limit, and one of the first solution codes is recently resolved When the sum of the number of the first solutions corresponding to the classification code and the number of the first solutions corresponding to the first solution code is not more than an upper limit value, the computing server updates the first solution code to This recently resolved classification code. The method for operating the solution search system of claim 6, further comprising: when the number of the first solutions corresponding to a first solution code is less than a lower limit, and one of the first solution codes is recently resolved When the sum of the number of the first solutions corresponding to the classification code and the number of the first solutions corresponding to the first solution code is not greater than an upper limit value, the computing server copies the first solution code and the Recently resolved the classification code update to a new resolution classification code.