TWI557576B - Method and System for Predicting Calculation of Timing Data - Google Patents

Description

Method and system for predicting calculation of time series data

The invention patent is mainly used to protect the cloud resource analysis service. The service is used in the cloud environment, and is also planned to be applied to the hicloud cloud computing, the virtual private cloud or the separate sale, and can conveniently and effectively evaluate whether the existing cloud resources are effectively Utilize and analyze cloud resource information to achieve monitoring and forecasting alarm functions.

In the simple design of the third-party software, when the system needs to calculate, the one-time data will be accessed through REST, etc., and the computing load will be processed once. When the computing demand is increased, the REST access times will be more, resulting in the mother. The load of software and self-system is improved. Therefore, under the current third-party software design, the calculation will be carried out by self-storing historical data. However, when more and more original data is collected, the search performance will be worse. When a calculation takes several hours, the availability of the system will become very low, and the data should be assembled, that is, tiered storage. When the raw materials that have been used in a small amount should be removed and compressed into compressed files, Into, and establish a relay data layer on the original data layer, relay data to extract the general information, the data size is larger, but the number of relay data is less than the original data. In addition, under the simple architecture, most calculations rely on the performance of the database. When the data becomes more and more large, the availability of the data will become very low. When different and a large number of user interaction queries, the calculation load The larger the user's computing needs should be predicted, if the system passively relies on the user to query and do things, it can not effectively solve a large number of simultaneous calculations.

In view of the shortcomings derived from the above-mentioned conventional methods, the inventor of the present invention has improved and innovated, and after years of painstaking research, he finally succeeded in researching and developing the method and system for predicting and calculating the time series data.

The invention provides a method and a system for predicting and calculating time series data, and performs multi-level group management on the collected time series data and members, and can quickly access and reuse relay data for group calculation to save system performance. And the benefits of time.

The invention provides a system for predicting and calculating time series data, including a user, a multi-level group and a time series data management system, and a timing data source. The multi-level group and time series data management system includes a data source management module, a time series data collection and analysis module, a data database, a multi-level group management module, a group computing demand input module, a directory management module, The relay data calculation module, the calculation result output module and the self-learning module, wherein the self-learning module comprises a common directory loading unit and a pre-calculation unit. The user operates the multi-level group and the time series data management system to define the time series data source through the data source management module. The time series data collection and analysis module collects the original data imported from the time series data source and stores the original data into the data database and analyzes the data. Incorporating components and creating optional members, users can perform member selection and group management through a multi-level group management module. The user initiates system operations through the group calculation input module, and establishes through the directory management module according to the user's computing needs. Or obtain the existing catalogue. According to the catalog definition, the relay data calculation module performs preliminary calculations of multi-level groups and members, and the catalog definition can be calculated and adjusted through time precision. early After the step operation ends, the calculation result output module performs the final result calculation and outputs it to the user. The whole operation process is monitored by the self-learning module, and the usage behavior can be analyzed and preloaded through the common directory loading unit. The commonly used directory is used to improve the efficiency of obtaining the catalog in the cache list, and the pre-calculation unit can predict the user's computing requirements to pre-calculate the relay data for reuse.

The invention provides a method for predicting and calculating time series data, which comprises the following steps: Step 1: The user issues a calculation request to the group; Step 2: The group calculation demand input module analyzes the calculation requirements and forms a directory; Step 3: Table of Contents The management module searches the common directory; in step 4, the common directory loading unit calculates the difference coefficient 1 minus the similarity coefficient through the fuzzy concept, and deducts the usage rate; step 5, searches the directory in the data database; step six, creates This directory; step seven, the directory management module increases the usage rate of the extracted directory by one; step eight, the relay data computing module arranges the computing requirements according to the group members; step IX, searches the relay computing data in the data database; Step 10: The relay data calculation module starts to calculate and mounts the calculated data in the directory; Step 11: When the relay data has been calculated, the calculation result output module performs the final settlement action, and the result is Output.

The method and system for predicting and calculating the time series data of the present invention have the following advantages when compared with other conventional technologies:

1. The present invention provides a computing data reuse mechanism for a multi-level group. When the system takes out the group member time series data, the preliminary calculation is performed to generate a relay data, and the relay data is stored in the data library for reuse. Reuse can occur in the user's repeated query and multi-level group operation, according to this operation mechanism, can save system computing efficiency and user waiting time.

2. The present invention provides a data acquisition and operation mechanism for a multi-level group. The preliminary data of the member is used to establish a relay data, and a directory is created for the relay data. The directory is established according to the rules of the user query. The directory can quickly find the relay data, which can be quickly reused by the system.

3. The present invention provides a mechanism for loading a cache data directory into a cache. When the amount of relay data is larger, the number of directories will also increase. Therefore, according to the historical experience of the system, the self-learning function is performed, and the load is most loaded. A frequently used catalogue, and through the fuzzy concept, adjust the list of most frequently used catalogues by means of the coefficient of difference.

4. Through the design of relay data, the original data can be removed and backed up, reducing the data load of the database. In addition to the rapid query of the data, the catalogue is deleted, and the related relay data is deleted. Following the data deletion mechanism.

5. The present invention provides a mechanism for predicting user requirements and pre-calculating. The system can predict and pre-calculate relay data according to historical usage experience and increase or decrease the usage of different directories, so that the system can be quickly reused in the next calculation. Pre-computation can also solve the performance problems caused by the large number of simultaneous operations.

100‧‧‧Users

200‧‧‧Multi-level group and time series data management system

201‧‧‧Source Management Module

202‧‧‧Time Series Data Collection and Analysis Module

203‧‧‧Data Database

204‧‧‧Multi-level group management module

205‧‧‧Group Computing Demand Input Module

206‧‧‧Directory Management Module

207‧‧‧Relay data calculation module

208‧‧‧ Calculation result output module

209‧‧‧Self-learning module

2091‧‧‧Common directory loading unit

2092‧‧‧ Pre-calculation unit

300‧‧‧Time source

S201~S211‧‧‧Step Diagram

S301~S304‧‧‧ Pre-calculated demand step diagram

The detailed description of the present invention and its accompanying drawings will be further understood. The technical content of the present invention and its functions are as follows: FIG. 1 is a system architecture diagram of a method and system for predicting time series data according to the present invention; 2 is a step diagram of a method and system for predicting calculation of time series data; FIG. 3 is a diagram of a pre-computation requirement step of a method and system for predicting calculation of time series data.

The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Hereinafter, the present invention will be further described with reference to the accompanying drawings:

Referring to FIG. 1, the present invention provides a system for predicting and calculating time series data, including a user 100, a multi-level group and timing data management system 200, and a time series data source 300.

The multi-level group and time series data management system 200 includes a data source management module 201, a time series data collection and analysis module 202, a data database 203, a multi-level group management module 204, and a group computing demand input module 205. The directory management module 206, the relay data calculation module 207, the calculation result output module 208, and the self-learning module 209; wherein the self-learning module 209 includes a common directory loading unit 2091 and a pre-computing unit 2092.

The user 100 operates the multi-level group and the time series data management system 200 to define the time series data source 300 through the data source management module 201, and the time series data collection and analysis module 202 collects the collected data from the time series data source 300. The original data is stored in the data database 203, and the data components are analyzed and optional members are created. The user 100 can perform member selection and group management through the multi-level group management module 204, and the user 100 calculates the input module through the group. 205 starts the system operation, establishes or obtains the existing directory through the directory management module 206 according to the computing requirements of the user 100, and according to the directory definition, the relay data calculation module 207 performs preliminary operations of the multi-level group and the members, and the directory definition thereof The calculation and adjustment can be performed through time precision. After the preliminary calculation is completed, the calculation result output module 208 performs the calculation of the final result and outputs it to the user 100. The overall calculation process is monitored by the self-learning module 209, and is commonly used. The directory loading unit 2091 can analyze the usage behavior and pre-load the commonly used directory to improve the efficiency of obtaining the directory in the cache list, and the pre-calculation unit 2092 can predict the user computing requirement to pre-calculate the relay data for reuse.

The method for predicting and calculating the time series data is as shown in FIG. 2 and FIG. 3, and the steps are as follows: Step 1: The user of S201 issues a calculation request to the group, which means that the user passes the multi-level group management module. Calculate the requirements for a group; Step 2: The S202 group calculates the demand input module to analyze the calculation requirements and form a catalog, which means the user's calculation requirements. The requirements should include the group, the time interval of the calculation, and the data. Granularity, results to be calculated (numerical output or sequence output), time definition (those to be excluded and allowed), and then a directory is created according to these requirements, wherein the result to be calculated is an optional item that the system can provide. For example, the maximum value, the minimum value, the average value, or the time series of a minimum value, a directory is generated based on a group of members and a computing requirement, and the organized directories can be grouped by different groups. Reuse; Step 3, S203 directory management module searches for common directories, which means that the above step 2 enters the common directory search according to the formed directory, and the search method is to compare the contents of all the directory definitions, and the output is to find the directory or check no. Table 4; Step 4, S204 common directory loading unit calculates the difference coefficient 1 minus the similarity coefficient through the fuzzy concept, and deducts the usage rate, which means that the difference coefficient is calculated according to the fuzzy concept created in the above step 3. And the difference coefficient is deducted from the usage rate of the commonly used catalogue, wherein the difference coefficient is 1 minus the similarity coefficient, the similarity coefficient uses the Dice coefficient to calculate the similarity coefficient of each definition in the catalog, and the average similarity coefficient is obtained to obtain the common catalogue pair. Comparing the similarity coefficient of the catalogue, the calculation of the difference coefficient is calculated for the common catalog (cache), instead of entering the data database to retrieve all the catalogues for calculation; step 5, S205 searching the catalogue in the data repository, meaning step After three investigations without a directory, this step enters the data database search, and its directory comparison The same as step 3, the output is to find the directory or check the directory; step six, S206 create this directory, meaning that when step 5 does not have a directory, this step creates this directory in the data database, its initial usage rate is zero, And output the directory; Step 7. The S207 directory management module adds 1 to the extracted directory, which means that the usage rate is added according to the directory extracted in the above steps 4 and 6, and is updated to the data database. And output this directory; Step 8: The S208 relay data calculation module arranges the calculation requirements according to the group members, which means that according to the directory entered in step 7 and for the group and members to be calculated by the user, the calculation time is based on the data granularity of the directory. The definition is cut into small segments, and the relay data calculation arrangement is performed according to each small time. This step can receive the pre-computation requirements transmitted by the component. This step is only responsible for arranging the calculation work of the relay data, when different users at the same time When the same directory is used for calculation, this step will also be performed in a queue. The calculation time must be greater than the data granularity of the directory to facilitate time cutting. Step 9. S209 searches for the relay calculation data in the data database, which means Step 8: The calculated work search data database 23 is checked to see if it has been calculated, and the output is checked for no calculation or calculated; Step 10, S210 relay data calculation module starts calculation and the calculated data is mounted on In the directory, it means that the calculation of the relay data is started according to the uncalculated work introduced in the above step S209, and the result is mounted on the phase. The directory and the data database 23 are stored, wherein when the raw data is converted into the relay data process, the definition and the cull form of the directory can be enabled or disabled through the adjustment time precision method; Step 11: S211 when the data is relayed The calculation has been completed, and the calculation result output module performs the final settlement action, and outputs the result, which means that the relay data to be used in this calculation is taken out from the data database, there is no shortage, and all relays are The data is integrated, the results are calculated according to the user's calculation requirements, and the results are calculated. Output.

It can be known from the above steps that the steps from the user inputting the calculation requirement to completing the calculation include the steps of reuse, fast calculation and pre-calculation. First, in step S201, the user issues a calculation request to a group (for example, calculating a numerical average over a period of time of the group), step S202 analyzes the calculation request and sets up a directory, and the directory is formed according to the group member and the user. Enter the time requirement. Step S203 searches according to the directory. If the same directory is found, the process proceeds to step S204. Step S204 calculates the difference coefficient for each common directory according to the fuzzy-based concept, and deducts the usage rate. The coefficient, if the common directory is [a, b, c, d], the user needs b, then the difference coefficient of a to b, c to b and d to b is calculated, and after calculating the coefficient value, The usage rates of a, c, and d are deducted, and the algorithm of the difference coefficient is (1-similarity coefficient). If the directory is not found, the process proceeds to step S205 to search for the database. If it is found, the process proceeds to step S204, and the function is as described above. If it is still not found, the directory is created in step S206. Step S207, after the required directory has been taken out, the usage rate is increased by one, and the process proceeds to step S208 to start the relay data calculation arrangement of the group member, and in step S209, the calculation required for searching the database already exists, if not exists, if not Then, in step S210, the data calculation is performed. When the relay data has been calculated, the process proceeds to step S211 to perform the final calculation and the result output.

In step S204, after waiting for a defined time, the common directory searches for the data database, and according to the usage rate of the directory, the directory with high usage rate is updated in the common directory table, and the step is to follow the loop. Constantly update the common directory table in the cache according to the usage rate. The concept is like Least Recently Used (LRU), but the implementation content is different from the traditional usage rate, and the difference coefficient is used to deduct the usage rate. This way the common table of contents is closer to most user needs.

The pre-calculation requirement in step S208 is as shown in FIG. 3, and the steps are as follows: Step 1: After S301 waits for a defined time, the pre-calculation unit periodically records the usage rate of the current common directory, which means that after waiting for a certain time. Record the current usage rate of commonly used directories, in which the record number and historical usage data are stored in the cache, not in the data database; Step 2, S302 when the number of records meets its defined number of times, meaning the number of current records is judged If the number of times is less than the defined number of times (at least three times), it will continue to wait for the next record. If the number of records is satisfied, the prediction step is performed. Step 3: If the number of times defined by S303 is satisfied, the pre-calculation unit will record the historical usage data in the current common directory. The number of pens, the catalogue that meets the number of defined pens, is used to predict the usage rate, and the high-predicted catalogue is taken out, which means that the historical usage rate (at least three strokes) is satisfied for the current common catalogue, and the historical usage growth value is taken out and calculated. Use this growth series to make predictions. After calculating the next growth value of the forecast, sort out the top three (maybe less than three) And output, if the history of the common directory is not satisfied with its forecast demand, then no prediction is needed, and the prediction directory is empty, wherein the prediction calculation is performed for the common directory (cache) instead of entering the data. The database retrieves all the catalogues for calculation; in step 4, S304 searches for the data repository based on the predicted catalog, retrieves the corresponding group members and performs pre-calculation, and returns to step one, meaning that the output predicted list is taken out from the data repository. Corresponding group The member of the group performs a pre-computation notification and sends the notification to the relay data calculation module. When the directory output in step 3 is empty, the pre-calculation is not performed, and the next return is performed, and the data size of the predicted directory is required. If the waiting time is less than the waiting time of step 1, the pre-calculation notification is performed; if the number of times is not satisfied, the step 1 is returned.

In summary, the loopback is performed. After the step S301 waits for a defined period of time, the pre-calculation unit periodically records the usage rate of the current common directory. In step S302, when the number of records meets the defined number of times, step S303 is pre-calculated. The unit will use the historical usage data in the current common directory to meet the definition of the number of records to use the rate prediction, take out the high-predicted directory, and check the current common directory, and take out the number of records that meet the certain record usage rate. The directory is predicted, but since the common directory is continuously updated in step S204, the number of records used in each new favorite directory is zero, so step S303 predicts according to the directory that satisfies the record usage rate, and the prediction is made. The method adopts gray prediction, and the gray prediction characteristic is fast prediction based on discrete data of a small number of pens. After predicting the next usage rate, the predicted values are sorted, the high usage rate directory is taken out, and the process proceeds to step S304 based on the prediction list. , search the data database, take out the corresponding group members and perform pre-calculation, search the data database This catalog was a member of the group, and before notifying step S208 pre-calculated.

According to the flow of FIG. 1, FIG. 2, and FIG. 3, a better example is implemented. When the user 1 inputs three time series data source definitions [source1, source2, source3], the time series data collection and analysis module 202 starts to continuously collect data. Such as: source1[object1, object2], source2[object1], source3[object1] Each object contains a series of time series data [raw data1 in timestamp1, raw data2 in timestamp2...], which increases according to the continuous increase of time. As shown in formula one:

The timing data collection and analysis module 202 analyzes the association of the source-object-raw data and fills in the data repository 203. The multi-level group management module 204 will express its association so that the user 100 can perform group management, for example, The user creates two groups [group1, group2], where group1 is [member1(object1 in source1), member2(object1 in source3)], group2 is [member1(group1)], and the member in group2 can be found as group1. This is a multi-level group design, and each member can have its own directory, as shown in Equation 2:

The user 100 inputs the calculation requirement through the group calculation demand input module 205. For example, the user wants to obtain the maximum value of the raw data of the group4 2014-04-13 00:00:00 to 2014-04-19 23:59:59. And the maximum value appears on that day, but the calculation only takes out the raw data from 08:00 to 16:59 daily. The group computing demand input module 205 analyzes the information input by the user and forms a relay data directory. The directory expression is as shown in Table 1 below:

Using the catalog for data cutting, the user's calculation demand is the maximum value that appears on that day, the catalogue data size is days, and will be 2014-04-13 00:00:00 to 2014-04-19 23:59:59 Cut into seven days in days. The definitions of year, month, week, hour and minute are all valid. The definition of day is 08-16 hours is valid, and the remaining 00-07, 17-23 hours is invalid, year, month, day, hour, minute and second. The culling form is empty, and a directory is generated based on a group of members and a computing requirement.

Searching the common directory form (cache) through the directory management module 206 to check whether there is any such directory, if not, searching to the data repository 203, if not, creating the directory; if the same directory is found in the common directory form or data repository 203 Then, the difference coefficient calculation is started. For example, the current common directory form is [directory 1, directory 2, directory 3, ..., directory 10], and the directory found is directory 11 (found in data database 203), and the common directory contains The entry unit 2091 starts the calculation of the difference coefficient between the directory 1 to the directory 11, the directory 2 to the directory 11 and the like to the directory 10 to the directory 11, and the difference coefficient is (1 - similarity coefficient), as shown in the formula 3:

Since the coefficient of difference is (1-similarity coefficient), the similarity coefficient needs to be calculated first. According to the fuzzy-base concept, the similarity between the two objects can be calculated by the difference between the two objects. The similarity coefficient of this example is calculated by Dice coefficient, as shown in Equation 4:

For directory 1 versus directory 11, the similarity factor for the calculation of the day definition is as follows:

Directory 1 (example):

Day (definition): 08-09 (vaild), Rest (invaild)

...

Directory 11 (example):

Day (definition): 08-16 (vaild), Rest (invaild)

...

Directory 1 (day definition) → directory 11 (day definition)

Day definition (dice similarity coefficient):

A value of 0.08 can be obtained, and the definitions of the remaining years, months, weeks, hours, and minutes, and the similarity coefficients of the shaving form and the data granularity can be calculated by analogy. All the similarity coefficient values are obtained and averaged, and the value is the similarity coefficient of the directory 11 to the directory 1, as shown in the formula:

After taking the similarity coefficient of all the common directories to the directory 11, the difference coefficient is obtained through the formula 3, for example, the directory 1 to the directory 11, the directory 2 to the directory 11, and the like to the directory 10 to the directory 11 is [0.8, 0.9, 0.1, 0.2, 0.2 , 0.1, 0.5, 0.2, 0.9, 0.4], subtract the difference coefficient from its catalog usage rate, such as the original usage rate of catalog 1 is 12.9, minus 12.1, 12.1, and other common catalogues, and so on. The directory 11 performs the action of adding 1 to the usage rate. In step S207, after waiting for a period of time, the common directory table is updated according to the directory usage rate. The main purpose of the action is to learn the behavior of the user, dynamically adjust the common directory, and deduct the difference coefficient. In this way, the similarity coefficient comparison can be performed only for the common directory in the cache, rather than for each directory in the data database, which can reduce the computational loss caused by the data comparison.

Step S208 starts to perform relay data calculation demand scheduling for the group. In this example, the member of group2 is group1, and since group1 is a group, it is further calculated for group1, which is a feature of the multi-level group, group1 The members are objectA (object1 in source1) and objectB (object1 in source3), searching for group1 and the directories of the two objects (based on this computing requirement) 2014-04-13 00:00:00 to 2014-04-19 23 : There is no calculated relay data between 59:59. According to the design of the patent directory structure, the calculated relay data can be reused, as shown in the following list 2:

The calculated data may be caused by the previous user's calculation for group1, so the interval for calculating the relay data is: objectA{2014-04-16,2014-04-17,2014-04-18,2014-04 -19}, objectB{2014-04-16,2014-04-17,2014-04-18,2014-04-19} and group1{2014-04-16,2014-04-17,2014-04-18 , 2014-04-19}.

After the relay data is data cut, the preliminary data is calculated according to the granularity of each data. For example, the user 100 needs to find the maximum value for a period of time (2014-04-13 00:00:00 to 2014-04) -19 23:59:59) and list the maximum value on that day, the system will divide this time into several small time (days), and then calculate the maximum value of each small time (day). The mechanism and relay data calculation will occur in steps S209 and S210. A relay data is converted from a group of raw data, and when the raw data is calculated, the definition of the directory and the culling form can be adjusted through time precision. The stringency of raw data varies according to the accuracy of time. For example, when the time precision is minute, the definition of the directory and the second elimination form are invalid. Therefore, when collecting the relevant raw data for calculation, the directory can be adjusted. Time accuracy to meet system design requirements, the time accuracy of this example is seconds (minimum Unit), so all definitions and culling forms in the directory are enabled for raw data. After the relay data has been calculated, step 411 is performed to perform the final rounding calculation, and the result is output. In this example, the maximum values of objectA and objectB are unified, and the maximum value is averaged to each data granularity of group1. In the calculation, the maximum value of each data granularity of group1 is compared, and finally the maximum value and the date of occurrence (days) are taken out and output, that is, the calculation is completed.

In step S208, the pre-calculation unit 2092 will perform user calculation demand prediction. After the system is running for a period of time, the usage change of the directory can reflect the calculation requirements of most users, so as shown in FIG. The calculation unit 2092 periodically records the usage change of the common directory, and performs the prediction action after the recording is completed a certain number of times (at least three times). According to the number of usage records recorded in each directory in the common directory, after the number of pens meets a certain number (minimum of three strokes), the prediction is started. The prediction method of this example uses gray prediction, and the gray prediction can utilize a small number of pens. The number can achieve the effect of fast prediction. First, the historical usage rate is subtracted before and after the increment series is taken. Because the gray prediction needs to be accumulated, the calculation series is calculated as Equation 6:

The gray difference equation is required according to the gray modeling, as shown in the formula: increment ( h ) + slope * z ( h ) = intercept h =1, 2,..., q , q < g (Equation 7)

z is the background value sequence, λ is the background value parameter, generally set to 0.5, as shown in Equation 8: z ( h ) = λ * accumulation ( h ) + (1 - λ ) * accumulation ( h -1) λ = 0.5 h =2,..., q , q < g (Equation 8)

Equation 7 can obtain the slope and intercept parameters according to the least squares method, and the time response can be obtained by mathematical derivation, as shown in formula IX:

Finally, through the reduction generation, the prediction series is gradually obtained. To predict the value of the next step, only the parameters of the formula 9 need to be adjusted, as shown in the formula:

In this example, it is only necessary to predict the value of the next step, that is, p=1. After predicting the next use rate increase value, the sorting takes out the top three most popular common directories (there may be less than three cases) , for pre-calculation. First, all the data databases are searched and the group members corresponding to the predicted directory are taken out for pre-calculation, that is, steps S304 and S209. The advantages of its pre-computation can be as shown in Table 2 above. The calculated relay data may be reused for the next user calculation to speed up the calculation of the output speed.

After the prediction directory is generated and notified to the relay data calculation module 207, the module determines the relay data calculation schedule based on the data granularity of the prediction directory and the waiting time of the pre-calculation unit 2092. For example, the waiting time of the pre-calculation unit 2092 is set to Step S301 is performed every month, and the data granularity of the predicted directory is less than monthly to perform pre-calculation. For example, step S208 predicts the directory 11 and predicts the relay data calculation module 207 at 2014-04-01, and the directory 11 data granularity is In the day, this module will pre-calculate all the days from 2014-04-01 to 2014-04-30 for the corresponding members of the forecast catalog. By 2014-05-01, component 6 will make the next forecast, and so on.

The detailed description of the present invention is intended to be illustrative of a preferred embodiment of the invention, and is not intended to limit the scope of the invention. The patent scope of this case.

To sum up, this case is not only innovative in terms of technical thinking, but also has many of the above-mentioned functions that are not in the traditional methods of the past. It has fully complied with the statutory invention patent requirements of novelty and progressiveness, and applied for it according to law. Approved this invention patent application, in order to invent invention, to the sense of virtue.

100‧‧‧Users

200‧‧‧Multi-level group and time series data management system

201‧‧‧Source Management Module

202‧‧‧Time Series Data Collection and Analysis Module

203‧‧‧Data Database

204‧‧‧Multi-level group management module

205‧‧‧Group Computing Demand Input Module

206‧‧‧Directory Management Module

207‧‧‧Relay data calculation module

208‧‧‧ Calculation result output module

209‧‧‧Self-learning module

2091‧‧‧Common directory loading unit

2092‧‧‧ Pre-calculation unit

300‧‧‧Time source

Claims (8)

A system for predicting and calculating time series data, the system comprising a data source management module, a time series data collection and analysis module, a data database, a multi-level group management module, a group computing demand input module, a directory management module The group, the relay data calculation module, the calculation output module and the self-learning module, wherein: the data source management module receives the user's input source setting such that the time series data collection and analysis module collects time series data. The original data transmitted from the source; the time series data collection and analysis module is to analyze the original data transmitted from the source of the time series data, and identify the data object association and write the data database; the multi-level group management The module is to retrieve the data object association from the data database, and let the user create a group, and establish a multi-level group feature, and the group has another group, the members can be included in different The structure of the group; the multi-level group, the multi-level group includes a group structure of single-level or single-group members; the group computing demand input module To identify the computing needs of the user and to form a directory, the directory is generated based on a group of members and a computing requirement; the directory management module is configured for the directory in which the group computing demand input module is formed. Search, add 1 to the searched directory usage rate and calculate the difference coefficient between the directory and the other commonly used directories, and the difference coefficient is deducted from the usage rate of the compared directory, wherein the difference coefficient is calculated for the common directory. Calculate instead of entering the data repository There is a directory for calculation; the relay data calculation module performs relay data calculation for the directory taken out by the directory management module and the group to be calculated; the calculation output module is for calculating the relay data. The relay data completed by the group is calculated and outputted; and the self-learning module is to periodically update the common directory and predict the computing needs of the user. For example, the system for predicting and calculating the time series data described in the first paragraph of the patent application, wherein the relay data and the directory cannot be shared and reused by different groups. The system for predicting and calculating the time series data according to the first aspect of the patent application, wherein the self-learning module further comprises a common directory loading unit and a pre-computing unit, wherein: the common directory loading unit is through the sorting data. The usage of the directory in the database is high and low, and the common directory is updated regularly; the pre-computing unit finds the historical usage rate of the common directory and uses the growth rate series to predict the next directory growth value. The user's usual computing needs are pre-calculated for reuse. A system for predicting and calculating time series data according to claim 3, wherein the common directory has a history usage record number of zero. The system for predicting and calculating the time series data according to the third aspect of the patent application, wherein the pre-calculation unit notifies the relay data calculation module of the predicted common directory, and performs pre-calculation on the predicted directory and related members. A system for predicting and calculating time series data as described in item 3 of the patent application, The pre-computation is calculated for common directories, rather than entering the data repository to retrieve all the catalogues for calculation. The system for predicting and calculating time series data according to claim 5, wherein the relay data calculation module determines a relay data calculation schedule based on a data granularity of the predicted directory and a waiting time of the pre-calculation unit. A method for predicting calculation of time series data comprises the following steps: Step 1: The user issues a calculation request to the group; Step 2: The group calculation demand input module analyzes the calculation requirements and forms a directory; Step 3: The directory management module Search for commonly used directories; Step 4: The common directory loading unit calculates the difference coefficient (1-similarity coefficient) for the remaining common directories through the fuzzy concept, and deducts the usage rate; Step 5: Search the directory in the data database; Step 6: Create this Table 7; Step 7: The directory management module adds 1 to the extracted directory; Step 8: The relay data calculation module arranges the calculation requirements according to the group members, wherein the computing requirements include: Step 1. Wait for a defined time Afterwards, the pre-calculation unit periodically records the usage rate of the current common directory; step 2. when the number of records meets its defined number of times; step 3. If the number of definitions is satisfied, the pre-calculation unit records the historical usage data in the current common directory. , satisfying the usage forecast of the directory whose number of definitions is defined, and taking out the directory of high prediction; Step 4. Based on the prediction directory, search the data database, take out the corresponding group members and perform pre-calculation, and go back to step 1; Step 5. If the number of definitions is not satisfied, go back to step 1; Step 9: Search relay calculation The data is in the data database; in step 10, the relay data calculation module starts to calculate and mounts the calculated data in the directory; step 11: when the relay data has been calculated, the calculation result output module performs the last The action is settled and the result is output.