JP6668657B2 - Data processing program, data processing method and data processing device - Google Patents

Description

  The present invention relates to a data processing program, a data processing method, and a data processing device.

  2. Description of the Related Art A CEP (Complex Event Processing) system is known as a technique for processing big data collected from various objects every moment. For example, the CEP engine processes a plurality of data received from various targets according to a predetermined rule (hereinafter, also referred to as “event processing”), and sequentially outputs the results of the event processing to an output destination.

  Big data includes data from which valid data to be output to an output destination cannot be extracted as a result of event processing. That is, the big data includes data necessary for the CEP system and unnecessary data. Data from which valid data to be output to the output destination cannot be extracted is data that does not need to be input to event processing.

  On the other hand, when the administrator manually selects necessary data and unnecessary data, the number of operations of the administrator increases, and the operation cost increases. In addition, since data selection depends on the skill of the administrator, necessary data and unnecessary data cannot be appropriately filtered, and the load on the system may remain high.

  Therefore, there is known a technique for monitoring event processing for events of each event type stored in the reception buffer and suppressing the number of events stored in the reception buffer based on the determined importance of each event type as a result of the monitoring. (For example, see Patent Document 1).

JP 2012-118928 A

  However, the determination as to whether or not the data should be input to the event processing changes during the operation of the CEP system, and the optimality of the initially set rules may deteriorate over time. For example, at the beginning of operation, "passenger car event" was regarded as unnecessary data. However, as operation proceeds, "passenger car event" has changed to necessary data.

  On the other hand, if the rules for discarding received events are fixed as in the cited document 1, necessary data can be discarded before event processing, and unnecessary data can be discarded before event processing. Or not. If the necessary data is discarded before the complex event processing, the load of the event processing can be reduced, but the usefulness and validity of the result of the event processing decrease. If unnecessary data cannot be discarded before event processing, the load on the system cannot be reduced.

  For example, when all the data of the received event is input to the event processing, the load of the CEP engine that performs the event processing increases, so that it is necessary to increase the capacity of the CEP system, and the operation cost increases. In addition, it is difficult to increase the capacity of the CEP system to the extent that it can quickly adapt to the rapidly increasing data input amount in recent years.

  Therefore, in one aspect, an object of the present invention is to appropriately set a filter condition for filtering a plurality of received events before performing event processing based on a predetermined rule.

In one aspect, a data processing program that causes a computer to perform an event process on received event processing target data in accordance with a predetermined rule and sequentially output data after the event process, the data processing program comprising : in the target data, we obtain an evaluation value on the data after the event processing for each first period type of the first event, in the plurality of types of said event processing data, different types of the first event About the second event, in the first long second period than the first event type event processing target data, and the second event type event processing target data correlation coefficient value of is calculated, it is identified from the evaluation value, first for the first event type of event processing target data Filter condition, and, based on the correlation coefficient value, sets the second filter condition regarding the event processing data type of the second event, the data processing program is provided.

  According to one aspect, it is possible to appropriately set a filter condition for filtering a plurality of received events before performing event processing based on a predetermined rule.

The figure which shows the outline | summary of streaming processing. The figure which shows an example of a CEP system. The figure for explaining one of the problems of a CEP system. The figure for explaining one of the problems of a CEP system. FIG. 2 is a diagram illustrating an example of a functional configuration of a server device according to an embodiment. FIG. 4 is a diagram for explaining an example of automatic analysis (feedback) of correlation according to one embodiment. FIG. 4 is a diagram illustrating an example of a time-series data table, a setting table, and a first correlation table according to the embodiment. 9 is a flowchart illustrating an example of event processing according to an embodiment. 9 is a flowchart illustrating an example of an automatic generation process according to an embodiment. 9 is a flowchart illustrating an example of a correlation analysis process (evaluation value) according to an embodiment. FIG. 4 is a diagram illustrating a method for automatically generating a filtering rule according to an embodiment. FIG. 6 is a diagram illustrating an expiration date of a filtering rule according to an embodiment. 9 is a flowchart illustrating an example of a correlation analysis process (new event) according to an embodiment. FIG. 6 is a diagram for explaining an example of automatic analysis of a correlation (new event) according to an embodiment. FIG. 6 is a diagram illustrating an example of a time-series data table and a second correlation table according to the embodiment. FIG. 4 is a diagram illustrating a method for automatically generating rules for a CEP engine according to one embodiment. FIG. 6 is a diagram illustrating an example of a time-series data table and a second correlation table according to the embodiment. FIG. 4 is a diagram illustrating a method for automatically generating rules for a CEP engine according to one embodiment. FIG. 2 is a diagram illustrating an example of a hardware configuration of a server device according to an embodiment.

  Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the specification and the drawings, components having substantially the same function and configuration are denoted by the same reference numerals, and redundant description is omitted.

[Overview of Streaming Processing]
Before describing a CEP system according to an embodiment of the present invention, an overview of streaming processing will be described with reference to FIG. A server device 90 that performs event processing in the CEP system can communicate with a traffic control server 20 that performs road traffic control via a network.

  To the server device 90, data of a plurality of events collected from various targets is distributed. FIG. 1 shows, as an example of a plurality of event processing target data, data of a car event (CarEvent), a truck event (TrackEvent), and a weather event (SensorEvent) having different event types. However, the plurality of events are not limited to this, and include various types of events. For example, the event information may be sensor information acquired from sensors such as a temperature sensor, a light detection sensor, an acceleration sensor, and a pressure sensor, and information collected from devices such as home appliances.

  In the server device 90, a CEP engine 91 operates. The CEP engine 91 is software that implements complex event processing (hereinafter, also referred to as “event processing”) for processing data of a plurality of received events based on a predetermined rule. As software that realizes complex event processing, for example, Esper is cited. The composite event processing may be referred to as ESP (Event Stream Processing), but is collectively referred to herein as CEP including ESP.

  The server device 90 stores an event processing rule 95 that defines conditions for performing event processing on a plurality of events. In FIG. 1, the server device 90 stores a rule A and a rule B as an example of the event processing rule 95.

  Rule A indicates a condition that the average speed of a passenger car event is 50 km / h or less and the weather is rainy. Rule B indicates a condition that the average speed of the track event is 50 km / h or less and the weather is rainy. The event processing rules 95 are manually set, for example, by the administrator of the CEP system.

  The server device 90 receives the plurality of distributed data items 22, 23, and 24. The CEP engine 91 performs event processing on each data according to the event processing rules 95 (rules A and B). When a car event satisfies rule A or a truck event satisfies rule B, the CEP engine 91 determines an action of "possible congestion" and recommends the determined "possibility of congestion". The information is transmitted to the traffic control server 20.

  When the traffic control server 20 receives the recommendation transmitted from the server device 90, the traffic control server 20 performs control to reduce traffic congestion on the assumption that traffic congestion will occur on the road on which traffic conditions are to be monitored. For example, the traffic control server 20 performs control to increase the period of the green light on the road on which traffic congestion is predicted for a signal provided on the monitored road, and determines the number of passenger cars that can pass on the monitored road. increase. In addition, for example, the traffic control server 20 notifies traffic information in which traffic congestion is predicted to occur on a road to be monitored, prompts a detour to another road, and controls to avoid traffic congestion.

  Big data includes data from which valid information to be output to an output destination cannot be extracted as a result of event processing. That is, the big data includes data necessary for the CEP engine 91 and unnecessary data.

  On the other hand, as shown in FIG. 2, consider a case where the administrator manually sets the filtering rule 96 (filter condition). In this case, the filter engine 92 removes data that satisfies the filtering rule 96 as unnecessary data from the distributed speed data 22, 23, the weather data 24, and the newly distributed speed data 25 of the motorcycle event, and removes the remaining data. Only to the CEP engine 91. However, if the administrator manually separates necessary data from unnecessary data, the number of operations of the administrator increases and the operation cost increases. In addition, since data selection depends on the skill of the administrator, necessary data and unnecessary data cannot be appropriately filtered, and the load on the system may remain high.

  On the other hand, when no filtering is performed and the CEP engine 91 processes all data of the received event, the load on the CEP engine 91 increases. In recent years, it has been difficult to quickly adapt to rapidly increasing data input volumes to enhance the CEP system.

  In addition, as shown in FIG. 3, consider a case where a threshold value serving as a filter condition of the filter engine 92 is a fixed value. In this case, the filter engine 92 regards the speed data 23 in which the number of track event speed data 23 input exceeds 600 events in one hour and the match rate of the CEP rule exceeds 70% as an event of low importance. to decide. As a result, the speed data 23 is filtered and not input to the CEP engine 91.

  If a fixed filter for each reception event (truck event or passenger car event) or a filter corresponding to the matching rate is provided in front of the CEP engine 91 like the event rate and the match rate in FIG. Can be reduced. On the other hand, the fixed filter and the filter corresponding to the matching rate cannot respond to the case where the initially set rules (filter conditions) are not appropriate rules due to the passage of time. Is less useful and relevant. In particular, if the filter of the CEP engine 91 does not correspond to the change in the importance of the received event, the filter becomes inappropriate.

  For example, the event rate of a track event varies greatly. Depending on the time zone, the day of the week, or the season, the event rate is always 600 events / hour or more, and many of the track events (important events) that need to be processed by the CEP engine 91 may be filtered. . As a result, the usefulness and validity of the output result of the CEP engine 91 may be significantly reduced, a useful recommendation may not be transmitted without detecting a sign of congestion, and congestion may occur.

  Therefore, if the data supplied to the CEP engine 91 is not excessively reduced by the filter engine 92 in order to maintain the usefulness and validity of the output result of the CEP engine 91, the load on the CEP engine 91 cannot be reduced. .

  In FIG. 4, as in FIG. 3, the filter engine 92 converts the speed data 23 in which the number of track event speed data 23 input exceeds 600 events / hour and the match rate of the CEP rule exceeds 70%. Judge as an event with low importance. As a result, the speed data 23 is filtered and not input to the CEP engine 91.

  For example, if the track event is an unimportant event unrelated to the traffic congestion factor, or changes to an unimportant event during operation, the speed data 23 of the track event that should be originally filtered out is not removed. It is input to the CEP engine 91. As a result, the load on the CEP engine 91 increases, and the CEP engine 91 cannot process any more, and the system operation stops.

  In view of the above problems, an embodiment of the present invention provides a CEP system capable of appropriately setting a filter condition for removing a plurality of received events before event processing based on a predetermined rule. Hereinafter, after an example of the functional configuration of the server device 10 is described, the operation of the CEP system including the server device and each process performed by the CEP system will be described.

[Functional configuration of server device]
An example of a functional configuration of the server device according to an embodiment of the present invention will be described with reference to FIG. The server device 10 is an example of a data processing device that performs event processing on a plurality of received events in accordance with a predetermined rule, and sequentially outputs the processed events.

  The server device 10 is a physical server that performs complex event processing, and is, for example, a data sensor or a server computer provided in each company. Data collected from various objects is distributed to the server device 10. The server device 10 filters data of a plurality of types of events received from various targets according to preset filter conditions. Next, the server device 10 performs an event process on data other than the data removed by the filtering, generates a recommendation according to a result of the process, and outputs the recommendation to an output destination.

  The server device 10 includes a communication control I / F (interface) unit 11, a storage unit 12, and a control unit 13. The communication control I / F unit 11 is an interface that has at least one port and controls communication with a network. The communication control I / F unit 11 receives event data distributed from various targets via a network. For example, the communication control I / F unit 11 receives the speed data 22 of the car event, the speed data 23 of the truck event, and the weather data 24 of the sensor event shown in FIG. The communication control I / F unit 11 transmits and receives various data to and from the traffic control server 20. For example, the communication control I / F unit 11 determines an action to be recommended when the CEP engine 15 determines that the event processing rule is met, and transmits the recommended data to the traffic control server 20.

  The storage unit 12 is a storage device such as a hard disk and an optical disk. The storage unit 12 stores various data. For example, the storage unit 12 stores an OS (Operating System) executed by the control unit 13 and a program for performing various processes such as event processing, automatic generation processing, and correlation analysis processing to be described later. Further, the storage unit 12 stores various data required for executing the program executed by the control unit 13.

  The storage unit 12 stores data of the time-series data table 17, the setting table 18, the first correlation table 19, and the second correlation table 219. The storage unit 12 generates time-series data and stores it in the time-series data table 17. The time-series data table 17 in FIG. 7A includes hourly average speed data 171 for a passenger car event, hourly evaluation value (feedback value) 172 for a car event, and average hourly speed for a truck event. Data 173 and hourly evaluation values (feedback values) 174 of track events are stored. For example, the evaluation value 172 is an index indicating whether traffic congestion has actually been avoided as a result of controlling signals and speed limits for recommendations from a CEP engine derived from a car event.

  As shown in FIG. 7B, the setting table 18 includes an evaluation value generation period (an example of a first period) 181, an evaluation value analysis period (an example of a third period) 182, a threshold value 183, and a validity period. The expiration date 184 is saved. The evaluation value analysis period is set to a period longer than the evaluation value generation period. By making the analysis period of the evaluation value longer than the generation period of the evaluation value, the number of data of the evaluation value to be analyzed can be made equal to or more than a predetermined value, thereby improving the accuracy of the analysis result.

  The first correlation table 19 has, as shown in FIG. 7C, feedback data (evaluation value) 191, event data 192, and a correlation coefficient value 193. The correlation coefficient value 193 is a value indicating a correlation between the feedback data 191 and the event data 192.

  For example, when the event is a passenger car event, data for one day of the evaluation value (feedback value) 172 of the passenger car event stored in the time-series data table 17 shown in FIG. A correlation coefficient value 193 indicating a correlation between the feedback data 191 and the event data 192 is calculated from the hourly average speed data 171 for one day.

  As an example of a method for calculating the correlation coefficient value 193, for example, the following equation (1) is given. When a data string composed of two sets of numerical values (x, y) is given, the correlation coefficient value is obtained by Expression (1). In the calculation of the correlation coefficient value 193, the average speed of the passenger car event every hour is substituted for x, and the evaluation value of the passenger car event every hour is substituted for y. Further, in the calculation of the correlation coefficient value 193, a data string including two sets of numerical values (x, y) is a set of an average speed of a passenger car event every hour and an evaluation value of the passenger car event every hour ( (x, y) shows a data string for one day. The same applies to track events.

  However, the method for calculating the correlation is not limited to this, and any known method for calculating a correlation can be used.

  Returning to FIG. 5, the control unit 13 has an internal memory for storing programs and control data that define various processing procedures, and executes various processing by using these. The control unit 13 includes a reception unit 4, an analysis unit 5, a filter unit 6, and an event processing unit 7.

  The receiving unit 4 receives various data including a plurality of events via a network. The receiving unit 4 receives an evaluation value evaluated by the traffic control server 20 for the recommendation transmitted from the server device 10. The receiving unit 4 is an example of an obtaining unit that obtains an evaluation value of data subjected to event processing based on an event processing rule described later for each first period. In the first period, a predetermined time such as one hour is set. The first period is set in advance by the administrator.

  The analysis unit 5 evaluates the event processing target data of the first event type in the third period longer than the first period in the plurality of event processing target data and the event processing target for each event type. A correlation coefficient value 193 with the data is calculated for each event type. The calculated correlation coefficient value 193 is stored in the first correlation table 19 in FIG. In addition, the analysis unit 5 may include, in a plurality of event processing target data, an event processing target data of a first event type in a second period longer than the first period and a second event processing target data different from the first event type. A correlation coefficient value 293 with the event processing target data of the second event type is calculated. The calculated correlation coefficient value is stored in the second correlation table 219 in FIG. Examples of the event processing target data of the first event type and the event processing target data of the second event type include speed data of a passenger car event and speed data of a truck event.

  The analysis unit 5 generates a filtering rule (filter condition) for filtering each data of a plurality of received events based on the correlation coefficient value 193 or the correlation coefficient value 293.

  The filter unit 6 filters each data of a plurality of events based on a filtering rule at a stage before the CEP engine 15 performs the event processing. The filter unit 6 discards a filter condition that has exceeded the expiration date of the filter condition. The filter unit 6 relates to the event processing target data of the second event type based on the first filter condition for the event processing target data of the first event type specified from the evaluation value and the correlation coefficient value. Set the second filter condition.

  The event processing unit 7 performs event processing (complex event processing: CEP) on the data (event data that has not been filtered) output from the filter unit 6. When the input data matches the event processing rule, the event processing unit 7 makes a recommendation to the traffic control server 20 that “there is a possibility of congestion”. The recommendation “possible congestion” is an example of a recommendation that is output to an output destination as a result of the server device 10 performing an event process on an input event.

  In the CEP system of FIG. 6 described below, the analysis engine 16 performs a third period longer than the first period of the evaluation value for each first period and each data of the plurality of events to be received. Corresponds to the analysis unit 5 that calculates a plurality of correlation coefficient values for each of a plurality of received events based on the correlation in. Further, the analysis engine 16 corresponds to the receiving unit 4 that receives the evaluation value of the event processed data for each first period.

  The filter engine 14 corresponds to the filter unit 6 that filters input data based on a filter condition.

  Further, the CEP engine 30 corresponds to the event processing unit 7 that processes data output from the filter unit 6 (event data that has not been filtered) based on an event processing rule.

[Overview of operation of CEP system including server device]
Next, an outline of the operation of the CEP system including the server device will be described with reference to FIGS. Hereinafter, a car event, a truck event, and a weather forecast event will be described as examples of big data. For example, speed data 22 indicating a speed of a passenger car detected by a speed sensor provided at a predetermined position on a road to be monitored for traffic conditions is distributed to the server device 10 as an example of a passenger car event. Further, as an example of a truck event, speed data 23 indicating a truck speed detected by a speed sensor provided at a predetermined position on a road to be monitored for traffic conditions is also delivered to the server device 10. . Further, as an example of a weather forecast event, the server device 10 is delivered with weather data 24 indicating predicted weather in an area including a road to be monitored, from a server of a provider of a weather forecast service.

(Event processing)
First, the analysis engine 16 receives a plurality of input events including speed data 22 of a car event, speed data 23 of a truck event, and weather data 24 of a weather forecast event. The analysis engine 16 generates time-series data from the input event and adds the data to the time-series data table 17. Thereby, as shown in FIG. 7A, time-series data is collected according to the type of the input event, and the average speed according to the type of the input event is calculated.

  Returning to FIG. 6, the speed data 22, 23 and the weather data 24 have a data structure of XML (Extensible Markup Language) format. The speed data 22 has, as tags, items with data names “eventTypeId” and “speed”. “EventTypeId” is a tag used when embedding the type of event that has occurred. In the speed data 22, “CarEvent” indicating that the passenger car has passed is embedded as a generated event in the tag of “eventTypeId”. “Speed” is a tag used when embedding the detected speed. In the speed data 22, a value indicating an hourly speed ([Km / h]) is embedded as a detected speed of the passenger car in a tag of “speed”. The speed data 22 indicates that the speed of the passenger car is 40 [Km / h].

  The speed data 23 has, as tags, items with data names of “eventTypeId” and “speed”. In the speed data 23, “TrackEvent” indicating that the track has passed is embedded as a generated event in the tag of “eventTypeId”. “Speed” is a tag used when embedding the detected speed. In the speed data 23, a value indicating an hourly speed ([Km / h]) is embedded as a speed of the detected track in a tag of “speed”. The speed data 23 indicates that the track speed is 40 [Km / h].

  Further, the weather data 24 has, as tags, items having data names of “eventTypeId” and “Rain”. “EventTypeId” is a tag used when embedding the type of event that has occurred, as described above. In the weather data 24, “SensorEvent” is embedded as a generated event in a tag of “eventTypeId”. “Rain” is a tag used when rain is predicted. In the weather data 24, when rain is predicted, “true” is embedded in the tag of “Rain”. Here, the weather data 24 indicates that rain is predicted.

  The analysis engine 16 outputs the input event to the filter engine 14. The filter engine 14 receives an input event and filters the input event based on a filtering rule 14a. The filter engine 14 outputs the unfiltered data to the CEP engine 15. At this point, it is assumed that the filtering rule 14a (filter condition) in FIG. 6 has not been automatically generated. The filtering process by the filter engine 14 after the filtering rule 14a is automatically generated will be described later.

  The CEP engine 15 inputs each data of a passenger car event, a truck event, and a weather forecast event, and performs event processing according to an event processing rule 15a. The event processing rule 15a is set for each event. For example, the CEP engine 15 performs event processing on the speed data 22 of the car event based on the event processing rule 15a for the car event. Similarly, the CEP engine 15 performs event processing on the speed data 22 of the track event based on the event processing rule 15a for the track event. Similarly, the CEP engine 15 performs event processing on the speed data 22 of the weather forecast event based on the event processing rule 15a for the weather forecast event. In this manner, the CEP engine 15 can perform evaluation for each event type by processing the event for each event type.

  For example, assume that the event processing rule 15a for a car event is a condition that the average speed of the car event is 50 km / h or less and the weather is rain. When this rule is satisfied, the CEP engine 15 transmits information of “possibility of congestion” to the traffic control server 20 (recommendation).

  Note that the CEP system can set an event to be evaluated and an event to be non-evaluated in advance. The CEP engine 15 excludes non-evaluation-targeted events from the input events from being subjected to event processing. Specifically, a system operation is possible in which a car event is an evaluation target and a weather forecast event is a non-evaluation target. In the present embodiment, the passenger car event, the truck event, and the weather forecast event are all events to be evaluated.

  When the traffic control server 20 receives a recommendation of “possible congestion” from the server device 10, the traffic control server 20 performs control to mitigate the congestion on the assumption that occurrence of congestion is predicted on a road to be monitored for traffic conditions. . For example, the traffic control server 20 performs control to increase the period of the green signal on the road side where traffic congestion is predicted and to control the speed limit for the signal provided on the road to be monitored, so that the traffic can be passed on the road to be monitored. Increase the number of passenger cars. In addition, for example, the traffic control server 20 notifies traffic information in which traffic congestion is predicted to occur on a road to be monitored, prompts a detour to another road, and controls to avoid traffic congestion.

  As described above, in the CEP system, the process in which the server device 10 inputs an input event, executes filter processing and event processing, and transmits a recommendation to the traffic control server 20 based on the processing result has been described.

  Next, the server device 10 receives feedback on the evaluation of “recommendation” from the traffic control server 20, automatically analyzes the correlation between the fed-back evaluation value and the input event, and automatically generates the filtering rule 14a from the analysis result. Will be described. By automatically generating the filtering rule 14a, data requiring event processing and unnecessary data in the CEP engine 15 can be automatically selected, and the processing load of the CEP engine 15 can be reduced.

(feedback)
(1) Feedback of the usefulness of the event processing rules for each event The traffic control server 20 evaluates whether or not the traffic congestion was actually avoided by the control for avoiding the traffic congestion, and transmits the evaluation value to the server device 10. Feedback to For example, the evaluation value data 21 transmitted by the traffic control server 20 shown in FIG. 6 is generated for each event processing rule (for passenger cars, trucks,...). In the present embodiment, the evaluation period (first period) is set to one hour set in advance in the generation period 181 of FIG. 7B, but the evaluation period is not limited to this and a certain time is set. It should be done. For example, in the present embodiment, the traffic control server 20 calculates an hourly evaluation value for a passenger car event and an hourly evaluation value for a truck event.

  The evaluation value data 21 has, as tags, items with data names of “eventTypeId”, “subject”, and “assessment”. “EventTypeId” is a tag used when embedding the type of event that has occurred. In the evaluation value data 21, a “feedback event” for returning an evaluation value every hour to the server device 10 is embedded in a tag of “eventTypeId”. The evaluation value indicates whether traffic congestion has actually been avoided as a result of controlling signals and speed limits for recommendations from the server device 10. “Subject” indicates the original event that generated the “feedback event”. Here, “CarEvent” is embedded in “subject”. An evaluation value (feedback value) for each event is embedded in “assessment”. Here, the evaluation value fed back to the passenger car event is “10”. In the present embodiment, the evaluation value is indicated in 10 levels from 1 to 10. An evaluation value of 10 indicates that the correlation is highest and the recommendation is most useful. When the evaluation value is 1, it shows that the correlation is the lowest and the recommendation is not the most useful.

(2) Generate Time Series Data The analysis engine 16 of the server device 10 receives an evaluation value that is an evaluation result fed back from the traffic control server 20. The analysis engine 16 generates time series data from the evaluation values and adds the time series data to the time series data table 17. Thereby, time-series data of the evaluation value fed back for each event is collected. In FIG. 7A, an hourly evaluation value 172 of a car event and an hourly evaluation value 174 of a truck event are stored.

(3) Automatic Analysis of Correlation Returning to FIG. 6, the analysis engine 16 automatically analyzes the correlation between the collected evaluation values and the input events based on the time-series data table 17, and determines the filtering rule 14a from the analysis result. Generate automatically. The analysis engine 16 calculates the evaluation values of various events for each analysis period (third period) of one day, for example, from the evaluation values of various events stored hourly based on the time-series data table 17. . Note that the third period is not necessarily one day as long as the period is longer than the above-described first period.

(4) Automatic Generation of Filtering Rule The analysis engine 16 compares the calculated correlation coefficient value 193 with a preset threshold value 183 shown in FIG. Is within the range of the threshold 183, the filtering rule 14a is automatically generated. For example, in the example shown in FIG. 7C, the correlation coefficient value 193 is “0.01”, and the threshold value 183 is “± 0.05”. Therefore, the analysis engine 16 determines that the correlation coefficient value 193 is within the range of the threshold value 183, and determines that the speed data of the passenger car event is data unnecessary for avoiding or reducing traffic congestion by traffic control. As a result, the analysis engine 16 automatically generates a filtering rule 14a that defines a filter condition for removing a car event. As a result, it is possible to automatically sort data that requires event processing in the CEP engine 15 from unnecessary data. As a result, the data from which unnecessary data has been removed is input to the CEP engine 15, whereby the processing load of the CEP engine 15 can be reduced.

  Specifically, when it is determined that the speed data of the passenger car event is unnecessary for avoiding or reducing traffic congestion by traffic control, the analysis engine 16 automatically generates the filtering rule 14a shown in FIG. The filtering rule 14a in FIG. 6 indicates that a passenger car event (CarEvent) is set as a target event, and when the speed data of the passenger car event indicates a value smaller than 0, the passenger car event is output. There are no passenger car events with a speed less than zero. Therefore, according to the filtering rule 14 a of FIG. 6, all the passenger car events are filtered and are not input to the CEP engine 15.

  An expiration date can be set in the filtering rule 14a. For example, if “1 year” is set in advance in the expiration date 184 of the setting table 18 shown in FIG. 7B, when one year has passed since the filtering rule 14a was automatically generated, the analysis engine 16 Automatically discards the filtering rule 14a.

  As described above, even when the importance of various event data changes, the filtering rule 14a can be automatically generated or automatically deleted according to the change. This makes it possible to appropriately set filter conditions for filtering unnecessary data from various event data at a stage prior to event processing by the CEP engine 15.

  The automatic generation of the filtering rule 14a has been described above. According to the present embodiment, the analysis engine 16 automatically generates the filtering rule 14a, so that the CEP engine 15 can automatically select data requiring event processing from unnecessary data. As a result, the amount of data input to the CEP engine 15 can be reduced, and the processing load of the CEP engine 15 can be reduced.

  Further, in the present embodiment, it is possible to automatically discard the filtering rule 14a that has exceeded the expiration date of the filter condition. In particular, when the reliability of the evaluation of the traffic control server 20 is high, the expiration date of the filtering rule 14a is set long, and when the evaluation result of the traffic control server 20 is fluid and the reliability of the evaluation is low, It is preferable to set a short expiration date. As a result, even if the optimality of the initially set rules deteriorates with the passage of time, the necessary data is discarded before the event processing, and the unnecessary data is not discarded before the event processing. A reduction in the usefulness and validity of the result can be avoided.

(Event trigger)
The event may be triggered when a vehicle enters the monitored road or when a certain amount of speed change occurs in each vehicle. The event occurrence trigger depends on the generation timing of a detection value output from an event generation source such as a sensor device. The input event of each vehicle according to the present embodiment is assumed to be a large amount of events such as 1,000,000 TPS (Transactions Per Second), but the number of events is not limited to this.

(Time series analysis opportunity)
The generation period 181 and the analysis period 182 of the time-series data depend on the operation status of the system. In the present embodiment, the generation period 181 is set to one hour and the analysis period 182 is set to one day, the average speed is calculated every hour, and the correlation analysis is performed every day. However, this unit is not necessarily required. . For example, correlation analysis can be performed weekly using the average of each week in each time zone as time-series data. The generation period 181 can be set in the setting table 18 in advance by an administrator such as an SE. However, the generation period is preferably set to be the same as the feedback interval for the entire CEP system.

[Event processing]
Next, a flow of event processing performed by the server device 10 according to the present embodiment will be described.
(Process of analysis engine)
When the event processing in FIG. 8 is started, first, in step S10, the analysis engine 16 (the receiving unit 4) inputs an input event. Next, in step S12, the analysis engine 16 (analysis unit 5) generates time-series data from the input event and adds it to the time-series data table 17.
(Processing of filter engine)
Next, in step S14, the filter engine 14 (the filter unit 6) determines whether the filtering rule 14a has exceeded the expiration date 184 based on the expiration date 184 of the setting table 18. When determining that the filtering rule 14a has exceeded the expiration date 184, the filter engine 14 discards the filtering rule 14a in step S16, and proceeds to step S18. In step S14, when the filter engine 14 determines that the filtering rule 14a has not exceeded the expiration date 184, the process proceeds to step S18.

Next, in step S18, the filter engine 14 determines whether the input event matches the filtering rule 14a. If the filter engine 14 determines that the input event matches the filtering rule 14a, the input engine is discarded in step S20, and the process proceeds to step S22. If the filter engine 14 determines that the input event does not match the filtering rule 14a, the process proceeds to step S22. In the example of the filtering rule 14a in FIG. 6, a passenger car event is filtered.
(CEP engine processing)
Next, in step S22, the CEP engine 15 (event processing unit 7) determines whether the input event matches the event processing rule 15a. If the CEP engine 15 determines that the input event matches the event processing rule 15a, it sends a recommendation to the traffic control server 20 in step S24, and ends this processing. When the CEP engine 15 determines that the input event does not match the event processing rule 15a, the CEP engine 15 ends this processing without transmitting a recommendation to the traffic control server 20.

[Automatic generation process]
Next, the flow of the automatic generation process performed by the server device 10 according to the present embodiment will be described with reference to FIG. The automatic generation process performed by the server device 10 of FIG. 9 is executed independently of the event process performed by the server device 10 of FIG. The event processing of FIG. 8 is executed every time an event occurs. On the other hand, in the automatic generation processing of FIG. 9, in a state where the evaluation values transmitted from the traffic control server 20 are accumulated for a predetermined time, the automatic generation of the filtering rule is performed in the correlation analysis processing (S38, S40). .
(Process of analysis engine)
When the automatic generation process is started, in step S30, the analysis engine 16 determines whether there has been feedback (reception of an evaluation value) from the traffic control server 20, which is the destination of the recommendation. If the analysis engine 16 determines that there has been feedback from the traffic control server 20, in step S32, it generates time-series data from the feedback evaluation values, stores it in the time-series data table 17, and proceeds to step S34. In step S30, when the analysis engine 16 determines that there is no feedback from the traffic control server 20, the process proceeds to step S34.

  Next, in step S34, the analysis engine 16 determines whether the time-series data stored in the time-series data table 17 has exceeded a certain time (a third period). When the analysis engine 16 determines that the time-series data does not exceed, for example, one day as a certain time, the analysis engine 16 ends the processing. When the analysis engine 16 determines that the time-series data has exceeded the predetermined time, the process proceeds to step S36, and determines whether the input event is a new event. When the analysis engine 16 determines that the input event is not a new event, the analysis engine 16 proceeds to a correlation analysis process (evaluation value) in step S38. On the other hand, when the analysis engine 16 determines that the input event is a new event, the analysis engine 16 proceeds to a correlation analysis process (new event) in step S40. Before the correlation analysis processing (new event) in step S40 is first performed, the correlation analysis processing (evaluation value) in step S38 is performed at least once.

  FIG. 10 shows a flowchart of the correlation analysis processing (evaluation value) called in step S38. FIG. 13 shows a flowchart of the correlation analysis process (new event) called in step S40. Hereinafter, the correlation analysis process (evaluation value) of FIG. 10 will be described first, and then the correlation analysis process (new event) of FIG. 13 will be described.

[Correlation analysis processing (evaluation value)]
When the correlation analysis process (evaluation value) in FIG. 10 is started, in step S48, the analysis engine 16 determines the phase of the evaluation value and the input event fed back from the traffic control server 20 based on the above equation (1). Calculate the relationship value. Next, in step S50, the analysis engine 16 determines whether the calculated correlation coefficient value 193 exceeds the range of the threshold 183. If the analysis engine 16 determines that the correlation coefficient value 193 exceeds the range of the threshold 183, the correlation between the evaluation value and the input event is high, and thus the analysis engine 16 determines that the data is necessary for event processing by the CEP engine 15. Then, the processing ends without generating the filtering rule 14a.

  In step S50, when the analysis engine 16 determines that the correlation coefficient value 193 is within the range of the threshold 183, in step S52, the analysis engine 16 determines that the correlation between the evaluation value and the input event is low, and automatically sets the filtering rule 14a. Then, the process ends. As a result, unnecessary event data can be removed from the events input to the CEP engine 15. As a result, the CEP engine 15 can process useful event data for avoiding traffic congestion, thereby transmitting useful recommendations to the traffic control server 20 and reducing the processing load of the CEP engine 15.

(Example of automatic generation of filtering rules)
An example of automatic generation of a filtering rule will be described with reference to FIGS. FIG. 11 illustrates a method for automatically generating a filtering rule according to an embodiment. For example, the administrator sets “all filter conditions” for each event type in the setting table 18 in advance. In the example shown in FIG. 11, an event name 185a and all filter conditions 185b are set in the setting table 18.

  In the automatic generation of the filtering rule, the analysis engine 16 sets a template from the setting table 18 and embeds the event name and the filter condition of the target event to be filtered. In FIG. 11, the analysis engine 16 embeds a car event and a filter condition in a template. As a result, the filtering rule can be automatically generated, so that the administrator does not need to manually generate the filtering rule.

(Set expiration date for each filtering rule)
Further, in the present embodiment, the “expiration date” can be set in the filtering rule 14a automatically generated as described above. For example, as shown in FIG. 12, the filtering target of the filtering rule 14a is a car event A. The filtering target of the filtering rule 14b is the passenger car event B.

  As shown in (1), (3), and (5) of FIG. 12, the filtering rule 14a is used to exclude a car event based on the correlation between the evaluation value fed back from the traffic control server 20a and the input event. Has been automatically generated.

  As shown in (2), (4), and (6) of FIG. 12, the filtering rule 14b is used to exclude a passenger car event based on the correlation between the evaluation value fed back from the traffic control server 20b and the input event. Has been automatically generated.

  The traffic control server 20a controls an existing road, and the feedback evaluation value has high reliability. On the other hand, the traffic control server 20b controls a recently formed road, and the reliability of the feedback evaluation value is low. Therefore, the validity period of the filtering rule 14a is set to a long period of, for example, one year, whereas the validity period of the filtering rule 14b is set to a short period of, for example, one week. In this manner, in this embodiment, the expiration date of the filtering rules 14a and 14b can be set for each traffic server. As a result, even when the evaluation value by the traffic server is fluid and the reliability is low, appropriate data selection can be performed while the operation of the CEP system is continued.

[Correlation analysis processing (new event)]
Next, a new event and an existing event when an event of a different type (hereinafter, also referred to as a “new event”) from an event of a type already processed (hereinafter, also referred to as an “existing event”) are input. Will be described with reference to FIGS. 13 and 14. FIG. FIG. 13 shows a flowchart of a correlation analysis process between a new event and an existing event. FIG. 14 shows, as an example of the automatic analysis of the correlation according to the embodiment, a new type of speed data 22 of a car event, speed data 23 of a truck event, and weather data 24 of a sensor event, which are examples of existing events. This shows a case where speed data 25 of a motorcycle event, which is an example of the event, is input.

  When the correlation analysis process (new event) in FIG. 13 is started, in step S58, the analysis engine 16 calculates a correlation coefficient value 293 between the existing event and the new event based on the above equation (1). 2 Stored in the correlation table 219. Next, in step S60, the analysis engine 16 determines, based on the second correlation table 219, whether the correlation coefficient value 293 exceeds a range of a preset threshold value 283. The threshold value 283 may be set to the same value as the threshold value 183, or may be set to a different value.

  The analysis engine 16 calculates a correlation coefficient value between the existing event and the new event for each type of the existing event, and determines the correlation between the existing event and the new event. As a result, for example, for each of the passenger car event and the truck event, the correlation with the new event is determined.

  The analysis engine 16 calculates a correlation coefficient value 293 between the new event data 194 and the existing event data 195 shown in FIG. 15B based on the time series data table 17 shown in FIG.

  For example, when the new event is a motorcycle event, one-hour average data of hourly speed data 171 of the passenger car event stored in the time-series data table 17 shown in FIG. A correlation coefficient value 293 between the motorcycle event data 194 and the car event data 195 is calculated from the data for one day of the average speed 175 for each day.

  As an example of a method of calculating the correlation coefficient value 293, the correlation coefficient value 293 can be calculated by the above equation (1). That is, when a data string including two sets of numerical values (x, y) is given, the correlation coefficient value is obtained by Expression (1). Here, x is the average speed of each type of event every hour, and y is the average speed of each motorcycle event every hour. Therefore, a data string composed of numerical values (x, y) indicates a data string for one day (average hourly speed of each type of event, average hourly rate of motorcycle event). The same applies to track events. However, the method for calculating the correlation is not limited to this, and any known method for calculating a correlation can be used.

  The analysis engine 16 automatically generates a filtering rule 14a from the analysis result. The analysis engine 16 evaluates each event of each type for each analysis period (an example of a second period) based on the evaluation value for each event stored hourly based on the time-series data table 17. The value is calculated and used for correlation analysis between existing events of each type and new events. In the present embodiment, the analysis period (an example of the second period) does not necessarily have to be one day as long as the period is longer than the first period. The second period and the third period may be the same period or different periods.

  Returning to FIG. 13, in step S60, when the analysis engine 16 determines that the correlation coefficient value 293 is within the range of the threshold value 283, the analysis engine 16 determines that the correlation between the new event and the existing event is low, and ends this processing. I do. When a high correlation is not recognized between the new event (bike event) and any existing event, the filtering rule and the event processing rule for the new event are not automatically generated.

  On the other hand, in step S60, when the analysis engine 16 determines that the correlation coefficient value 293 exceeds the range of the threshold 283, the analysis engine 16 determines that the correlation between the new event and the existing event is high, and proceeds to step S62. In step S62, the analysis engine 16 determines whether there is a filtering rule for an existing event having the highest correlation.

  In step S62, when the analysis engine 16 determines that there is a filtering rule for an existing event having the highest correlation, in step S64, the analysis engine 16 automatically generates a filtering rule for a new event in accordance with the filtering rule for the existing event having the highest correlation. When there is one correlation coefficient value 293 exceeding the threshold 283 range, a new event filtering rule is automatically generated in accordance with the existing event filtering rule corresponding to the correlation coefficient value 293 exceeding the threshold 283 range.

  For example, in the second correlation table 219 shown in FIG. 15B, it can be seen from the correlation coefficient value 293 that the new motorcycle event has the highest correlation with the car event. Thus, for example, when a strong correlation such that the absolute value of the correlation coefficient value 293 exceeds 0.95 is recognized, the speed of the motorcycle increases when the speed of the passenger car increases, and the speed of the motorcycle decreases when the speed of the passenger car decreases. Is also expected to fall. In this case, it can be expected that a motorcycle event having a high correlation with a passenger car event having a low correlation with the traffic congestion has no relation or a low relation with the traffic congestion similarly to the passenger car event.

  The above prediction will be specifically described. The traffic control server 20 calculates, for each event, an evaluation value indicating whether or not the traffic congestion was actually avoided by the control for avoiding the traffic congestion according to the recommendation from the server device 10, and transmits the evaluation value to the server device 10. For example, when the evaluation value transmitted from the traffic control server 20 to the server device 10 is an evaluation value of a car event, the server device 10 calculates the evaluation value of the car event, the speed data of the car event, and 193 is used to calculate the correlation coefficient value 193 of the passenger car event. When the correlation coefficient value 193 of the car event is equal to or smaller than the threshold 183, the server device 10 determines that the correlation between the evaluation value of the car event and the speed data of the car event is low. As a result, the server device 10 automatically generates a filtering rule for a car event, as shown in step S52 of FIG. Thereby, the speed data of the car event is discarded based on the filtering rule of the car event. In this manner, unnecessary events can be discarded based on the filtering rule specified by the evaluation value for each event type.

  The correlation coefficient value 293 indicates a correlation between a new event and an existing event. The server device 10 calculates a correlation coefficient value 293 between the new event and the existing event based on the equation (1). If it is determined that the new event and the existing event have a high correlation based on the correlation coefficient value 293, a filtering rule for the new event is automatically generated using the filtering rule for the existing event that has been evaluated based on the calculated evaluation value. I do. For example, when the new event is a motorcycle event and the existing event is a passenger car event, and the correlation coefficient value 293 between the motorcycle event and the passenger car event exceeds the threshold value 283, the server device 10 determines the correlation between the motorcycle event and the passenger car event. Is determined to be high. In this case, a motorcycle event having a high correlation with a passenger car event having a low correlation with the traffic congestion can be expected to have no relation or a low relation with the traffic congestion similarly to the passenger car event. As a result, as shown in step S64 in FIG. 13, the server device 10 automatically generates a filtering rule for a motorcycle event, which is a new event, using a filtering rule for a car event, which is an existing event. In the example shown in FIG. 14, the analysis engine 16 automatically generates a filtering rule 14c for filtering a motorcycle event using the filtering rule 14a for a car event. In this way, based on the correlation coefficient value 293 between the existing event and the new event, the filtering rule of the existing event that has been evaluated based on the calculated evaluation value is used for the new event determined to be highly correlated with the existing event. Then, a new event filtering rule can be automatically generated. As a result, it is possible to appropriately set a filter condition for filtering a plurality of received events before performing event processing based on a predetermined rule.

  Note that an existing event is an example of event processing target data of a first event type, and a new event is an example of event processing target data of a second event type. The existing event filtering rule is an example of a first filter condition for a first event type. The new event filtering rule is an example of a second filter condition for the second event type. The generated first filter condition and second filter condition indicate filter conditions for filtering a plurality of event processing target data received by the server apparatus 10 before inputting the data to the event processing in FIG.

  In the above description, the second filter condition for the new event is automatically generated based on the correlation coefficient value 293 between the new event and the existing event and the first filter condition for the existing event specified from the evaluation value of the existing event. However, it is not limited to this. For example, a correlation coefficient value 293 between the first existing event and a second existing event having a different type from the first existing event, and a first existing event related to the first existing event specified from the evaluation value of the first existing event. The second filter condition for the second existing event may be automatically generated based on the above filter condition. For example, based on a first filter condition of a car event as an example of an existing event and a correlation coefficient value 293 of a truck event and a car event as another example of an existing event, a second filter condition for a truck event is automatically set. May be generated. According to this, it is possible to appropriately set a rule for discarding an input event including an existing event and a new event that changes according to a time zone or a road condition.

  In the example illustrated in FIG. 14, the analysis engine 16 automatically generates a motorcycle filtering rule 14c having the same filter condition as the passenger car filtering rule 14a.

  As a result, unnecessary event data can be removed from the events input to the CEP engine 15. As a result, the CEP engine 15 can process useful event data for avoiding traffic congestion, thereby transmitting useful recommendations to the traffic control server 20 and reducing the processing load of the CEP engine 15. In particular, according to the present embodiment, even when the data amount of an input event increases rapidly due to the occurrence of a new event, an increase in the load on the CEP system can be suppressed.

  Returning to FIG. 13, next, in step S66, the analysis engine 16 automatically generates the event processing rule 15a, and ends this processing.

  The automatically generated filtering rule has an expiration date, and the filtering rule after the expiration date is automatically deleted. Therefore, when the filtering rule is automatically discarded, a new event that is no longer filtered is input to the CEP engine 15. Therefore, in order to process a new event that is no longer filtered, an event processing rule 15a is automatically generated in advance in step S66. Thereby, the CEP engine 15 can process the new event input to the CEP engine 15 as an event to be recommended for the traffic control server 20.

  For example, in FIG. 16, the analysis engine 16 predicts that a motorcycle event that has a high correlation with a truck event related to traffic congestion will probably be related to the traffic condition, and the event processing rule 15c for the motorcycle event. Is automatically generated. The analysis engine 16 automatically generates a motorcycle event processing rule having the same condition as the track event event processing rule. Thus, the motorcycle event can be processed by the CEP engine 15.

  If a new event (here, a bike event) does not have a high correlation with any existing event, an event processing rule for the bike event is not automatically generated. In this case, it is preferable that the administrator manually generates the event processing rule 15a for the new event as needed.

  In step S62, when the analysis engine 16 determines that there is no filtering rule for the existing event having the highest correlation, the process proceeds to step S66, in which the event processing rule 15c for a new event is automatically generated, and the process ends. .

  According to this, even if a new event is input to the analysis engine 16 during operation and it takes time until the next evaluation value for feedback is obtained, an automatically generated event for the new event is used. Based on the processing rule 15c, the correlation between the existing event and the new event can be automatically analyzed. As a result, the analysis engine 16 can automatically generate the filtering rule 14c, and can automatically select required data and unnecessary data.

  After the automatic generation of the filtering rule 14c shown in (3) of FIG. 14 and the automatic generation of the event processing rule 15c shown in (4) of FIG. 14, as shown in (5) of FIG. Alternatively, an email may be sent to the administrator to notify the automatic generation of each rule. When the automatic generation of each rule is not performed, or when the automatic generation of each rule fails, the administrator may be notified by e-mail.

  It is considered that it takes time until a reliable evaluation value is obtained, especially because a new event is added and the number of new events is small. In such a case, when the traffic control server 20 feeds back the evaluation value to the server device 10 every day, there is a concern that the load of the CEP system is unnecessarily increased. Therefore, the traffic control server 20 may reduce the transmission frequency according to the situation, for example, setting the first Monday every month to a day when the evaluation value is transmitted to the server device 10 so as not to increase the load on the system. .

(Example of automatic generation of event processing rules)
An example of automatic generation of an event processing rule will be described with reference to FIG. FIG. 18 illustrates a method for automatically generating an event processing rule according to an embodiment. For example, in the automatic generation of the event processing rule of the new event, the analysis engine 16 copies the event processing rule of the existing event for the event processing rule of the new event, and renames the event name of the target event. In FIG. 18, the analysis engine 16 copies the event processing rule 15a for the track event for the event processing rule 15c for the bike event, and changes the event name of the target event from the track event to the bike event. A rule parsing tool is used to automatically generate event processing rules. As a result, the event processing rule 15c for a new event can be automatically generated, and the trouble of manually generating the event processing rule can be omitted.

[Example of effect]
For example, if the administrator manually performs filter conditions for extracting necessary data and unnecessary data in order to narrow down data to be input to the CEP engine, the operation cost of the system increases. The setting of the filter condition depends on the development skill of the administrator. For this reason, appropriate data may not be selected as data to be input to the CEP engine. In addition, data required by the CEP engine and unnecessary data may change during operation. In this case, if the filter condition is fixed, it is not possible to follow a change in the situation during operation, and it becomes impossible to input data corresponding to the change in the situation to the CEP engine. As a result, even if the data becomes unnecessary, it may be input to the CEP engine, and a situation may occur in which the system load is not reduced. In addition, when new data increases rapidly during operation, if it takes time to select new data, the system load increases, and the processing performance of the entire system decreases.

On the other hand, the CEP system according to the present embodiment has the following functions (1) to (3) to solve the above problem, reduce the load on the system, and deal with the rapidly increasing big data. Complex event processing can be performed at low cost.
(1) Function of automatic analysis and selection of necessary and unnecessary data (2) Function of automatic generation of filtering rules for selected unnecessary data (3) Function of automatic analysis and automatic selection when a new event occurs In this embodiment, the analysis of the correlation between the evaluation value fed back from the destination (for example, the traffic control server 20) using the recommendation and the time series data of the input event is automatically performed in the present embodiment. Can be done Thereby, important events and non-important events can be automatically sorted out, and non-important events can be appropriately filtered. Further, in the present embodiment, the filter condition can be dynamically changed based on the evaluation value fed back from the traffic control server 20. For this reason, it is possible to follow the situation change during operation, remove unnecessary data in response to the situation change by filtering, and input necessary data to the CEP engine. As a result, the load on the CEP system can be reduced.

  Further, in the present embodiment, the analysis of the correlation between the new event and the existing event can be automatically performed. As a result, when a high correlation is found between the new event and the existing event, unnecessary new events can be automatically eliminated by applying the same conditions as the filtering rules of the existing event to the new event. In addition, by applying the same conditions as the event processing rules of the existing event to the new event, it is possible to provide the traffic control server 20 with useful recommendations regarding the new event.

(Example of hardware configuration)
Finally, an example of a hardware configuration of the server device 10 according to the present embodiment will be described with reference to FIG. FIG. 19 illustrates an example of a hardware configuration of the server device 10 according to the present embodiment.

  As shown in FIG. 19, the server device 10 includes an input device 101, a display device 102, an external I / F 103, a RAM (Random Access Memory) 104, a ROM (Read Only Memory) 105, a CPU (Central Processing Unit) 106, An I / F 107 and an HDD (Hard Disk Drive) 108 are provided. Each part is mutually connected by a bus B.

  The input device 101 includes a keyboard, a mouse, and the like, and is used to input each operation signal to the server device 10. The display device 102 includes a display and displays various processing results.

  The communication I / F 107 is an interface for connecting the server device 10 to a network. Thus, the server device 10 can perform data communication with the traffic control server 20 via the communication I / F 107.

  The HDD 108 is a nonvolatile storage device that stores programs and data. The stored programs and data include basic software and application software for controlling the entire apparatus. For example, the HDD 108 stores various DB information, programs, and the like.

  The external I / F 103 is an interface with an external device. The external device includes a recording medium 103a and the like. Thereby, the server device 10 can read and / or write on the recording medium 103a via the external I / F 103. The recording medium 103a includes a CD (Compact Disk), a DVD (Digital Versatile Disk), an SD memory card (SD Memory card), a USB memory (Universal Serial Bus memory), and the like.

  The ROM 105 is a nonvolatile semiconductor memory (storage device) that can retain internal data even when the power is turned off. The ROM 105 stores programs and data such as network settings. The RAM 104 is a volatile semiconductor memory (storage device) that temporarily stores programs and data. The CPU 106 is an arithmetic unit that reads out programs and data from the storage device (for example, the “HDD 108” or the “ROM 105”) onto the RAM 104 and executes processing to realize control of the entire apparatus and implement functions.

  With the above hardware configuration, the CEP system according to the present embodiment can perform event processing, automatic generation processing of filtering rules, automatic generation processing of event processing rules, and the like. For example, the CPU 106 executes event processing, automatic generation processing of a filtering rule, automatic generation processing of an event processing rule, and the like using data and programs stored in the ROM 105 and the HDD 108. The time-series data table 17, the setting table 18, the first correlation table 19, and the second correlation table 219 are stored in the RAM 104 or the HDD 108, or stored in a cloud server connected to the server device 10 via a network. Can be stored.

  The data processing program, the data processing method, and the data processing device have been described in the above embodiments. However, the data processing program, the data processing method, and the data processing device according to the present invention are not limited to the above embodiments, and various modifications and improvements can be made within the scope of the present invention. In addition, matters described in the above embodiments can be combined within a range that does not contradict. Further, the data processing program, the data processing method, and the data processing device may be configured by hardware, may be configured by software, or may be configured by combining hardware and software.

  For example, the configuration of the CEP system according to the above-described embodiment is an example, and does not limit the scope of the present invention. Needless to say, there are various system configuration examples according to applications and purposes. For example, a system configuration in which the server device 10 and the traffic control server 20 are connected to each other via a network is one mode of the CEP system according to the present invention, and is not limited to this. For example, the number of server devices 10 included in the CEP system according to the present invention may be one or two or more. When a plurality of server apparatuses 10 are installed, the event processing, the automatic generation processing of the filtering rules, the automatic generation processing of the event processing rules, and the like can be performed by the plurality of server apparatuses 10 in a distributed manner. The processing functions may be selectively integrated in one of the plurality of server devices 10 according to the purpose or purpose. In the CEP system according to the above embodiment, the case where the CEP engine 15 operates on the server device 10 has been described as an example. However, one or a plurality of VMs (Virtual Machines) are The CEP engine 15 may operate on the above.

  Further, the form of the service provided by the server device 10 is not limited to the recommendation for avoiding traffic congestion. As another service form performed by the server device 10, for example, sensor information generated from various sensors (for example, a sensor for detecting rain) placed in a home or information generated from a home appliance is input as an event. Processing. As a result, the server device 10 can provide a service for displaying a recommendation according to the event processing result on a display of each home. An example of a recommendation is "It's raining nearby. I recommend drying my room." "The weather is bad, so I recommend drying my room before going out." You can hang it before you go out. "

With respect to the above description, the following items are further disclosed.
(Appendix 1)
A data processing program for causing a computer to execute a process of sequentially performing an event process on received event process target data according to a predetermined rule and sequentially outputting data after the event process,
Acquiring an evaluation value of the data after the event processing for each first period of the first event type among the plurality of event processing target data;
In the plurality of event processing target data, in a second period longer than the first period, an event processing target data of the first event type and a second event type different from the first event type Calculate the correlation coefficient value with the event processing target data of
A first filter condition for the event processing target data of the first event type, which is specified from the evaluation value, and a second filter condition for the event processing target data of the second event type based on the correlation coefficient value. Set filter conditions,
Data processing program.
(Appendix 2)
The first filter condition and the second filter condition are filter conditions for filtering data of the plurality of received events before inputting the data to the event processing.
The data processing program according to supplementary note 1.
(Appendix 3)
Discarding the first filter condition or the second filter condition that has exceeded the expiration date of the filter condition;
The data processing program according to Supplementary Note 1 or 2.
(Appendix 4)
A data processing method for causing a computer to perform a process of sequentially receiving received event processing data according to a predetermined rule and sequentially outputting data after the event processing,
Acquiring an evaluation value of the data after the event processing for each first period of the first event type among the plurality of event processing target data;
In the plurality of event processing target data, in a second period longer than the first period, an event processing target data of the first event type and a second event type different from the first event type Calculate the correlation coefficient value with the event processing target data of
A first filter condition for the event processing target data of the first event type, which is specified from the evaluation value, and a second filter condition for the event processing target data of the second event type based on the correlation coefficient value. Set filter conditions,
Data processing method.
(Appendix 5)
The first filter condition and the second filter condition are filter conditions for filtering data of the plurality of received events before inputting the data to the event processing.
The data processing method according to supplementary note 4.
(Appendix 6)
Discarding the first filter condition or the second filter condition exceeding the expiration date of the filter condition,
6. The data processing method according to supplementary note 4 or 5.
(Appendix 7)
A data processing device that performs event processing on received event processing target data according to a predetermined rule, and sequentially outputs data after the event processing,
An acquisition unit configured to acquire an evaluation value of data after the event processing for each first period of a first event type among the plurality of event processing target data;
In the plurality of event processing target data, in a second period longer than the first period, event processing target data of the first event type and a second event type different from the first event type An analysis unit that calculates a correlation coefficient value with the event processing target data of
A first filter condition for the event processing target data of the first event type, which is specified from the evaluation value, and a second filter condition for the event processing target data of the second event type based on the correlation coefficient value. A filter section for setting filter conditions,
A data processing device comprising:
(Appendix 8)
The first filter condition and the second filter condition are filter conditions for filtering data of the plurality of received events before inputting the data to the event processing.
The data processing device according to attachment 7.
(Appendix 9)
Discarding the first filter condition or the second filter condition exceeding the expiration date of the filter condition,
The data processing device according to attachment 7 or 8.

1: image forming apparatus 4: receiving section 5: analyzing section 6: filtering section 7: event processing section 11: communication control I / F section 12: storage section 13: control section 14: filter engine 14a, 14b, 14c: filtering rule 15: CEP engine 15a, 15c: Event processing rule 16: Analysis engine 17: Time series data table 18: Setting table 19: First correlation table 20: Traffic control server 22: Speed data (passenger car event)
23: Speed data (track event)
24: Weather data 219: Second correlation table

Claims (5)

A data processing program for causing a computer to execute a process of sequentially performing an event process on received event process target data according to a predetermined rule and sequentially outputting data after the event process,
A plurality of types of the of the event processing data, and obtains the evaluation value related data after the event processing for each first period type of the first event,
In a plurality of types of said event processing data, different for the type of the first event and a second event, the long second period than the first period, the type of the first event calculated and event processing data, the correlation coefficient value between the second event type event processing target data,
The specified from the evaluation value, the first Event Type Event processing target data for the first filter condition, and, based on the correlation coefficient value, the related event processing target data type of the second event Set the filter condition of 2,
Data processing program. The first filter condition and the second filter condition are filter conditions for filtering data of the plurality of received events before inputting the data to the event processing.
A data processing program according to claim 1. Discarding the first filter condition or the second filter condition that has exceeded the expiration date of the filter condition;
A data processing program according to claim 1. A data processing method in which a computer performs a process of performing event processing on received event processing target data according to a predetermined rule and sequentially outputting data after the event processing,
A plurality of types of the of the event processing data, and obtains the evaluation value related data after the event processing for each first period type of the first event,
In a plurality of types of said event processing data, different for the type of the first event and a second event, the long second period than the first period, the type of the first event calculated and event processing data, the correlation coefficient value between the second event type event processing target data,
The specified from the evaluation value, the first Event Type Event processing target data for the first filter condition, and, based on the correlation coefficient value, the related event processing target data type of the second event Set the filter condition of 2,
Data processing method. A data processing device that performs event processing on received event processing target data according to a predetermined rule, and sequentially outputs data after the event processing,
A plurality of types of the of the event processing data, an acquisition unit that acquires an evaluation value on the data after the event processing for each first period type of the first event,
In a plurality of types of said event processing data, different for the type of the first event and a second event, the long second period than the first period, the type of the first event and event processing data, and analyzing unit for calculating a correlation coefficient value between the second event type event processing target data,
The specified from the evaluation value, the first Event Type Event processing target data for the first filter condition, and, based on the correlation coefficient value, the related event processing target data type of the second event A filter unit for setting a filter condition of 2,
A data processing device comprising:

Images