CN106446941A - Unconventional emergency dynamic priority method based on model matching - Google Patents

Abstract

The invention discloses an unconventional emergency dynamic priority method based on model matching, and the method comprises the steps: carrying out the matching of real-time processing data flows through a set which comprises various key information rules of key information model identification key information, and filtering the key information meeting the rules through matching; assigning different values to different key information because different types of key information are different in importance in the system, wherein the size of the assigned values represents the importance of the key information in the system; calculating the frequencies of the key information, and the number of occurrence times of the key information in a time unit, i.e., the occurrence frequency of each type of key information when the key information rules are in a key information time window; calculating the priority of the key information, the arithmetic mean value of the frequencies of the key information and the arithmetic mean value of weight products, wherein the frequencies and weights of the key information decide the importance of the key information in the system.

Description

Unconventional accident dynamic priority method based on Model Matching

Technical field

The present invention relates to a kind of unconventional accident dynamic priority method based on Model Matching, belong to data processing skill Art field.

Background technology

Natural Science Fund In The Light committee of China started in 2008 and implements " unconventional sudden incidents report research " weight Since big project, numerous scholars is studied to unconventional accident with scientific worker.

Unconventional accident identification and the important step, scholar and the technical work that are the discovery that unconventional accident research Person passes through continuous research, obtains many progress in terms of unconventional accident identification with discovery.

By the effort of numerous scholars and scientific worker, in terms of unconventional accident research, achieve many Achievement.In existing achievement in research, unconventional accident recognition methodss have a common trait with the research of model aspect, The focus point being exactly these achievements in research is algorithm and model itself, how existing achievement in research is efficiently applied to believe online Breath is processed, and corresponding research has also lacked.

Content of the invention

Goal of the invention：For problems of the prior art, the present invention calls pass the data of unconventional accident Key data, critical data correspond to key message, emphasis process around unconventional accident online information in data flow knowledge Not, the frequency that data weighting, data occur and priority carry out analysis, research, and proposition is a kind of based on Model Matching very Rule accident dynamic priority method, using method proposed by the present invention, is that unconventional accident is applied at online information Reason provides a kind of alternative thinking and scheme.

Technical scheme：A kind of unconventional accident dynamic priority method based on Model Matching, knows including critical data Not, the process that key message weight calculation, the calculating of key message frequency and key message priority calculate.

Critical data identification process

Key message model KMODEL：The set of the various key message rules of mark key message, to real-time processing number Mated according to stream, filtered out the key message meeting rule by coupling.

KMODEL=M (KRULE)=x | x ∈ KRULE } (4)

Real-time stream is mated by key message model, is calculated and identification, mistake by key message logic rule Leach the data acquisition system meeting key message rule.

Key message model is a dynamic model, and in different system, the recognition rule of key message is different, passes through Model configuration file record, description key message rule, build key message model.

Before identification critical data, first key message model configuration file is loaded and parsed, obtain and close Key information model data MS, key message model data is a key message regular collection, a key message rule parsing Result corresponds to a key message mark.

Key message weight

Key message weight：Different types of key message, significance level in systems is different, to different crucial letters Breath gives different numerical value, represents key message significance level in systems by the size giving numerical value.

Initial data is mated through key message model, identifies key message, and key message is mapped to be had The interval of different weights.

Key message weight dynamic change, in different system, different business, the weighted value of key message is different, leads to Cross the relation of key message weight configuration file record, description key message data weighting and key message.

Before identification critical data, load weight configuration file, weight configuration file is parsed, calculate weight mark Will, obtains weight mark set WS, and WS is put into internal memory.Implement step as follows：

1) load weight configuration file；

2) parsing is entered to weight configuration file；

3) calculate weight mark；

4) weight mark puts into internal memory list.

A kind of weight type corresponds to a weight mark, represents different weight marks by different numerical value (weighted value), leads to Cross the weight type that different weighted values represent different key messages, the relationship of the two corresponds.

Weight mark configuration file and key message configuration file are associated by key message rule flag.

In conjunction with MS and WS, the data flow after original data stream, process is identified, recognition result passes through 1,0 two-value table Show, 1 expression identification data belongs to key message, 0 represents that identification data is non-critical information, if data meets key message rule Then, then the data being identified is key message, is not otherwise, algorithm such as (5)：

Wherein：

KS_nIt is that key message rule starts in current time in recognition cycle, every class key message that system identification goes out Number of times set, DATA is identified data, and MS is model data.

Key message frequency

Key message frequency：The number of times occurring in the key message unit interval.I.e. key message rule is in key message Between in window, every class key message frequency of occurrences.Frequency formula such as (6):

Pre=FFrequency (KS_n, t)=KS_n/t (6)

Wherein：

KS_nThe set of the number of times that every class key message identifies, t be key message rule recognition cycle start to The time span of current time.

Key message priority

Key message priority：Arithmetic average, key message frequency and weight that key message frequency is amassed with weight, certainly Determine key message significance level in systems.

In a time slice, change over time, the continual analysis of mass data, the priority of key message is not Break and change, Mobile state adjustment is entered by the key message priority that visualization is shown, real-time, preferential display priority is high Data, by being averaged with the product of weight to the frequency of key message, average is big, and priority is high, and priority data calculates It is a lasting process, priority dynamically adjusts formula such as (7)：

Wherein：

Pre is the frequency of key message, is weight initial value, and WS is weight mark.

In real-time analyzer, in real time, continually enter system, Pri value is big, then key message is excellent for mass data First level is higher, and data is more crucial.

Brief description

Fig. 1 is key message rule and key message relationship model figure；

Fig. 2 is model, key message is regular and key message rule flag；

Fig. 3 maps for key message to weight；

Fig. 4 identifies internal memory logic chart for key message；

Fig. 5 is clock memory graph of a relation；

Fig. 6 is key message priority forming process relationship memory figure；

Fig. 7 is threshold method result figure；

Fig. 8 is key message frequency；

Fig. 9 is key message priority.

Specific embodiment

With reference to specific embodiment, it is further elucidated with the present invention it should be understood that these embodiments are merely to illustrate the present invention Rather than restriction the scope of the present invention, after having read the present invention, the various equivalences to the present invention for the those skilled in the art The modification of form all falls within the application claims limited range.

A kind of unconventional accident dynamic priority method based on Model Matching, abbreviation Model Matching dynamic priority algorithm (PMADP), calculate and key message priority meter including critical data identification, key message weight calculation, key message frequency The process calculated.Wherein critical data identification link, existing algorithm and model can be applied to this as data recognition rule Bright method.

Critical data identification process

Key message time window：Key message starts to calculate to key message the time interval being treated to end point, this It it is the vital stage of key message, key message is once processed (carry out identification data using key message recognition rule to obtain respectively Class key message, calculates the frequency of every class key message by key message frequency computing formula, designs every class key message Weight, by critical data weight, calculates the priority that the frequency obtaining to calculate every class key message.) after, the mark of key message Note will reset to original state.

Key message rule KRULE：The logical relation of mark key message, logical relation contains number in real-time stream According to calculated relationship, incidence relation, judgment criterion, algorithm and the computation model with independent function.Logical relation can be divided into：

Simple logic relation：The logical relation of only single logical operationss relation.

DATA=data | data ∈ R } (1)

In formula (1), R is a kind of simple relation, and DATA is all data acquisition systems meeting certain relation R, and formula (1) represents full All data acquisition systems of sufficient relation R.

The production logic of formula (1) is represented by：

R—>DATA IF R THEN DATA (2)

Wherein：R is unity logic relation, and DATA is the data meeting R relation.

Compound logic relation：Passed through by multiple simple logic operation relations or and, the complex relation that constitutes such as XOR.

In formula (3), R1, R2, R3, R4 are a kind of simple relations, and DATAC is all data acquisition systems meeting these relations, Formula (3) represents all and meets relation R1, R2, all data acquisition systems of the various combination of R3, R4.

Key message model KMODEL：The set of the various key message rules of mark key message, to real-time processing number Mated according to stream, filtered out the key message meeting rule by coupling.

KMODEL=x | x ∈ KRULE } (4)

Key message rule is a point total relation with key message model, the relation of key message rule and key message model As Fig. 1.Real-time stream is mated by key message model, is calculated and identification by key message logic rule, filters Go out to meet the data acquisition system of key message rule.

Key message model is a dynamic model, and in different system, the recognition rule of key message is different, passes through Model configuration file record, description key message rule, build key message model.

Before identification critical data, first key message model configuration file is loaded and parsed, obtain and close Key information model data MS, key message model data is a key message regular collection, a key message rule parsing Result corresponds to a key message mark, resolution logic such as Fig. 2.

Key message model property：

Key message model is key message regular collection, that is, key message model by numerous key messages rule with The rule composition of these key message rule association relations is described；

One key message rule corresponds to a key message rule flag, after model configuration file loads, regular and mark The one-to-one corresponding known is indicated in calculator memory.

Key message weight

Key message weight：Different types of key message, significance level in systems is different, to different crucial letters Breath gives different numerical value, represents key message significance level in systems by the size giving numerical value.

Initial data is mated through key message model, identifies key message, and key message is mapped to be had The interval of different weights, such as Fig. 3.

Key message weight dynamic change, in different system, different business, the weighted value of key message is different, leads to Cross the relation of key message weight configuration file record, description key message data weighting and key message.

Before identification critical data, load weight configuration file, weight configuration file is parsed, calculate weight mark Will, obtains weight mark set WS, and WS is put into internal memory.Implement step as follows：

1) load weight configuration file；

2) parsing is entered to weight configuration file；

3) calculate weight mark；

4) weight mark puts into internal memory list.

A kind of weight type corresponds to a weight mark, represents different weight marks by different numerical value (weighted value), leads to Cross the weight type that different weighted values represent different key messages, the relationship of the two corresponds.Key message and weight type, power Weight values relation such as Fig. 4.

Weight mark configuration file and key message configuration file are associated by key message rule flag.

In conjunction with MS and WS, the data flow after original data stream, process is identified, recognition result passes through 1,0 two-value table Show, 1 expression identification data belongs to key message, 0 represents that identification data is non-critical information, if data meets key message rule Then, then the data being identified is key message, is not otherwise, algorithm such as (5)：

Wherein：

KS_nIt is that key message rule starts in current time in recognition cycle, every class key message that system identification goes out Number of times set, DATA is identified data, and MS is model data.Key message identification internal memory logic is shown in Fig. 4.

Key message frequency

Key message frequency：The number of times occurring in the key message unit interval.I.e. key message rule is in key message Between in window, every class key message frequency of occurrences.Frequency formula such as (6):

Pre=KS_n/t (6)

Wherein：

KS_nThe set of the number of times that every class key message identifies, t be key message rule recognition cycle start to The time span of current time, clock memory relation is shown in Fig. 5.

Key message priority

Key message priority：Arithmetic average, key message frequency and weight that key message frequency is amassed with weight, certainly Determine key message significance level in systems.

In a time slice, change over time, the continual analysis of mass data, the priority of key message is not Break and change, Mobile state adjustment is entered by the key message priority that visualization is shown, real-time, preferential display priority is high Data, by being averaged with the product of weight to the frequency of key message, average is big, and priority is high, and priority data calculates It is a lasting process, priority dynamically adjusts formula such as (7)：

Wherein：

Pre is the frequency of key message, is weight initial value, and WS is weight mark, Pre_iRepresent i-th key message Frequency.

In real-time analyzer, in real time, continually enter system, Pri value is big, then key message is excellent for mass data First level is higher, and data is more crucial, and key message priority forming process relationship memory is shown in Fig. 6.

Experimental analysiss

According to the achievement in Nsfc Major project " unconventional sudden incidents report research " Statistics, in 2001 to 2010 years especially great natural disasters of China, forest fire tops the list.Threshold method is to grind in warning system Study carefully, apply one of more method, experiment is compared analysis for references object to this algorithm and threshold method with forest fire.

The process logic such as formula (8) of threshold method

IF R>=X THEN Doing (8)

If R is more than or equal to X, then do logical process.In formula 8, X is the threshold value of setting, and R is initial data or various calculation The end value that method calculates.

Air humidity, temperature, the hydrocarbonaceous amount of in the air, wind-force, thunder and lightning are to determine the key factor that forest fire occurs.Gloomy Cigarette in woods, flame are the direct embodiments of forest fires.

Data configuration situation explanation：

Air humidity：It may occur however that fire during less than 61.6%；

Area and temperature coefficient：R=0.367, P>0.01 it may occur however that fire；

The hydrocarbonaceous amount of in the air：During more than 0.2%, easily there is fire；

Wind-force：It may occur however that fire during less than 2 grades；

Thunder and lightning：During more than 50 kilo-ampere, easy initiation fire；

Cigarette：It is fiery point；

Flame：It is fiery point.

Area of woods is big, and temperature, humidity, wind-force, gas concentration, cigarette, fiery data change at any time, and this is accomplished by sensor Quantity is many, and acquisition time interval is short, therefore contains much information, forest fire is effectively monitored, and application system needs to accomplish：

Alarm interference is few；

The high integration stress of priority, preferential display.

Data type, weight and threshold value relation are shown in Table 1.

Air humidity, area and temperature coefficient, hydrocarbonaceous amount, wind-force, thunder and lightning, cigarette, fiery data are built by analog form.Phase Hope that experimental data can be tried one's best in the time dimension of 200 minutes and reflect the actual data change situation of a long period, because This experimental data has larger undulatory property.

Fig. 7 is air humidity, area and temperature coefficient, hydrocarbonaceous amount, wind-force, thunder and lightning, cigarette, the result of fiery threshold method.

According to formula 8, using the identical data of Fig. 7, air humidity, area and temperature coefficient, hydrocarbonaceous amount, wind-force, thunder and lightning, Cigarette, the frequency of fire are shown in Fig. 8.

According to formula 9, using the identical data of Fig. 7, air humidity, area and temperature coefficient, hydrocarbonaceous amount, wind-force, thunder and lightning, Cigarette, the priority of fire are shown in Fig. 9.

As seen in Figure 7, using threshold method, alarm interference is many, and key message emphasis does not project.

By Fig. 8,9 as can be seen that using new algorithm, invalid warning information reduces, the information priorities highest of most critical, It is displayed by priority and pay close attention to.

Table 3.1 data type, weight and threshold value relation

Claims (5)

1. a kind of unconventional accident dynamic priority method based on Model Matching it is characterised in that：Know including critical data Not, the process that key message weight calculation, the calculating of key message frequency and key message priority calculate；

Critical data identification process

Key message model KMODEL：The set of the various key message rules of mark key message, to real-time processing data stream Mated, filtered out the key message meeting rule by coupling；

Key message weight

Different types of key message, significance level in systems is different, gives different numerical value to different key messages, Key message significance level in systems is represented by the size giving numerical value；

Key message frequency

Key message frequency, the number of times occurring in the key message unit interval；I.e. key message rule is in key message time window Interior, every class key message frequency of occurrences.

Key message priority

Key message priority, arithmetic average, key message frequency and weight that key message frequency is amassed with weight, determine Key message significance level in systems.

2. the unconventional accident dynamic priority method based on Model Matching as claimed in claim 1 it is characterised in that：? Before identification critical data, first key message model configuration file is loaded and parsed, obtained key message model Data MS, key message model data is a key message regular collection, a key message rule parsing result corresponding Individual key message mark.

3. the unconventional accident dynamic priority method based on Model Matching as claimed in claim 1 it is characterised in that：Former Beginning data is mated through key message model, identifies key message, and key message is mapped to there is different weights Interval；Key message weight dynamic change, in different system, different business, the weighted value of key message is different, by closing The relation of key information weight configuration file record, description key message data weighting and key message.

Before identification critical data, load weight configuration file, weight configuration file is parsed, calculate weight mark, obtain Obtain weight mark set WS, and WS is put into internal memory.Implement step as follows：

1) load weight configuration file；

2) parsing is entered to weight configuration file；

3) calculate weight mark；

4) weight mark puts into internal memory list.

A kind of weight type corresponds to a weight mark, represents different weight marks by different numerical value (weighted value), by not Represent the weight type of different key messages with weighted value, the relationship of the two corresponds；

Weight mark configuration file and key message configuration file are associated by key message rule flag；

In conjunction with MS and WS, the data flow after original data stream, process is identified, recognition result is represented by 1,0 two-value, 1 Represent that identification data belongs to key message, 0 represents that identification data is non-critical information, if data meets key message rule, The data being then identified is key message, is not otherwise, algorithm such as (5)：

${\mathrm{KS}}_n=FRecognition(DATA,MS)=\left\{\begin{array}{c}({\mathrm{\Σ}}_{i=1}^{n-1}{\mathrm{KS}}_i+1,IFdata\∈MS\\ {\mathrm{\Σ}}_{i=1}^{n-1}{\mathrm{KS}}_i,IFdata\∈MS\end{array}\right.---\left(5\right)$

Wherein：

KS_nIt is that key message rule starts in current time in recognition cycle, the number of times of every class key message that system identification goes out Set, DATA is identified data, and MS is model data.

4. the unconventional accident dynamic priority method based on Model Matching as claimed in claim 3 it is characterised in that：Frequently Rate formula such as (6):

Pre=FFrequency (KS_n, t)=KS_n/t (6)

Wherein：

KS_nIt is the set of the number of times that every class key message identifies, t is that key message rule starts to current in recognition cycle The time span of time.

5. the unconventional accident dynamic priority method based on Model Matching as claimed in claim 4 it is characterised in that：

In a time slice, change over time, the continual analysis of mass data, the priority of key message is constantly sent out Changing, by entering Mobile state adjustment, real-time, preferential display priority is high number to the key message priority of visualization display According to by being averaged with the product of weight to the frequency of key message, average is big, and priority is high, and it is one that priority data calculates Individual lasting process, priority dynamically adjusts formula such as (7)：

$\mathrm{Pr}i=FFreDynamicAdj(\mathrm{Pr}e,WS)=\frac{{\mathrm{\Σ}}_{i=1}^n({\mathrm{Pre}}_i*WS)}{n}---\left(7\right)$

Wherein：

Pre is the frequency of key message, is weight initial value, and WS is weight mark；

In real-time analyzer, in real time, continually enter system, Pri value is big, then key message priority for mass data Higher, data is more crucial.