CN106127331A - Civil aviaton based on rough set theory radio interference Forecasting Methodology - Google Patents

Abstract

A kind of civil aviaton based on rough set theory radio interference Forecasting Methodology.First it carry out attributive classification to civil aviaton's radio interference report data, extracts and is disturbed process association attributes, establish conditional attribute collection and decision attribute, sets up interference decision data table.Then, each attribute value contained by interference decision data table is carried out data prediction, pre-processed results uses special reduction method based on data base's design carry out attribute reduction and Value reduction, obtains interfering well cluster rule list.Finally, utilize belief function to try to achieve the conditional attribute value influence degree table to decision attribute, utilize decision rules table and influence degree table, jamming report data are carried out running status diagnosis and provides the interference early warning value in area contained by these data.The inventive method accuracy that predicts the outcome is higher than based on seasonal effect in time series Forecasting Methodology, and is independent of assumed condition, to precise information not requirement.

Description

Civil aviaton based on rough set theory radio interference Forecasting Methodology

Technical field

The invention belongs to data mining and civil aviaton's radio interference analysis technical field, particularly relate to a kind of based on coarse Civil aviaton's radio interference Forecasting Methodology that collection is theoretical.

Background technology

Along with the high speed development of Civil Aviation Industry, the flight flow of air traffic control system (ATCS) management and control constantly increases, therefore civil aviaton Industry is also improving constantly for the reliability requirement of blank pipe communication and navigation monitoring system, particularly with the call of beechnut The indexs such as quality, coverage, system stability and reliability propose requirements at the higher level.Powerful however as China's economy Increasing, radio extensively changes in social each system, specialized application becomes normality, thus the electromagnetism ring of Civil Aviation System Border is the most complicated, and radio interference has become the main cause affecting communication and navigation monitoring system Efficient Operation.Cause dry at present That disturbs is a lot of because have, and Civil Aviation System is the most likely caused by the system such as broadcasting and TV, electric power transmission or even common people's home electronics Radio interference.Meanwhile, inside Civil Aviation System, the intermodulation of admittance equipment transmitter, radio's hydraulic performance decline etc. Factor causes radio interference the most increasing.

Research worker in Civil Aviation Industry conducts in-depth research in terms of strengthening civil aviaton's radio system reliability, improves nothing The work of the aspects such as line electrical interference investigation ability is also quickly being carried out.Such as it is erected at civil aviaton of the U.S. radio interference inspection of the whole America Examining system (Interference Monitoring Detection System, IMDS) possesses the strongest interference monitoring investigation energy Power, China is also making great efforts to strengthen the capacity of resisting disturbance of radio of civil aviaton simultaneously, but either strengthens equipment dependability still The work improving interference monitoring technical merit does not the most possess the ability according to radio interference data analysis predicted interference.Along with boat The surge of class's flow, Civil Aviation Industry is being increased sharply by the number of times of radio interference, and the radio interference of real-time report is reported as Analysis interference Producing reason provides valuable data.Utilize the theoretical method of rough set attribute reduction, to radio interference report Accusing data to carry out excavating to obtain decision rules, the report data in conjunction with monitoring in real time are analyzed contrast, can be prediction nothing The situation that line electrical interference occurs provides to be supported.

From long-term radio interference report data it can be seen that major part radio interference to be airborne very high frequency(VHF) air-ground logical Telephone system experienced interference, disturbs the information such as signal characteristic to be only capable of being reflected to controller by flight unit.This type of is done Disturbing, field technicians cannot realize monitoring radio interference scenarios in real time.In terms of control operational angle, area control Experienced interference is most frequent, thus wide flight range is for capturing the concrete frequency domain character pole of interference signal timely and accurately For unfavorable.On the other hand, flight unit and civil aviaton radio station management personnel are different for the cognitive assessment of interference so that The interference characteristic reported up there are differences between describing.As can be seen here, above-mentioned factors all can cause radio interference to be reported There is inexactness, imperfection and the characteristic such as sudden in data.

Due to rough set theory relative to other method inexact knowledge expression, learn, the aspect such as conclusion has prominent The feature gone out, so research civil aviaton based on rough set theory radio interference Forecasting Methodology has important theory and reality should By value.And not yet occur at present utilizing rough set theory that civil aviaton's radio interference report data are predicted analytical technology Correlational study achievement.

Summary of the invention

In order to solve the problems referred to above, it is an object of the invention to provide a kind of civil aviaton based on rough set theory radio and do Disturb Forecasting Methodology.

In order to achieve the above object, civil aviaton based on the rough set theory radio interference Forecasting Methodology bag that the present invention provides Include the following step carried out in order:

1) collect the civil aviaton's radio interference report data in a certain duration, set up jamming report tables of data JRT'；

2) attribute contained in above-mentioned jamming report tables of data JRT' is analyzed, removes the genus unrelated with the process that is disturbed Property, is made up of decision table DT' the remaining attribute relevant to the process that is disturbed and property value, and determine conditional attribute collection E' and Decision attribute D'；

3) according to the Time segments division relevant to civil aviaton, the time belonging to conditional attribute in above-mentioned decision table DT' is carried out continuously Data Discretization, utilizes regular expression specification handles conditional attribute collection E' and the value of decision attribute D', thus to decision table In DT', data carry out pretreatment, obtain pretreated decision table PDT'；

4) for step 3) in the pretreated decision table PDT' that obtains use and utilize data base based on rough set theory The reduction method of design carries out attribute reduction and Value reduction, obtains decision rules table DRT'；

5) utilize belief function methods analyst above-mentioned decision rules table conditional attribute difference value for disturbance ground The influence degree in district, obtains property value influence degree Table I T'；

6) utilize step 4) in the decision rules table DRT' that obtains and step 5) in the property value influence degree table that obtains Civil aviaton's radio interference report data are carried out running status diagnosis by IT', to provide interference early warning value in area contained by data, from And make prediction.

In step 3) in, the described time according to the Time segments division relevant to civil aviaton to belonging to conditional attribute in decision table The method carrying out continuous data discretization is:

Define according to the period, conditional attribute time difference value is equivalent to different periods, specifically to represent 0:01 morning To 6:00 time point, to represent 6:01 to 8:00 time point morning, to represent 8:01 to 11:00 time point the morning, with generation at noon Table 11:01 to 13:00 time point, to represent 13:01 to 17:00 time point afternoon, to represent 17:01 to the 19:00 time at dusk Point, to represent 19:01 to 24:00 time point evening.

In step 4) in, described for step 3) in the pretreated decision table PDT' that obtains use based on rough set The reduction method that theory utilizes data base to design carries out attribute reduction and Value reduction, the method obtaining decision rules table DRT' For:

Input: pretreated decision table PDT₁=(U₁,A₁=E₁∪D₁,V₁,f₁), wherein U₁For domain, for object x's Set；A₁For community set；E₁For conditional attribute collection, for the set of conditional attribute e；D₁For decision attribute, for decision attribute values d Set；V₁For attribute codomain, f₁:U₁×A₁→V₁For specifying domain U₁In the information function of each object attribute values；

Output: the decision rules table DRT' of pretreated decision table PDT'；

Concrete grammar is as follows:

(1) analysis condition attribute necessity, removes pretreated decision table PDT' conditional attribute e₁And all take , if there are two corresponding identical decision attribute values d of incomplete same combination in the combination of remaining conditional attribute value in value₁Situation, Then conditional attribute e₁For indispensable attributes, being otherwise unnecessary, all conditions attribute in ergodic condition property set, with pretreated Whole essential condition attributes of decision table PDT' and value set up core value analytical table DT_c；

(2) analyze core value, remove core value analytical table DT_cMiddle object x_iConditional attribute e_jIf remaining conditional attribute combines Value in the situation of like combinations correspondence difference decision attribute values d, then conditional attribute e occur_jValue be object x_iCore value, no Do not consider, traverse object x_iTravel through all objects after all conditions attribute, preserve all core values；

(3) obtain decision rules, mark all core values in pretreated decision table PDT', with certain decision attribute values d_aRight All object set { the x answered_i, (i=1,2 ... I) be scope, remove the exactly the same conditional attribute without core value of value and After value, at object x_iIn make core value add its non-core conditional attribute e_jValue be a combination, if possessing in the object of like combinations Other conditional attribute values are different, then retain combination, do not consider, traverse object x_iTraverse object collection after all non-core attributes Close { x_iAll objects in }, finally travel through the object set of all decision attribute values d, set up decision-making rule according to withed a hook at the end combination Then table DRT'.

In step 5) in, described utilizes belief function methods analyst above-mentioned decision rules table conditional attribute difference value For the influence degree in disturbance area, the method obtaining property value influence degree Table I T' is:

For somewhere, certain state of conditional attribute is equivalent to this state to this ground phase to the influence degree that this ground is disturbed Answer the degree of membership of conditional attribute value, take statistical sample method, by influence degree according to jamming report large sample observing matrix Statistic frequency approximation；Conditional attribute e_jQ-th state be e_jq, q=1,2 ..., Q_j, the value of decision attribute mush area Collection is combined into D={d₁,d₂,…d_m, wherein the v mush area is d_v；Represent that the sample being made up of jamming report data is empty In, by conditional attribute e_jqAffect mush area d_vDisturbance records number of times,Represent by conditional attribute e_jqImpact Total disturbance records number of times of each department, it is known that conditional attribute e_jqTo mush area d_vThe influence degree being disturbedFor:

$p_v^q\left(e_j\right)=\frac{u_{qv}^{\left(j\right)}}{\underset{v=1}{\overset{m}{\Σ}}{u}_{qv}^{\left(j\right)}}=\frac{u_{qv}^{\left(j\right)}}{u_q^{\left(j\right)}}.$

In step 6) in, described utilizes step 4) in the decision rules table DRT' that obtains and step 5) in the genus that obtains Property value influence degree Table I T', civil aviaton radio interference report data are carried out running status diagnosis, to provide area contained by data Interference early warning value, thus the method made prediction is:

For certain mush area, running status diagnosis is based on judging that somewhere section interference data are disturbed decision-making rule to this ground Property value contained by degree of membership then, i.e. decision rules and interference data similar situation；To belong to the mush area of decision attribute values d_vAs a example by, its early warning value being disturbed is shown below:

$W_v=\underset{l=1}{\overset{L_v}{\Σ}}\underset{t=1}{\overset{T_v}{\Σ}}\underset{j=1,q=1}{\overset{J,Q_j}{\Σ}}\frac{c_{l,t,j,q}\·p_v^q\left(e_j\right)}{L_v\·T_v}$

In formula, W_vFor certain section interference data to mush area d_vIt is disturbed early warning value, L_vFor in data about d_vFeature become Amount number of combinations, T_vFor d_vDecision rules bar number, J is the t article decision rules non-null attribute number, Q_jContain for jth conditional attribute Property value number；c_l,t,j,qIt is characterized in this segment data, mush area d_vThe combination of the l characteristic variable in, jth feature becomes Amount and mush area d_vThe similarity of the l article decision rules corresponding conditions attribute；For jth conditional attribute q state To mush area d_vThe influence degree being disturbed, i.e. conditional attribute status weights, these weights are by step 5) in the genus that obtains Property value influence degree Table I T ' try to achieve.

The present invention is directed to inexactness that civil aviaton radio interference report data possess, imperfection and the spy such as sudden Property, utilize rough set theory (Rough Set) relative to other method inexact knowledge expression, learn, the aspect such as conclusion The advantage possessed, it is provided that a kind of civil aviaton based on rough set theory radio interference Forecasting Methodology: first, to civil aviaton's radio Jamming report data carry out attributive classification, extract and be disturbed process association attributes, establish conditional attribute collection and decision attribute, Set up interference decision data table.Then, each attribute value contained by interference decision data table is carried out data prediction, to pretreatment Result uses special reduction method based on data base's design to carry out attribute reduction and Value reduction, obtains interfering well cluster rule Table.Finally, (conditional attribute status is weighed to the influence degree of decision attribute to utilize belief function to try to achieve conditional attribute value Value) table, utilize above-mentioned decision rules table and influence degree table, jamming report data are carried out running status diagnosis and provides this data The interference early warning value in contained area.By the temporary this method such as conditional attribute value predict the outcome with based on time series forecasting skill Predicting the outcome of art compares, and this method accuracy that predicts the outcome is higher than based on seasonal effect in time series Forecasting Methodology, and is independent of Assumed condition, to precise information not requirement.After using the conditional attribute value weight tried to achieve based on belief function, this method Predict the outcome compared with the situation using the power such as conditional attribute value, for the actual prediction effect being disturbed the less area of number of times Really have and be obviously improved, it was predicted that accuracy has increased, and demonstrates the effectiveness of this method further.

Accompanying drawing explanation

Civil aviaton based on the rough set theory radio interference Forecasting Methodology flow chart that Fig. 1 provides for the present invention.

Fig. 2 be based on the inventive method, commonly use based on time series Predicting Technique interference predicting method predict the outcome with Practical situation comparison diagram.

Fig. 3 is that in the case of conditional attribute value etc. is weighed, certain information district is disturbed situation and prediction case comparison diagram.

Fig. 4 is disturbed situation and prediction case contrast for certain information district in the case of considering the impact of conditional attribute difference value Figure.

Detailed description of the invention

Civil aviaton based on the rough set theory radio interference with specific embodiment, the present invention proposed below in conjunction with the accompanying drawings Forecasting Methodology is described in detail.

As it is shown in figure 1, civil aviaton based on the rough set theory radio interference Forecasting Methodology that the present invention provides includes by suitable The following step that sequence is carried out:

1) collect civil aviaton's radio interference report (abbreviation jamming report) data in a certain duration, set up jamming report number According to table JRT'；

Table 1 is jamming report tables of data JRT' example, and it comprises record date, is disturbed frequency, is disturbed time, interference Degree, interference classification, interference signal characteristic, interference source type, mush area, impact, take measures, report without committee's time And remarks are at interior attribute.

Table 1, jamming report tables of data JRT' example

2) attribute contained in above-mentioned jamming report tables of data JRT' is analyzed, removes the genus unrelated with the process that is disturbed Property, is made up of decision table DT' the remaining attribute relevant to the process that is disturbed and property value, and determine conditional attribute collection E' and Decision attribute D'；

It is analyzed above-mentioned jamming report tables of data JRT' understanding, therein impacts, take measures, when reporting without committee Between and remarks for interference consequence, unrelated with being disturbed process, therefore these attributes are removed, by remaining attribute and property value structure Becoming the decision table DT' shown in table 2, domain U' is all data messages in decision table DT', and object x' is any one interference in table Record, arranging conditional attribute collection E' is { frequency, time, degree, classification, interference signal characteristic, interference source type }, decision attribute D' is the set of mush area, and the value of decision attribute D' is different mush area d', and decision rules is above-mentioned conditional attribute The valued combinations of collection E' causes decision attribute D' to be disturbed.

Table 2, decision table DT'

3) according to the Time segments division relevant to civil aviaton, the time belonging to conditional attribute in above-mentioned decision table DT' is carried out continuously Data Discretization, utilizes regular expression specification handles conditional attribute collection E' and the value of decision attribute D', thus to decision table In DT', data carry out pretreatment, obtain pretreated decision table PDT'；

Table 3 is to the decision table PDT' set up after table 2 pretreatment, divides according to flight relevant time period, takes with " morning " etc. Value substitutes the value of attribute time, is specially to represent 0:01 to 6:00 time point morning, when representing 6:01 to 8:00 with morning Between point, to represent 8:01 to 11:00 time point the morning, to represent 11:01 to 13:00 time point noon, represent 13:01 with afternoon To 17:00 time point, to represent 17:01 to 19:00 time point, to represent 19:01 to 23:59 time point evening at dusk.Foundation Station name in standard chart, is " H ground " by the value lack of standardization " H ground " of mush area D', " near H ground " specification in table 2, In the case of large sample, utilize regular expression specification handles conditional attribute collection and decision attribute in data base's computing module of design Value.

Table 3, pretreated decision table PDT'

4) for step 3) in the pretreated decision table PDT' that obtains use and utilize data base based on rough set theory The reduction method of design carries out attribute reduction and Value reduction, obtains decision rules table DRT'；

Input: pretreated decision table PDT'=(U₁,A₁=E₁∪D₁,V₁,f₁), wherein U₁For domain, for object x's Set；A₁For community set；E₁For conditional attribute collection, for the set of conditional attribute e；D₁For decision attribute, for decision attribute values d Set；V₁For attribute codomain, f₁:U₁×A₁→V₁For specifying domain U₁In the information function of each object attribute values.

Output: the decision rules table DRT' of pretreated decision table PDT'；

Concrete grammar is as follows:

(1) analysis condition attribute necessity, removes pretreated decision table DRT' conditional attribute e₁And all take , if there are two corresponding identical decision attribute values d of incomplete same combination in the combination of remaining conditional attribute value in value₁Situation, Then conditional attribute e₁For indispensable attributes, being otherwise unnecessary, all conditions attribute in ergodic condition property set, with pretreated Whole essential condition attributes of decision table DRT' and value set up core value analytical table DT_c；

(2) analyze core value, remove core value analytical table DT_cMiddle object x_iConditional attribute e_jIf remaining conditional attribute combines Value in the situation of like combinations correspondence difference decision attribute values d, then conditional attribute e occur_jValue be object x_iCore value, no Do not consider, traverse object x_iTravel through all objects after all conditions attribute, preserve all core values；

(3) obtain decision rules, mark all core values in pretreated decision table DRT', with certain decision attribute values d_aRight All object set { the x answered_i, (i=1,2 ... I) be scope, remove the exactly the same conditional attribute without core value of value and After value, at object x_iIn make core value add its non-core conditional attribute e_jValue be a combination, if possessing in the object of like combinations Other conditional attribute values are different, then retain combination, do not consider, traverse object x_iTraverse object collection after all non-core attributes Close { x_iAll objects in }, finally travel through the object set of all decision attribute values d, set up decision-making rule according to withed a hook at the end combination Then table DRT'.

Said method reduced unitized table 3 is utilized to obtain the decision rules table DRT' shown in table 4.

Table 4, decision rules table DRT'

5) utilize belief function methods analyst above-mentioned decision rules table conditional attribute difference value for disturbance ground The influence degree in district, obtains property value influence degree Table I T'；

For somewhere, certain state of conditional attribute is equivalent to this state to this ground phase to the influence degree that this ground is disturbed Answer the degree of membership of conditional attribute value, take statistical sample method, by influence degree according to jamming report large sample observing matrix Statistic frequency approximation.

Conditional attribute e_jDisturbance records statistics as shown in table 5, wherein conditional attribute e_jQ-th state be e_jq, q=1, 2,…,Q_j, the value collection of decision attribute mush area is combined into D'={d₁,d₂,…d_m, wherein the v mush area is d_v。 In representing the sample space being made up of jamming report data, by conditional attribute e_jqAffect mush area d_vDisturbance records number of times,Represent by conditional attribute e_jqTotal disturbance records number of times of each department of impact, it is known that conditional attribute e_jqTo interference Area d_vThe influence degree being disturbedFor:

$p_v^q\left(e_j\right)=\frac{u_{qv}^{\left(j\right)}}{\underset{v=1}{\overset{m}{\Σ}}{u}_{qv}^{\left(j\right)}}=\frac{u_{qv}^{\left(j\right)}}{u_q^{\left(j\right)}}---\left(3\right)$

The interference source type difference value obtained according to table 3 is as shown in table 6, due to table to mush area influence degree example Contained by 3, record strip number is less, causes occurring that partial value is the situation of 0 to indivedual areas influence degree.Decision-making in actual application Table is collected by mass data and obtains, and conditional attribute collection status and mush area quantity are more, and each state can to different regions Reliability situation specifically and truly, therefore the most tired is stated.

Table 5, conditional attribute e_jDisturbance records statistical table

Table 6, interference source type value are to mush area influence degree example

6) utilize step 4) in the decision rules table DRT' that obtains and step 5) in the property value influence degree table that obtains Civil aviaton's radio interference report data are carried out running status diagnosis by IT', to provide interference early warning value in area contained by data, from And make prediction.

For certain mush area, running status diagnosis is based on judging that somewhere section interference data are disturbed decision-making rule to this ground Property value contained by degree of membership then, i.e. decision rules and interference data similar situation, the early warning value that somewhere is disturbed such as formula (4) institute Show:

$W_v=\underset{l=1}{\overset{L_v}{\Σ}}\underset{t=1}{\overset{T_v}{\Σ}}\underset{j=1,q=1}{\overset{J,Q_j}{\Σ}}\frac{c_{l,t,j,q}\·p_v^q\left(e_j\right)}{L_v\·T_v}---\left(4\right)$

In formula (4), W_vFor certain section interference data to mush area d_vIt is disturbed early warning value, L_vFor in data about intrusively District d_vCharacteristic variable number of combinations, T_vFor mush area d_vDecision rules bar number, J is the t article non-null attribute of decision rules Number, Q_jFor jth conditional attribute number Han property value.c_l,t,j,qIt is characterized in this segment data, mush area d_vCorresponding l In characteristic variable combination, jth characteristic variable value and mush area d_vThe t article decision rules corresponding conditions attribute residing for shape The similarity of state.For jth conditional attribute q state to mush area d_vThe influence degree being disturbed, i.e. conditional attribute Status weights, these weights are by step 5) in property value influence degree Table I T' that obtains try to achieve.

Utilize the interference on certain continuous ten days interference data prediction H ground, H ground in table 7.Table 7 is containing 7 data, and in table 4, H ground only contains Article 1, possess time and the decision rules of the non-null attribute of classification the two, then L_v=7, T_v=1, J=2.Attribute time contains 7 attributes Value (morning, morning, the morning, noon, afternoon, at dusk, evening), classification is containing 3 property values (aerial, ground, vacant lot), then Q₁= 7,Q₂=3.Consider the power situations such as conditional attribute status,For Q_jInverse, thenTable C in 8_l,t,j,qFor the similarity situation of data in table 7 and H ground decision rules, with c_1,1,1,6As a example by, table 7 the 1st data frequency becomes Amount has 1 identical value at dusk with H ground decision rules conditional attribute, and property value is the 6th state, i.e. l of attribute time at dusk =1, t=1, j=1, q=6, then c_1,1,1,6=1.By c in above-mentioned element value and table 8_l,t,j,qBring in formula (4), obtain H ground Interference early warning value is 0.16327.

Certain the ten days interference tables of data in table 7, H ground

Table 8, H ground decision rules similarity information slip

When decision attribute values is other each department, decision rules may be a plurality of, need to analyze each rule similarity successively. In practical situation, decision rules table is to obtain based on the study to a large amount of interference data, and use when being predicted the most dry The data volume disturbing data is relatively big, tires out the most in a tabular form and states.

Measured result and analysis

The effect of civil aviaton based on the rough set radio interference Forecasting Methodology that the present invention provides can be passed through and actual number Further illustrate according to confirmatory experiment.Actual measurement parameter is arranged: data basis is the civil aviaton of continuous 120 days durations in certain navigational intelligence district Radio interference report data, Time Series Forecasting Methods dimension is 5, and damped coefficient uses 0.3, different Forecasting Methodology results pair Than time the present invention provide weights based on rough set Forecasting Methodology conditional property set status for wait power.

1, predict the outcome contrast with classical Forecasting Methodology

The Forecasting Methodology that Fig. 2 provides for utilizing the present invention with classical based on Time Series Forecasting Methods predict the outcome right Than figure, experimental data is civil aviaton's radio interference report data of in certain information district 120 days, and contrast is with reference to being in this information district 7 Individual the actual of the main station is disturbed number of times.Wherein use rolling average and exponential smoothing based on Time Series Forecasting Methods Two methods, its dimension is set to 5, and damped coefficient uses 0.3.It can be seen that the present invention provides based on rough set theory pre- The accuracy that predicts the outcome of survey method is higher than traditional algorithm, and the tendency for 7 main station early warning values is disturbed with actual Number of times is basically identical.

2, conditional attribute value etc. temporary predict the outcome and actual interference situation contrasts

Fig. 3 is based on the inventive method, is reported as predicting the outcome and actual interference of data with the interference in 120 days of certain information district Situation comparison diagram, the influence degree of decision attribute is weighed by the conditional attribute difference value wherein used in formula (4) for waiting.From figure 3 understand, actual be disturbed number of times more (be disturbed number of times not less than 2) in the case of, it was predicted that tendency and the reality of result are subject to Interference number of times situation of change is more consistent.But to being disturbed the station area of number of times less (less than 2), it was predicted that result compares reality There is certain jitter phenomenon in situation, this is less relevant with the disturbance records sample of the part station in data analysis, and this can cause writing from memory The power situations such as each conditional attribute value recognized exist relatively greatly partially for the influence degree of the station with actual conditional attribute value Difference, needs to utilize step 5 in formula (4)) belief function mentioned analyzes attribute value weights.

3, predict the outcome after utilizing belief function analysis condition attribute value weights and actual interference situation contrasts

Fig. 4 is to use under the data premise of identical 120 days with Fig. 3, uses step 5) in the belief function mentioned analyze After the attribute value weights arrived, predicting the outcome and practical situation comparison diagram of the inventive method.Comparison diagram 3 and Fig. 4 understands, for It is disturbed the station area of number of times less (less than 2), it was predicted that the jitter phenomenon of result tendency is effectively improved, illustrates to use and trust letter Number analysis condition attribute values weights be effective, through revised predict the outcome than Fig. 3 predict the outcome on the whole Tendency more closely, precision of prediction increases, this result is consistent with the result of theory analysis.

First civil aviaton's radio is done by civil aviaton based on the rough set theory radio interference Forecasting Methodology that the present invention provides The projects disturbing report are analyzed, and remove the redundant attributes unrelated with the process that is disturbed, and determine conditional attribute collection and decision-making Attribute, utilization is disturbed process association attributes and value sets up decision table, will be converted to the problem of jamming report learning knowledge The problem that decision table is obtained decision rules；Then decision table is carried out pretreatment, attribute reduction and Value reduction to obtain certainly Plan rule, associating jamming report data carry out running status diagnosis and obtain interference early warning value, and will predict the outcome and based on time Between sequence method predict the outcome and actual interference situation contrasts；Result shows, in the case of the power such as conditional attribute value, The method is obtained in that be closer to practical situation predicts the outcome, and predicts the main station being disturbed often more Accurately.Additionally, in the case of using the attribute value weights obtained based on belief function, for being disturbed the less station of number of times Prediction effect be obviously improved, and to overall prediction effect and the practical situation more adjunction of mush area contained by interference data Closely.

Claims (5)

1. civil aviaton based on a rough set theory radio interference Forecasting Methodology, it is characterised in that: it includes carrying out in order The following step:

1) collect the civil aviaton's radio interference report data in a certain duration, set up jamming report tables of data JRT'；

2) attribute contained in above-mentioned jamming report tables of data JRT' is analyzed, removes the attribute unrelated with the process that is disturbed, It is made up of decision table DT' the remaining attribute relevant to the process that is disturbed and property value, and determines conditional attribute collection E' and determine Plan attribute D'；

3) according to the Time segments division relevant to civil aviaton, the time belonging to conditional attribute in above-mentioned decision table DT' is carried out continuous data Discretization, utilizes regular expression specification handles conditional attribute collection E' and the value of decision attribute D', thus in decision table DT' Data carry out pretreatment, obtain pretreated decision table PDT'；

4) for step 3) in the pretreated decision table PDT' that obtains use and utilize data base to design based on rough set theory Reduction method carry out attribute reduction and Value reduction, obtain decision rules table DRT'；

5) utilize belief function methods analyst above-mentioned decision rules table conditional attribute difference value for disturbance area Influence degree, obtains property value influence degree Table I T'；

6) utilize step 4) in the decision rules table DRT' that obtains and step 5) in property value influence degree Table I T' that obtains, Civil aviaton's radio interference report data are carried out running status diagnosis, to provide interference early warning value in area contained by data, thus does Go out prediction.

Civil aviaton based on rough set theory the most according to claim 1 radio interference Forecasting Methodology, it is characterised in that: Step 3) in, described carries out consecutive numbers according to the Time segments division relevant to civil aviaton to the time belonging to conditional attribute in decision table According to the method for discretization it is:

Define according to the period, conditional attribute time difference value be equivalent to different periods, specifically to represent 0:01 to 6 morning: 00 time point, to represent 6:01 to 8:00 time point morning, to represent 8:01 to 11:00 time point the morning, to represent 11 noon: 01 to 13:00 time point, to represent 13:01 to 17:00 time point afternoon, to represent 17:01 to 19:00 time point at dusk, with Represent 19:01 to 24:00 time point evening.

Civil aviaton based on rough set theory the most according to claim 1 radio interference Forecasting Methodology, it is characterised in that: Step 4) in, described for step 3) in the pretreated decision table PDT' that obtains use and utilize number based on rough set theory Carrying out attribute reduction and Value reduction according to the reduction method of storehouse design, the method obtaining decision rules table DRT' is:

Input: pretreated decision table PDT'=(U₁,A₁=E₁∪D₁,V₁,f₁), wherein U₁For domain, for the set of object x； A₁For community set；E₁For conditional attribute collection, for the set of conditional attribute e；D₁For decision attribute, for the collection of decision attribute values d Close；V₁For attribute codomain, f₁:U₁×A₁→V₁For specifying domain U₁In the information function of each object attribute values；

Output: the decision rules table DRT' of pretreated decision table PDT'；

Concrete grammar is as follows:

(1) analysis condition attribute necessity, removes pretreated decision table PDT' conditional attribute e₁And whole values, if There are two corresponding identical decision attribute values d of incomplete same combination in the combination of remaining conditional attribute value₁Situation, then condition Attribute e₁For indispensable attributes, being otherwise unnecessary, all conditions attribute in ergodic condition property set, with pretreated decision table Whole essential condition attributes of PDT' and value set up core value analytical table DT_c；

(2) analyze core value, remove core value analytical table DT_cMiddle object x_iConditional attribute e_jIf, taking of remaining conditional attribute combination Value occurs the situation of like combinations correspondence difference decision attribute values d, then conditional attribute e_jValue be object x_iCore value, the most not Consider, traverse object x_iTravel through all objects after all conditions attribute, preserve all core values；

(3) obtain decision rules, mark all core values in pretreated decision table PDT', with certain decision attribute values d_aCorresponding All object set { x_i, (i=1,2 ... I) it is scope, remove the exactly the same conditional attribute without core value of value and value After, at object x_iIn make core value add its non-core conditional attribute e_jValue be a combination, if possessing in the object of like combinations other Conditional attribute value is different, then retain combination, do not consider, traverse object x_iTraverse object set after all non-core attributes {x_iAll objects in }, finally travel through the object set of all decision attribute values d, set up decision rules according to withed a hook at the end combination Table DRT'.

Civil aviaton based on rough set theory the most according to claim 1 radio interference Forecasting Methodology, it is characterised in that: Step 5) in, described utilizes belief function methods analyst above-mentioned decision rules table conditional attribute difference value for different dry Disturbing the influence degree in area, the method obtaining property value influence degree Table I T' is:

For somewhere, certain state of conditional attribute is equivalent to this state bar corresponding to this ground to the influence degree that this ground is disturbed The degree of membership of part attribute value, takes statistical sample method, by the system of influence degree foundation jamming report large sample observing matrix Meter frequency approximation；Conditional attribute e_jQ-th state be e_jq, q=1,2 ..., Q_j, the value set of decision attribute mush area For D={d₁,d₂,…d_m, wherein the v mush area is d_v；In representing the sample space being made up of jamming report data, By conditional attribute e_jqAffect mush area d_vDisturbance records number of times,Represent by conditional attribute e_jqAffect is each Total disturbance records number of times in area, it is known that conditional attribute e_jqTo mush area d_vThe influence degree being disturbedFor:

$p_v^q\left(e_j\right)=\frac{u_{qv}^{\left(j\right)}}{\underset{v=1}{\overset{m}{\Σ}}{u}_{qv}^{\left(j\right)}}=\frac{u_{qv}^{\left(j\right)}}{u_q^{\left(j\right)}}.$

Civil aviaton based on rough set theory the most according to claim 1 radio interference Forecasting Methodology, it is characterised in that: Step 6) in, described utilizes step 4) in the decision rules table DRT' that obtains and step 5) in the property value that obtains affect journey Civil aviaton's radio interference report data are carried out running status diagnosis by degree Table I T', to provide interference early warning in area contained by data Value, thus the method made prediction is:

For certain mush area, running status diagnosis is based on judging that somewhere section interference data are disturbed decision rules to this ground Property value contained by degree of membership, i.e. decision rules and interference data similar situation；To belong to the mush area d of decision attribute values_vFor Example, its early warning value being disturbed is shown below:

$W_v=\underset{l=1}{\overset{L_v}{\Σ}}\underset{t=1}{\overset{T_v}{\Σ}}\underset{j=1,q=1}{\overset{J,Q_j}{\Σ}}\frac{c_{l,t,j,q}\·p_v^q\left(e_j\right)}{L_v\·T_v}$

In formula, W_vFor certain section interference data to mush area d_vIt is disturbed early warning value, L_vFor in data about d_vCharacteristic variable group Close number, T_vFor d_vDecision rules bar number, J is the t article decision rules non-null attribute number, Q_jFor jth conditional attribute containing attribute Value number；c_l,t,j,qIt is characterized in this segment data, mush area d_vThe combination of the l characteristic variable in, jth characteristic variable with Mush area d_vThe similarity of the l article decision rules corresponding conditions attribute；For jth conditional attribute q state to dry Disturb area d_vThe influence degree being disturbed, i.e. conditional attribute status weights, these weights are by step 5) in the property value that obtains Influence degree Table I T ' try to achieve.