CN106294038A

CN106294038A - The generation of a kind of fault spectrum, detection method based on fault spectrum and device

Info

Publication number: CN106294038A
Application number: CN201510272657.8A
Authority: CN
Inventors: 刘迅
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2015-05-25
Filing date: 2015-05-25
Publication date: 2017-01-04
Anticipated expiration: 2035-05-25
Also published as: WO2016188279A1; CN106294038B

Abstract

The embodiment of the present application provides the generation of a kind of fault spectrum, detection method based on fault spectrum and device, and this generation method includes: obtain the first work order data of one or more classification；Each first work order data include fault message and detection information；For every class the first work order data, from described detection information, extract public characteristic word, as characteristic vector；For every class the first work order data, learn the logical relation between described fault message and described characteristic vector, it is thus achieved that every class fault spectrum model；Every class fault spectrum model is carried out pruning modes, it is thus achieved that every class fault spectrum.The embodiment of the present application is by setting up fault spectrum, make to support concurrently to detect according to these one or more detection paths during subsequent detection, decrease detection time-consuming, improve the efficiency of detection, simultaneously, the detection of application and trouble spectrum is simple to operate, greatly reduces the frequency of artificial participation, reduces the consuming of user's energy.

Description

The generation of a kind of fault spectrum, detection method based on fault spectrum and device

Technical field

The application relates to the technical field of computer, particularly relate to a kind of generation method of fault spectrum, one Plant detection method based on fault spectrum, the generating means of a kind of fault spectrum and a kind of detection based on fault spectrum Device.

Background technology

Along with the fast development of science and technology, various products, such as fictitious host computer, cloud platform etc., extensively enter The life of people, learn, the field such as work.

Generally, when product breaks down, user can submit work order to WorkForm System, carries out detecting, tieing up Protect, and then solve fault.

Existing WorkForm System is mainly made up of two subsystems: autonomous answer system and customer service answer are System.

In WorkForm System, user needs oneself consult Help Center's document or solve according to guide prompting Fault.

Owing to user needs to operate investigation step by step according to document or prompting, i.e. serial investigation, expends relatively Many time, the speed of fault detect is slow；Further, the general quantity of technical documentation that WorkForm System is accumulated A lot, operation complexity, need to expend the substantial amounts of energy of user；Additionally, reading technique document needs to need more Knowledge in field will be had accumulation, technical threshold is higher, and the user weak for technology foundation or customer service are very Difficult solving problems by themselves.

Summary of the invention

In view of the above problems, it is proposed that the embodiment of the present application is to provide one to overcome the problems referred to above or extremely Partially solve the generation method of a kind of fault spectrum of the problems referred to above, a kind of detection side based on fault spectrum Method and the generating means of corresponding a kind of fault spectrum, a kind of detection device based on fault spectrum.

In order to solve the problems referred to above, the embodiment of the present application discloses a kind of generation method of fault spectrum, including:

Obtain the first work order data of one or more classification；Each first work order data include that fault is believed Breath and detection information；

For every class the first work order data, from described detection information, extract public characteristic word, as feature Vector；

For every class the first work order data, learn the logic between described fault message and described characteristic vector Relation, it is thus achieved that every class fault spectrum model；

Every class fault spectrum model is carried out pruning modes, it is thus achieved that every class fault spectrum.

Preferably, described fault spectrum includes root node and the leaf node being connected, and described root node characterizes Fault message, described leaf node characterizes detection information, has logic and close between at least part of leaf node System, described leaf node has one or more father node.

Preferably, the described step extracting public characteristic word from described detection information includes:

Described detection information is carried out word segmentation processing, it is thus achieved that one or more first participles；

Add up the word frequency of the described first participle；

The weight of the described first participle is calculated by the word frequency of the described first participle；

According at least part of first participle of described weight extraction as public characteristic word.

Preferably, the described step extracting public characteristic word from described detection information also includes:

The one or more first participle is used to mate in preset disabling in dictionary；

Remove the first participle that the match is successful.

Preferably, the described step that every class fault spectrum model is carried out pruning modes includes:

Identical subtree is searched in described fault spectrum model；Described subtree is one or more leaf node Set；

When finding, the father node of identical subtree is connected to one of them subtree；

In identical subtree, cut off other the subtree outside the subtree connected.

Preferably, the described step that every class fault spectrum model is carried out pruning modes also includes:

According to default prune approach, described fault spectrum model is carried out pruning modes.

The illegal leaf node of logical relation is cut off from described fault spectrum model.

The embodiment of the present application also discloses a kind of detection method based on fault spectrum, including:

When receiving the second work order data, from described second work order extracting data key word；

Search the fault spectrum that described second work order data generic is corresponding；

In described fault spectrum, according to described keyword lookup one or more detection path；

Detect according to the one or more detection path, it is thus achieved that testing result.

Preferably, the described step from described second work order extracting data key word includes:

Described second work order data are carried out word segmentation processing, it is thus achieved that one or more second participles；

Identify the part of speech of the one or more the second participle；

From the one or more second participle, key word is extracted according to described part of speech.

Preferably, the described step from described second work order extracting data key word also includes:

Remove the second participle that the match is successful.

Preferably, described in described fault spectrum, search one or more detection roads according to described Feature Words The step in footpath includes:

In described fault spectrum, search the root node with described Keywords matching；

One or more leaf nodes that traversal is connected with described root node, it is thus achieved that one or more detection roads Footpath.

Preferably, described according to the one or more detection path detect, it is thus achieved that testing result Step includes:

For each detection path, obtain what the one or more leaf nodes in described detection path characterized Detection information；

The detection information characterized according to current leaf node detects, it is thus achieved that couple candidate detection result；

Search next leaf node of mate with described couple candidate detection result of logical relation, return execution according to The detection information that current leaf node characterizes carries out the step detected, until performing to final leaf joint Point；

The couple candidate detection result of final leaf node is set to testing result.

Preferably, described fault spectrum generates in the following manner:

The embodiment of the present application also discloses the generating means of a kind of fault spectrum, including:

Work order data acquisition module, for obtaining the first work order data of one or more classification；Each One work order data include fault message and detection information；

Public characteristic word extraction module, for for every class the first work order data, from described detection information Extract public characteristic word, as characteristic vector；

Fault spectrum model learning module, for for every class the first work order data, learns described fault message And the logical relation between described characteristic vector, it is thus achieved that every class fault spectrum model；

Module pruned by fault spectrum model, for every class fault spectrum model is carried out pruning modes, it is thus achieved that every class Fault spectrum.

Preferably, described public characteristic word extraction module includes:

First participle processing module, for carrying out word segmentation processing to described detection information, it is thus achieved that one or many The individual first participle；

Word frequency statistics module, for adding up the word frequency of the described first participle；

Weight computation module, for calculating the power of the described first participle by the word frequency of the described first participle Weight；

The first participle extracts submodule, is used for according at least part of first participle of described weight extraction as public affairs Feature Words altogether.

Preferably, described public characteristic word extraction module also includes:

First matched sub-block, for using the one or more first participle to disable dictionary preset In mate；

First removes submodule, for removing the first participle that the match is successful.

Preferably, described fault spectrum model pruning module includes:

Sub-tree search submodule, for searching identical subtree in described fault spectrum model；Described subtree Set for one or more leaf nodes；

Connexon module, for when finding, is connected to one of them by the father node of identical subtree Subtree；

First prunes submodule, in identical subtree, cut off outside the subtree connected other Subtree.

Preferably, described fault spectrum model pruning module also includes:

Second prunes submodule, for pruning described fault spectrum model according to default prune approach Process.

Preferably, described fault spectrum model pruning module also includes:

3rd prunes submodule, for cutting off logical relation illegal leaf joint from described fault spectrum model Point.

The embodiment of the present application also discloses a kind of detection device based on fault spectrum, including:

Keyword extracting module, for when receiving the second work order data, from described second work order data Middle extraction key word；

Fault spectrum searches module, for searching the fault spectrum that described second work order data generic is corresponding；

Detection path searching module, in described fault spectrum, according to described keyword lookup one or Multiple detection paths；

Detection module, for detecting according to the one or more detection path, it is thus achieved that testing result.

Preferably, described keyword extracting module includes:

Second word segmentation processing submodule, for carrying out word segmentation processing to described second work order data, it is thus achieved that one Individual or multiple second participles；

Part of speech identification submodule, for identifying the part of speech of the one or more the second participle；

Second participle extracts submodule, is used for according to described part of speech from the one or more second participle Extract key word.

Preferably, described keyword extracting module also includes:

Second matched sub-block, for using the one or more first participle to disable dictionary preset In mate；

Second removes submodule, for removing the second participle that the match is successful.

Preferably, described detection path searching module includes:

Root node matched sub-block, in described fault spectrum, searches the root with described Keywords matching Node；

Leaf node traversal submodule, the one or more leaves joint being connected with described root node for traversal Point, it is thus achieved that one or more detection paths.

Preferably, described detection module includes:

Detection acquisition of information submodule, for for each detection path, obtains in described detection path The detection information that one or more leaf nodes characterize；

Couple candidate detection result obtains submodule, carries out for the detection information characterized according to current leaf node Detection, it is thus achieved that couple candidate detection result；

Leaf node searches submodule, for searching under logical relation mates with described couple candidate detection result One leaf node, returns and calls couple candidate detection result acquisition submodule, until performing to final leaf joint Point；

Testing result arranges submodule, for the couple candidate detection result of final leaf node is set to inspection Survey result.

Preferably, described fault spectrum generates with lower module by calling:

The embodiment of the present application includes advantages below:

The embodiment of the present application is by setting up fault spectrum so that support during subsequent detection concurrently according to this or Multiple detection paths are detected, and decrease and detect time-consuming, that raising detects efficiency, meanwhile, and application event The detection of barrier spectrum is simple to operate, greatly reduces the frequency of artificial participation, reduces the consuming of user's energy, Meanwhile, utilize the knowledge point handling failure in the knowledge base that the work order data of magnanimity are formed, be substantially reduced Technical threshold, facilitates the weak user of technology foundation or customer service solving problems by themselves.

Accompanying drawing explanation

Fig. 1 is the flow chart of steps of the generation embodiment of the method for a kind of fault spectrum of the application；

Fig. 2 A and Fig. 2 B is the pruning exemplary plot of a kind of fault spectrum model of the application；

Fig. 3 A and Fig. 3 B is the pruning exemplary plot of a kind of fault spectrum model of the application；

Fig. 4 is the flow chart of steps of a kind of based on fault spectrum the detection method embodiment of the application；

Fig. 5 is a kind of exemplary plot detecting path of the application；

Fig. 6 A is existing a kind of detection example figure；

Fig. 6 B is a kind of detection example figure of the application；

Fig. 7 is the structured flowchart of the generating means embodiment of a kind of fault spectrum of the application；

Fig. 8 is the structured flowchart of a kind of based on fault spectrum the detection device embodiment of the application.

Detailed description of the invention

Understandable, below in conjunction with the accompanying drawings for enabling the above-mentioned purpose of the application, feature and advantage to become apparent from With detailed description of the invention, the application is described in further detail.

With reference to Fig. 1, it is shown that the flow chart of steps of the generation embodiment of the method for a kind of fault spectrum of the application, Specifically may include steps of:

Step 101, obtains the first work order data of one or more classification；

In actual applications, the first work order data of magnanimity in history can be stored, to this magnanimity First work order data analysis sum up after, typical first work order data are write as knowledge point, are saved in and know Know in storehouse.

It is said that in general, each first work order data may include that date, ID, product, problem The key elements such as classification, problem (fault message), solution (detection information), communication record.

Wherein, fault message can be by recording the information of the fault occurred, and detection information can be to record How to carry out detection and solve the information of this fault, both are corresponding.

Such as, in certain work order data, fault message is that " DB (Database, data base) accesses Slowly ", detection information is " woulding you please first detect network congestion ".

The first sufficient amount of, to belong to a classification together work order data can be extracted by Question Classification, make For training sample.

Step 102, for every class the first work order data, extracts public characteristic word from described detection information, As characteristic vector；

Public characteristic word, for word common in the first work order data of part in such, may be used for characterizing The feature of detection information, as the parameter of training sample.

In a preferred embodiment of the present application, step 102 can include following sub-step:

Sub-step S11, carries out word segmentation processing to described detection information, it is thus achieved that one or more first participles；

In implementing, word segmentation processing can be carried out in the following manner:

1, segmenting method based on string matching: refer to the Chinese being analysed to according to certain strategy Entry in the machine dictionary that word string is preset with mates, if finding certain word in dictionary Symbol string, then the match is successful (identifying a word).

2, feature based scanning or the segmenting method of mark cutting: refer to preferential in character string to be analyzed Middle identification and be syncopated as some words with obvious characteristic, using these words as breakpoint, can be by former word Symbol string is divided into less string to enter mechanical Chinese word segmentation again；Or participle and part-of-speech tagging are combined, Utilize abundant grammatical category information that participle decision-making provides help, and in annotation process the most in turn Word segmentation result is tested, adjusts.

3, based on understand segmenting method: refer to by allow computer mould personification distich understanding, Reach to identify the effect of word.Its basic thought carries out syntax, semantic analysis exactly while participle, Utilize syntactic information and semantic information to process Ambiguity.

4, segmenting method based on statistics: the frequency to each combinatorics on words of co-occurrence adjacent in language material Add up, calculate their information that appears alternatively, and the adjacent co-occurrence calculating two Chinese characters X, Y is general Rate.The information of appearing alternatively can embody the tightness degree of marriage relation between Chinese character.When tightness degree is higher than During some threshold value, just it is believed that this word group may constitute a word.

Certainly, above-mentioned word segmentation processing mode is intended only as example, when implementing the embodiment of the present application, Can arrange other word segmentation processing modes according to practical situation, this is not limited by the embodiment of the present application System.It addition, in addition to above-mentioned word segmentation processing mode, those skilled in the art can also be according to reality Needing to use other word segmentation processing mode, this is not any limitation as by the embodiment of the present application.

Sub-step S12, adds up the word frequency of the described first participle；

Sub-step S13, calculates the weight of the described first participle by the word frequency of the described first participle；

In actual applications, TF-IDF (term frequency inverse document can be passed through Frequency, a kind of conventional weighting technique prospected for information retrieval and information) calculate first point The weight of word.

Specifically, TF-IDF may be used for assessing a words for a file set or a language material The significance level of a copy of it file in storehouse, the importance of words occurs hereof along with it Number of times is directly proportional increase, but can be inversely proportional to decline along with the frequency that it occurs in corpus simultaneously.

Sub-step S14, according at least part of first participle of described weight extraction as public characteristic word.

If the weight calculating the first participle by TF-IDF, then can extract the front N (N that weight is the highest For positive integer, such as 10) the individual first participle is as public characteristic word.

For generally speaking, fault message and the public characteristic word thereof of each classification can be obtained.

Such as: for the classification of connection failure, the public characteristic word of extraction is as follows:

Intercept, error log, part failure ..., white list；

Report an error, authentication failed ..., password；

Connection failure, access reject ..., port.

In another preferred embodiment of the present application, step 102 can also include following sub-step:

Sub-step S15, uses the one or more first participle to carry out in dictionary in preset disabling Join；

Sub-step S16, removes the first participle that the match is successful.

Disable that can to store the frequency of occurrences in dictionary the highest, but the word that practical significance is the most little, refer mainly to pair Word, function word, modal particle etc., as "Yes", " but " etc..

In the embodiment of the present application, before sub-step S12, the first participle can be filtered off by stop words In insignificant word.

Such as, detection information " would you please first to network congestion detect " can be divided into " asking ", " you ", " first ", " to ", " network congestion ", " detection ", the first participle such as " ", by disabling dictionary, Can remove " asking ", " you ", " first ", " to ", the insignificant word such as " ".

Step 103, for every class the first work order data, learns described fault message and described characteristic vector Between logical relation, it is thus achieved that every class fault spectrum model；

Application the embodiment of the present application, can pre-set training aids, may be used for learning the number of each dimension According to the logical relation of (i.e. fault message, characteristic vector), such as support vector machine (Support Vector Machine, SVM), decision tree (Decision Tree), random forest (Random Forest) etc. Deng, this is not any limitation as by the embodiment of the present application.

Wherein, support vector machine is by nonlinear mapping p, sample space be mapped to one high In dimension or even infinite dimensional feature space (Hilbert space) so that non-thread in original sample space The problem of the linear separability that the problem that property can be divided is converted in feature space.

Random forest, is to set up a forest by random manner, has a lot of decision tree groups inside forest Become, between each decision tree of random forest be do not have related.After obtaining forest, when having one The when that individual new input sample entering, each decision tree in forest is just allowed the most once to sentence Disconnected, look at which kind of (for sorting algorithm) this sample should belong to, then look at which kind of is selected Select at most, just predict that this sample is that class.

Decision tree is on the basis of known various situation probability of happening, asks for only by constituting decision tree The expected value of the present worth probability more than or equal to zero, assessment item risk, it is judged that the decision analysis of its feasibility Method, is a kind of diagram method intuitively using probability analysis.

When training aids is trained, fault spectrum model can be fitted, if error (CP) is less than one Error threshold set in advance (such as 0.001), stops matching, the fault spectrum model such as Fig. 2 A at institute's training Shown in, it is tree structure, including root node and leaf node, has between at least part of leaf node and patrol The relation of collecting.

As shown in Figure 2 A, if " 2.5 ", " 3.1 " they are node (including root node, leaf node), table Levy characteristic vector, the characterization logic relation such as " mmax ＜ 6100 ", " syct ＞=360 ".

Additionally, root node characterizes fault message, leaf node characterizes detection information.

Specifically, as shown in Figure 3A, " A:DB connection failure " is root node, characterizes fault letter Breath, " E: local detection ", " B: network breaks " are leaf node, characterize detection information, with root node There is not logical relation in " A:DB connection failure ".

And the child node that " H: configuration investigation ", " C: repair network " be " B: network is disconnected ", i.e. " B: Network breaks " it is " H: configuration investigation ", the father node of " C: repair network ", " H: configuration investigation ", There is logical relation (not shown on figure) in " C: repair network " and " B: network breaks ".

Step 104, carries out pruning modes to every class fault spectrum model, it is thus achieved that every class fault spectrum.

In actual applications, according to actual demand, fault spectrum model can be carried out pruning modes, it is thus achieved that therefore Barrier spectrum.

Wherein, can include root node and the leaf node being connected in fault spectrum, root node can characterize event Barrier information, leaf node can characterize detection information, can have logic between at least part of leaf node Relation, leaf node can have one or more father node.

The fault spectrum pruned, can be stored in fault spectrum warehouse (data base).

In the one of the application is preferable to carry out, step 104 can include following sub-step:

Sub-step S21, searches identical subtree in described fault spectrum model；

Wherein, described subtree can be the set of one or more leaf node；

Sub-step S22, when finding, is connected to one of them subtree by the father node of identical subtree；

Sub-step S23, in identical subtree, cuts off other the subtree outside the subtree connected.

In the embodiment of the present application, owing to some subtree may have repetition, therefore, it can recurrence inspection has The subtree repeated, after discovery, points to another subtree at a node, the subtree of deletion simultaneously itself, Some leaf node is made to have multiple father node (representing that a phenomenon may be caused by many reasons), (DAG refers to that a directed graph cannot be from certain summit to form class tree structure, i.e. directed acyclic graph This point is returned to) through some limits.

The fault spectrum of class tree structure not tree structure, such as binary tree, in binary tree, some subtree may Other subtrees to repeat, the structure causing binary tree is tediously long, and branch is too much, and logical relation is unintelligible；But There is not the subtree of repetition in DAG, because if being repeated, father node can delete this subtree, and Pointing to other subtrees repeated, less accordingly, with respect to tree structure levels such as binary trees, logic is more clear Clear.

(i.e. repeat) to refer to that leaf node is identical it should be noted that identical, patrolling between leaf node The relation of collecting is identical.

As shown in Figure 3A, " H: configuration investigation ", " F: white list ", " D: detection password " and " J: Detection port " subtree of this four leaf nodes composition repeats, as shown in Figure 3 B, can delete wherein One subtree so that father node " B: network breaks " and " E: the local detection " of this subtree are pointed to same Individual subtree.

In the another kind of the application is preferable to carry out, step 104 can also include following sub-step:

Sub-step S24, carries out pruning modes according to default prune approach to described fault spectrum model.

Generally, training fault spectrum model out may be likely to result in inspection containing deeper level The step surveyed is various.

In the embodiment of the present application, can be by the prune approach preset, such as prune () function, to fault Spectrum model carries out pruning modes, in the range of acceptable detection error, the level of fault spectrum model is subtracted Low, reduce the complexity of fault spectrum model, reduce the step of detection.

Such as, the level of fault spectrum model as shown in Figure 2 A is 6 layers, prunes it by prune approach After, it is thus achieved that 4 layers of fault spectrum model as shown in Figure 2 B.

Sub-step S25, cuts off the illegal leaf node of logical relation from described fault spectrum model.

In the embodiment of the present application, can investigate by manual confirmation or by legal logical relation, cut Remove leaf node illegal to some logical relations, improve accuracy rate.

With reference to Fig. 4, it is shown that the step stream of a kind of based on fault spectrum the detection method embodiment of the application Cheng Tu, specifically may include steps of:

Step 401, when receiving the second work order data, crucial from described second work order extracting data Word；

The embodiment of the present application can be applied in virtual customer service system, and this is virtual overcomes system can use stream Formula processes framework strom in real time, it is ensured that complete detection in minimum delay.

Certainly, except strom, it is also possible to application S4 (Simple Scalable Streaming System), The streamings such as MillWheel, Kinesis process in framework in real time, and this is not any limitation as by the embodiment of the present application.

In implementing, user can submit the second work order to by modes such as browser, independent application Data give virtual customer service system, and virtual customer service system can be removed unrelated interruptions information, purify the second work Forms data.

Generally, these the second work order data allow to comprise a fault message, solves a problem, if using Family proposes a problem, and the information unrelated with this problem may be considered interference information.

Furthermore, under certain conditions, the second work order data can be according to product classification, and currently The unrelated information of product can the second work order data.

Such as, in certain second work order data, user's query: " the UDF function in my SQL Why can not perform？" answer of virtual customer service system: " because the most not opening the power of UDF Limit, therefore your UDF can not perform." additionally, user inquires again: " understand.Another problem: How my daily record downloads？”

In this example, the 2nd problem of user is unrelated with first problem, belongs to interference information.

For the second work order data after filtration interference information, can extract key word, this key word is permissible For embodying the information of the second work order data (i.e. fault message) feature.

In a preferred embodiment of the present application, step 401 can include following sub-step:

Described second work order data are carried out word segmentation processing, it is thus achieved that one or more second by sub-step S31 Participle；

1, segmenting method based on string matching.

2, feature based scanning or the segmenting method of mark cutting.

3, based on the segmenting method understood.

4, segmenting method based on statistics.

Sub-step S32, identifies the part of speech of the one or more the second participle；

Sub-step S33, extracts key word according to described part of speech from the one or more second participle.

In the embodiment of the present application, the second participle can be carried out part of speech analysis, it is thus achieved that each second participle Part of speech, such as noun, verb, adjective, adverbial word, preposition, conjunction, auxiliary word etc..

Wherein it is possible to form key word by noun with verb, noun is determined for destination object, Verb can deduce main semanteme.

Such as, in certain second work order data, user A inquires " the UDF letter either with or without maximum Number ", in this example, verb be " either with or without ", noun is " UDF function ", the i.e. problem of user A (i.e. key word) is to ask, " either with or without UDF function ".

The most such as, in certain second work order data, user B inquires " how debugging UDF function ", In this example, verb is " debugging ", and noun is that " UDF function ", the i.e. problem of user B (i.e. close Keyword) it is to ask, the method for " debugging UDF function ".

In a preferred embodiment of the present application, step 401 can also include following sub-step:

Sub-step S34, uses the one or more first participle to carry out in dictionary in preset disabling Join；

Sub-step S35, removes the second participle that the match is successful.

In the embodiment of the present application, before sub-step S32, the second participle can be filtered off by stop words In insignificant word.

Step 402, searches the fault spectrum that described second work order data generic is corresponding；

Application the embodiment of the present application, can be stored in fault spectrum warehouse (data base) with training in advance fault spectrum In.

In actual applications, the second work order data generic can be searched by text similarity corresponding Fault spectrum, i.e. searches the fault spectrum similar to the key word in the second work order data in fault spectrum warehouse.

Such as, if key word is: " DB ", " inquiry ", " wait ", " slowly " etc., then can be according to literary composition The fault spectrum of " DB accesses slow " this classification on this similarity mode.

In the embodiment of the present application, can include root node and the leaf node being connected in fault spectrum, root saves Point can characterize fault message, and leaf node can characterize detection information, between at least part of leaf node Can have logical relation, leaf node can have one or more father node.

Step 403, in described fault spectrum, according to described keyword lookup one or more detection path；

In implementing, the information of detection mode and detection ordering can be recorded in this detection path.

In a preferred embodiment of the present application, step 403 can include following sub-step:

Sub-step S41, in described fault spectrum, searches the root node with described Keywords matching；

Sub-step S42, one or more leaf nodes that traversal is connected with described root node, it is thus achieved that Or multiple detection path.

In the embodiment of the present application, fault spectrum is a class tree structure (directed acyclic graph DAG), because of This, can carry out using " node " to mate during retrieval detection path under upper cause.

Root node can be positioned in fault spectrum according to key word, be to be passed through from this root node traversal down All leaf nodes, composition detection path.

Such as, fault spectrum as shown in Figure 3 B exists two detection paths, is distributed as " A → E → H → F/D/J " and " A → B → C/H → F/D/J ".

Owing to the leaf node of root node Yu next layer is generally not present logical relation, therefore, this next layer The quantity of leaf node is the most identical with the quantity in detection path.

It should be noted that be connected to refer to that root node is joined directly together with child node, it is also possible to refer to root node It is indirectly connected to child node.

Such as, as shown in Figure 3 B, " A:DB is even with root node for leaf node " E: local detection " Connect failure " it is joined directly together, leaf node " H: configuration investigation " and root node " A:DB connection failure " It is indirectly connected to.

Step 404, detects according to the one or more detection path, it is thus achieved that testing result.

In application embodiment, support concurrently to detect according to these one or more detection paths, reduce Detection is time-consuming, improves the efficiency of detection.

In a preferred embodiment of the present application, step 404 can include following sub-step:

Sub-step S51, for each detection path, obtains the one or more leaves in described detection path The detection information that child node characterizes；

Sub-step S52, the detection information characterized according to current leaf node detects, it is thus achieved that Hou Xuanjian Survey result；

Sub-step S53, searches next leaf node that logical relation is mated with described couple candidate detection result, Return and perform sub-step S51, until performing to final leaf node；

Sub-step S54, is set to testing result by the couple candidate detection result of final leaf node.

In the embodiment of the present application, the detection information that logical relation and leaf node characterize can referred to as be advised Then, i.e. when meet certain condition (logical relation) just do execution certain operation (detection information).

Such as, when DB accesses slow, just check bandwidth and the flow of network, here it is a logic is closed System.

Rule is previously defined in regulation engine (rule engine), such as JBoss Rules (business Regulation engine), once condition triggers, and regulation engine can perform this rule, such as: perform network state Sense command ifstat.

In implementing, in detection path, logically relation successively detects, until final leaf joint Point, can avoid performing not meet the parton node of logical relation.Wherein, final leaf node can To refer to the leaf node not having next layer of leaf node, the not necessarily leaf of the bottom in detection path Node.

Such as, in detection path as shown in Figure 5, root node " A:DB accesses slow " characterizes event Barrier information, key element is network, and DB loads, SQL (Structured Query Language, structuring Query language).

First have to determine that network is the most problematic according to child node " B: network congestion detection ", if network Problematic (i.e. " Y "), remaining key element is all difficult to work and (is i.e. performed without that " C:DB loads inspection Survey "), manually solve according to child node " H: contact net work "；Secondly, if network is not asked According to child node " C:DB load detecting ", topic (i.e. " N "), then judge that DB load is the highest, as Really " DB loads height ", though SQL itself no problem (being i.e. performed without " D: slow SQL detection "), The most also can show the slow situation that accesses, then confirm SQL according to child node " detection of J:SQL thread " Thread whether normal operation；If " DB load is low ", then according to child node " D: slow SQL detection " Detect, be distributed as the detection of leaf node " K: index ", the detection of " M: implement plan " and The detection of " N: lock ".

If go to child node " H: contact net work ", " detection of J:SQL thread ", " K: index ", " M: implement plan " and " N: lock ", then can be with termination detection, it is thus achieved that testing result.

Showing according to a certain data statistics, the work order data of 70% are the most basic FAQs, have bigger Repeatability, processes these problems with existing WorkForm System and need to take the customer service resource of more than 50%, spends A large amount of duplications of labour, efficiency is the lowest, one of purpose of the embodiment of the present application be automatization solve this 70% FAQs.

Existing WorkForm System usually user comes booting problem investigation, the i.e. side of guiding according to the experience of oneself To for " people → problem ", operation complexity, need to expend the substantial amounts of energy of user, and, technical threshold High.

The embodiment of the present application is to allow System guides troubleshooting procedure, is by the fault of the work order data genaration of magnanimity Spectrum instructs user to investigate problem, is inverse guiding, i.e. guide direction is " system → problem → people ", Simple to operate, greatly reduce the frequency of artificial participation, reduce the consuming of user's energy, utilize meanwhile Knowledge point in the knowledge base that the work order data lock of magnanimity is formed processes problem, greatly reduces technology door Sill, facilitate the weak user of technology foundation or customer service solving problems by themselves.

As shown in Figure 6A, existing WorkForm System, when serial is investigated, does not i.e. have any internal logic to close System investigates on ground, needs node all of in Fig. 6 A is carried out out of order investigation one by one, such as: A → B → C → D → E, until pinpointing the problems.Therefore, during time complexity, O (N), N are the quantity of node, The most all of node will be investigated.

In the embodiment of the present application, as shown in Figure 6B, concurrently investigated (as concurrently by fault spectrum Perform B, C) time, because there being logical relation between node, typically need not all of node one One investigation, saves and much investigates step, e.g., if guiding D in C detection, then need not perform E, Otherwise, if guiding E in C detection, then need not perform D.For the structure of fault spectrum, utilize branch Structure (every time divided by 2), can save and much investigate step, and investigation complexity is minimum up to O (log₂ ^N), N is the quantity of node.

In a preferred embodiment of the present application, described fault spectrum can generate in the following manner:

Sub-step S61, obtains the first work order data of one or more classification；Each first work order data Include fault message and detection information；

Sub-step S62, for every class the first work order data, extracts public characteristic from described detection information Word, as characteristic vector；

Sub-step S63, for every class the first work order data, learn described fault message and described feature to Logical relation between amount, it is thus achieved that every class fault spectrum model；

Sub-step S64, carries out pruning modes to every class fault spectrum model, it is thus achieved that every class fault spectrum.

In actual applications, described fault spectrum can include root node and the leaf node being connected, described Root node can characterize fault message, and described leaf node can characterize detection information, at least part of leaf Can have logical relation between node, described leaf node can have one or more father node.

In a preferred embodiment of the present application, sub-step S62 can include following sub-step:

Sub-step S621, carries out word segmentation processing to described detection information, it is thus achieved that one or more first point Word；

Sub-step S622, adds up the word frequency of the described first participle；

Sub-step S623, calculates the weight of the described first participle by the word frequency of the described first participle；

Sub-step S624, according at least part of first participle of described weight extraction as public characteristic word.

In another preferred embodiment of the present application, sub-step S62 can also include following sub-step:

Sub-step S625, uses the one or more first participle to carry out in preset disabling in dictionary Coupling；

Sub-step S626, removes the first participle that the match is successful.

In a preferred embodiment of the present application, sub-step S64 can include following sub-step:

Sub-step S641, searches identical subtree in described fault spectrum model；Described subtree is one Or the set of multiple leaf node；

Sub-step S642, when finding, is connected to one of them son by the father node of identical subtree Tree；

Sub-step S643, in identical subtree, cuts off other the subtree outside the subtree connected.

In another preferred embodiment of the present application, sub-step S64 can also include following sub-step:

Sub-step S644, carries out pruning modes according to default prune approach to described fault spectrum model.

Sub-step S645, cuts off the illegal leaf node of logical relation from described fault spectrum model.

It should be noted that for embodiment of the method, in order to be briefly described, therefore it is all expressed as one it be The combination of actions of row, but those skilled in the art should know, and the embodiment of the present application is not by described The restriction of sequence of movement because according to the embodiment of the present application, some step can use other orders or Person is carried out simultaneously.Secondly, those skilled in the art also should know, embodiment described in this description Belong to preferred embodiment, necessary to involved action not necessarily the embodiment of the present application.

With reference to Fig. 7, it is shown that the structured flowchart of the generating means embodiment of a kind of fault spectrum of the application, Specifically can include such as lower module:

Work order data acquisition module 701, for obtaining the first work order data of one or more classification；Often Individual first work order data include fault message and detection information；

Public characteristic word extraction module 702, for for every class the first work order data, from described detection letter Breath extracts public characteristic word, as characteristic vector；

Fault spectrum model learning module 703, for for every class the first work order data, learns described fault Logical relation between information and described characteristic vector, it is thus achieved that every class fault spectrum model；

Module 704 pruned by fault spectrum model, for every class fault spectrum model is carried out pruning modes, it is thus achieved that Every class fault spectrum.

In implementing, described fault spectrum can include root node and the leaf node being connected, described Root node can characterize fault message, and described leaf node can characterize detection information, at least part of leaf Can have logical relation between node, described leaf node can have one or more father node.

In a preferred embodiment of the present application, described public characteristic word extraction module 702 can include Following submodule:

In a preferred embodiment of the present application, described public characteristic word extraction module 702 can also wrap Include following submodule:

Matched sub-block, for using the one or more first participle to disable in dictionary preset Row coupling；

Remove submodule, for removing the first participle that the match is successful.

In a preferred embodiment of the present application, described fault spectrum model is pruned module 704 and can be included Following submodule:

In a preferred embodiment of the present application, described fault spectrum model is pruned module 704 and can also be wrapped Include following submodule:

With reference to Fig. 8, it is shown that the structural frames of a kind of based on fault spectrum the detection device embodiment of the application Figure, specifically can include such as lower module:

Keyword extracting module 801, for when receiving the second work order data, from described second work order Extracting data key word；

Fault spectrum searches module 802, for searching the fault that described second work order data generic is corresponding Spectrum；

Detection path searching module 803, in described fault spectrum, according to described keyword lookup one Individual or multiple detection paths；

Detection module 804, for detecting according to the one or more detection path, it is thus achieved that detection Result.

In a preferred embodiment of the present application, described keyword extracting module 801 can include as follows Submodule:

In a preferred embodiment of the present application, described keyword extracting module 801 can also include as Lower submodule:

In a preferred embodiment of the present application, described detection path searching module 803 can include as Lower submodule:

In a preferred embodiment of the present application, described detection module 804 can include following submodule:

In a preferred embodiment of the present application, described fault spectrum can be raw by calling with lower module Become:

In a preferred embodiment of the present application, described public characteristic word extraction module can also include as Lower submodule:

In a preferred embodiment of the present application, described fault spectrum model is pruned module and can be included as follows Submodule:

In a preferred embodiment of the present application, described fault spectrum model prune module can also include as Lower submodule:

For device embodiment, due to itself and embodiment of the method basic simlarity, so the comparison described Simply, relevant part sees the part of embodiment of the method and illustrates.

Each embodiment in this specification all uses the mode gone forward one by one to describe, and each embodiment stresses Be all the difference with other embodiments, between each embodiment, identical similar part sees mutually ?.

Those skilled in the art are it should be appreciated that the embodiment of the embodiment of the present application can be provided as method, dress Put or computer program.Therefore, the embodiment of the present application can use complete hardware embodiment, completely Software implementation or the form of the embodiment in terms of combining software and hardware.And, the embodiment of the present application Can use and can be situated between with storage at one or more computers wherein including computer usable program code The upper computer journey implemented of matter (including but not limited to disk memory, CD-ROM, optical memory etc.) The form of sequence product.

In a typical configuration, described computer equipment includes one or more processor (CPU), input/output interface, network interface and internal memory.Internal memory potentially includes computer-readable medium In volatile memory, the shape such as random access memory (RAM) and/or Nonvolatile memory Formula, such as read only memory (ROM) or flash memory (flash RAM).Internal memory is computer-readable medium Example.Computer-readable medium includes removable media permanent and non-permanent, removable and non- Information storage can be realized by any method or technology.Information can be computer-readable instruction, Data structure, the module of program or other data.The example of the storage medium of computer includes, but Be not limited to phase transition internal memory (PRAM), static RAM (SRAM), dynamic random are deposited Access to memory (DRAM), other kinds of random access memory (RAM), read only memory (ROM), Electrically Erasable Read Only Memory (EEPROM), fast flash memory bank or other in Deposit technology, read-only optical disc read only memory (CD-ROM), digital versatile disc (DVD) or other Optical storage, magnetic cassette tape, tape magnetic rigid disk storage other magnetic storage apparatus or any its His non-transmission medium, can be used for the information that storage can be accessed by a computing device.According to herein Defining, computer-readable medium does not include the computer readable media (transitory media) of non-standing, Data signal and carrier wave such as modulation.

The embodiment of the present application is with reference to the method according to the embodiment of the present application, terminal unit (system) and meter The flow chart of calculation machine program product and/or block diagram describe.It should be understood that can be by computer program instructions Each flow process in flowchart and/or block diagram and/or square frame and flow chart and/or square frame Flow process in figure and/or the combination of square frame.Can provide these computer program instructions to general purpose computer, The processor of special-purpose computer, Embedded Processor or other programmable data processing terminal equipment is to produce One machine so that performed by the processor of computer or other programmable data processing terminal equipment Instruction produce for realizing at one flow process of flow chart or multiple flow process and/or one square frame of block diagram or The device of the function specified in multiple square frames.

These computer program instructions may be alternatively stored in and computer or other programmable datas can be guided to process In the computer-readable memory that terminal unit works in a specific way so that be stored in this computer-readable Instruction in memorizer produces the manufacture including command device, and this command device realizes flow chart one The function specified in flow process or multiple flow process and/or one square frame of block diagram or multiple square frame.

These computer program instructions also can be loaded into computer or other programmable data processing terminals set Standby upper so that on computer or other programmable terminal equipment, to perform sequence of operations step in terms of producing The process that calculation machine realizes, thus the instruction performed on computer or other programmable terminal equipment provides and uses In realizing in one flow process of flow chart or multiple flow process and/or one square frame of block diagram or multiple square frame The step of the function specified.

Although having been described for the preferred embodiment of the embodiment of the present application, but those skilled in the art being once Know basic creative concept, then these embodiments can be made other change and amendment.So, Claims are intended to be construed to include preferred embodiment and fall into the institute of the embodiment of the present application scope There are change and amendment.

Finally, in addition it is also necessary to explanation, in this article, the relational terms of such as first and second or the like It is used merely to separate an entity or operation with another entity or operating space, and not necessarily requires Or imply relation or the order that there is any this reality between these entities or operation.And, art Language " includes ", " comprising " or its any other variant are intended to comprising of nonexcludability, so that Process, method, article or terminal unit including a series of key elements not only include those key elements, and Also include other key elements being not expressly set out, or also include for this process, method, article or The key element that person's terminal unit is intrinsic.In the case of there is no more restriction, statement " include one It is individual ... " key element that limits, it is not excluded that including the process of described key element, method, article or end End equipment there is also other identical element.

Generation method, a kind of detection based on fault spectrum to a kind of fault spectrum provided herein above Method, the generating means of a kind of fault spectrum and a kind of detection device based on fault spectrum, carried out detailed Jie Continuing, principle and the embodiment of the application are set forth by specific case used herein, above reality Execute the explanation of example to be only intended to help and understand the present processes and core concept thereof；Simultaneously for ability The those skilled in the art in territory, according to the thought of the application, the most all can Change part, and in sum, this specification content should not be construed as the restriction to the application.

Claims

1. the generation method of a fault spectrum, it is characterised in that including:

Method the most according to claim 1, it is characterised in that described fault spectrum includes being connected Root node and leaf node, described root node characterize fault message, described leaf node characterize detection letter Breath, has logical relation between at least part of leaf node, described leaf node has one or more father Node.

Method the most according to claim 1 and 2, it is characterised in that described from described detection letter The step extracting public characteristic word in breath includes:

Add up the word frequency of the described first participle；

Method the most according to claim 3, it is characterised in that described from described detection information The step extracting public characteristic word also includes:

Remove the first participle that the match is successful.

5. according to the method described in claim 1 or 2 or 4, it is characterised in that described to the event of every class Barrier spectrum model carries out the step of pruning modes and includes:

In identical subtree, cut off other the subtree outside the subtree connected.

Method the most according to claim 5, it is characterised in that described to every class fault spectrum model The step carrying out pruning modes also includes:

8. a detection method based on fault spectrum, it is characterised in that including:

Method the most according to claim 8, it is characterised in that described fault spectrum includes being connected Root node and leaf node, described root node characterize fault message, described leaf node characterize detection letter Breath, has logical relation between at least part of leaf node, described leaf node has one or more father Node.

Method the most according to claim 8, it is characterised in that described from described second work order number Include according to the step of middle extraction key word:

Identify the part of speech of the one or more the second participle；

11. methods according to claim 10, it is characterised in that described from described second work order The step of extracting data key word also includes:

Remove the second participle that the match is successful.

12. methods according to claim 9, it is characterised in that described in described fault spectrum, The step searching one or more detection paths according to described Feature Words includes:

13. methods according to claim 12, it is characterised in that described according to one or Multiple detection paths are detected, it is thus achieved that the step of testing result includes:

Method described in 14. according to Claim 8 or 9 or 10 or 11 or 12 or 13, its feature exists In, described fault spectrum generates in the following manner:

The generating means of 15. 1 kinds of fault spectrums, it is characterised in that including:

16. devices according to claim 15, it is characterised in that described fault spectrum includes phase Root node even and leaf node, described root node characterizes fault message, and described leaf node characterizes detection Information, has logical relation between at least part of leaf node, described leaf node has one or more Father node.

17. according to the device described in claim 15 or 16, it is characterised in that described public characteristic word Extraction module includes:

18. devices according to claim 17, it is characterised in that described public characteristic word extracts Module also includes:

19. according to the device described in claim 15 or 16 or 18, it is characterised in that described fault Spectrum model is pruned module and is included:

20. devices according to claim 19, it is characterised in that described fault spectrum model is pruned Module also includes:

21. devices according to claim 19, it is characterised in that described fault spectrum model is pruned Module also includes:

22. 1 kinds of detection devices based on fault spectrum, it is characterised in that including:

23. devices according to claim 22, it is characterised in that described fault spectrum includes phase Root node even and leaf node, described root node characterizes fault message, and described leaf node characterizes detection Information, has logical relation between at least part of leaf node, described leaf node has one or more Father node.

24. devices according to claim 22, it is characterised in that described keyword extracting module Including:

25. devices according to claim 24, it is characterised in that described keyword extracting module Also include:

26. devices according to claim 23, it is characterised in that described detection path searching mould Block includes:

27. devices according to claim 25, it is characterised in that described detection module includes:

28. according to the device described in claim 22 or 23 or 24 or 25 or 26 or 27, its feature Being, described fault spectrum generates with lower module by calling: