CA2129510C - Automatic temporospatial pattern analysis and prediction in a telecommunications network using rule induction - Google Patents
Automatic temporospatial pattern analysis and prediction in a telecommunications network using rule inductionInfo
- Publication number
- CA2129510C CA2129510C CA002129510A CA2129510A CA2129510C CA 2129510 C CA2129510 C CA 2129510C CA 002129510 A CA002129510 A CA 002129510A CA 2129510 A CA2129510 A CA 2129510A CA 2129510 C CA2129510 C CA 2129510C
- Authority
- CA
- Canada
- Prior art keywords
- rule
- set forth
- particular data
- telecommunications system
- rule set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M3/00—Automatic or semi-automatic exchanges
- H04M3/22—Arrangements for supervision, monitoring or testing
- H04M3/36—Statistical metering, e.g. recording occasions when traffic exceeds capacity of trunks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M3/00—Automatic or semi-automatic exchanges
- H04M3/22—Arrangements for supervision, monitoring or testing
- H04M3/2254—Arrangements for supervision, monitoring or testing in networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04Q—SELECTING
- H04Q3/00—Selecting arrangements
- H04Q3/0016—Arrangements providing connection between exchanges
- H04Q3/0062—Provisions for network management
- H04Q3/0075—Fault management techniques
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04Q—SELECTING
- H04Q3/00—Selecting arrangements
- H04Q3/0016—Arrangements providing connection between exchanges
- H04Q3/0062—Provisions for network management
- H04Q3/0075—Fault management techniques
- H04Q3/0079—Fault management techniques involving restoration of networks, e.g. disaster recovery, self-healing networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04Q—SELECTING
- H04Q2213/00—Indexing scheme relating to selecting arrangements in general and for multiplex systems
- H04Q2213/13521—Indexing scheme relating to selecting arrangements in general and for multiplex systems fault management
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04Q—SELECTING
- H04Q2213/00—Indexing scheme relating to selecting arrangements in general and for multiplex systems
- H04Q2213/13544—Indexing scheme relating to selecting arrangements in general and for multiplex systems modeling or simulation, particularly of networks
Abstract
Automatic Temporospatial Pattern Analysis And Prediction In A Telecommunications Network Using Rule Induction Abstract A facility is provided for enhancing an operations support system so that, based on data generated as a result of an event occurring in an associatedtelecommunications network, the operations support system can predict the likelihood of the event occurring again in the network.
Description
2~
- 1 ~
ol- ~t;c Tc.~ ,atial Pattern Analysis And I'~"-qc~i~n In A T~lc~ t;on~ Network Using Rule Ind~
Field of the Inv~l~t;~
The invendon relates to IJluCeS~il g data gC-Ihaled by a 5 te1~ cc,-~ -;r~tinn~ network and applying the result of such ~l~f.,~ g to other network data ~o pre~ict the pos~ y of a pard ular event h ~p~ ~li ng in the network.
Back~round of the I~ ..t;~
A large tf 1eco..-...~ ;rln~ network typically Eene~Ps a large number of reports over a short period of time, e.g., twenty-four hours. Such reports include 10 data relating to, for ey~3pl~ telephrJne traffic, circuit alarrns, etc. A Network n~ge...f n~ System ~NMS) ~soci~ d with the telPçomm--n;c~tion~ network ~rCum~ t~s and ploces~s the data to ~ack and manage the p~r - .-~ ce of the network. Such m~n~gem~.nt includes (a) identifying and repailing network problems when they occur, and (b) trends in telephone traffic pa~terns. For eY~npte~ consicler 15 the way in which an NMS processes trouble alarms, which may be c~ ifi~l as either ll, --s'~nt or non~ ;Gnt Non-lsansi~ alarms are usually indicative of a "hard" fault. T~ ansient ala~s, on the o~her hand, are usually indicative of either a ~,OCCU~ g and nonl~occ~ g network problem. A no~ oc..~ g alarm is usually the result of aln in~o.~ f ~~ 1 fault and, ~ .5fo~ may be ignored. The 20 reason for this is that a nonreocc~lrrin~ alarm is most likely the result of some exte~nal con!li tion~ such as, ~or e I ' ~, a s~ce ~ lightnin~, causing 2 m~m~n~loss of a carrier signal at a point in the network. A reoccluSrin~ l",nc~ t alanT~, on the other hand, is usually indica~ive of a chronic problem, for example, a network facility or circuit tha~ fails when particular conflition~ occlar. Such a problem might 2S affect dle ~quali~r of ~ ui-ic~tionc when it ~curs.
Ln most i~ t~ s~ ho~e~,r, it is difficult to rlictingui~h a chronic alan~
from a ~ si~ , non~hronic alarm. Moreover, the Illajc,lily o~ alanns ~a~ are gvn~al~d in a tel~co~ n~ network are of the W.l.~C"., ;.~ ~ype. Because ~ this, an NM~ may spend an inordinate amounlt of ~in~ and n~ vs aLt~,m~lillg to30 identi3~y the sources ~Ic~lJec-liYe no~ ;cul~in~ ala~ns, ~ereby possibly delaying the rep~ of chronic ~luble~Js.
Summary of the In~ention An ~l~n~e..)~,n~ in ~he art of teleco...~ ~1On~ is achiewd by arran~g an operations supp~rt system so that, in a~cord with an aspeet of the 35 invention, it gen. l~t~s a rule set firom data generated by an ~
tclPcol,..l~ ir~ions system and ~hen applies the rule se~ to da~a subseq ~~ntly . . = , . ~ ~
.. . .. . . ..
.
: "
, .~ , CA 02129~10 1998-02-2~
generated by the telecommunicationssystem to predict that a particular event(s) will Iikely occur in the system. In an illustrative embodiment of the invention, a set of predetermined measurements is generated for the data that is initially obtained from the telecommunications system and then supplied to a particular rule induction method which 5 then generates a number of rules as a function of such measurements and selects a best rule set from the generated rules such that the selected rule set is capable of predicting accurately the occurrence of past events characterized by the data. The rule set is then used to predict future occurrences of the events.
In accordance with one aspect of the present invention there is provided a 10 method of processing particular data generated by a telecommunications system, said particular data being indicative of different types of events that occurred in said telecommunications system, said method comprising the steps of: obtaining said particular data from said telecommunications system, segmenting said particular data into a plurality of groups, processing said particular data contained within at least a first 15 group of said plurality of groups and automatically generating values for respective predetermined measurements, independently generating at least one rule set as a function of said predetermined measurements, and processing said particular data in said first group using said at least one rule set to identify said different types of events and to predict measurements in at least a second group.
20 Brief Description of the D~
FIG. 1 shows a broad block diagram of an operations system;
FIG. 2 shows in flow chart form the program which implements the invention in the rule generator of FIG. 1; and FIG. 3 identifies di~lc;l-l types of performance measurements used in the 25 system of FIG. 1.
Detailed Description Turning now to FIG. 1 there is shown a broad block diagram of a conventional telecommunicationsnetwork 200 and operations support system 300. Operations support system 300, more particularly, supports the overall operation of network 200. One 30 aspect of such support deals with tracking the performance of network 200 by accumulating and processing alarm messages generated by individual facilities comprising network 200. An alarm message may be generated by a facility as a result of one of a number of different conditions. Such conditions include (a) a spurious problem CA 02129~10 1998-02-2~
that occurs for just a short period of time, (b) a chronic problem that occurs periodically, but which can cause an associated facility to slowly degrade over time, (c) a "hard"
failure which may cause the associated facility to become inoperable, etc. The number of such alarms that net~vork 200 may generate within a particular period of time, e.g., 24 5 hours, could be very high. Such alarms are accumulated by system 300 and stored in database 30. An alarm message includes information identifying a particular problem and the identity of the network 200 facilities that might be affected by the problem.
The alarm messages that are stored in d~t~b~ce 30 may be accessed by facility type. For example, all alarm messages associated with a particular type of facility, such 10 as echo cancelers, may be accessed by presenting a request identifying the particular type of facility type to database 30. Such access may be refined by also specifying aparticular period of time in the request. In response to the request, database 30 outputs an alarm message that occurred over the specified period of time for the identified type of facility.
One dilemma that an operations support system faces is which of the large number of alarms that are generated on a daily basis should be addressed first. It is appa~ l that alarms that are generated as a result of a "hard" fault are usually easy to identify and are addressed first. However, the majority of alarms that are generated daily are typically due to spurious conditions and chronic problems. Moreover, it is difficult to determine initially whether an alarm occurred as a result of a spurious condition or a chronic problem. As such, much effort is sometimes directed to tracking down the cause of spurious alarms since they comprise the majority of alarms that occur each day.
Accordingly, alarms identifying a chronic problem may not be dealt with until the facility generating such alarms degrades to a hard fault.
Recognizing that problem, we have adapted an operations support system so that it processes alarms to identify chronic problems and prioritize the repair of those problems. To that end, Operations Support System (OSS) 300 includes rule generator 40 and operations system processor 50 which operate in accordance with the invention to identify and prioritize chronic problems. More particularly, rule generator 40, which may be, for example, embodied in a SPARC station 2 available from Sun Microsystems, Inc., responds to receipt of a command from an external source, e.g., terminal T2, by sending an access request identifying a facility type and period of time to d~t~b~ee 30.
As mentioned above, d~t~b~ee 30, in response to such a request, outputs all alarm CA 02129~10 1998-02-2~
messages that occurred over the specified period of time for the type of facility identified - in the request, in which the facility type may be, for example, echo cancelers, Tl carriers, T3 carriers, toll switches, etc. The specified period of time is typically one to several weeks and may be identified by a starting date and an ending date.
Assuming that the requested period of time equals two weeks and the facility type is Tl carriers, then the number of alarm messages of the requested type that database 30 outputs could be very large if network 200 employs an appreciable number of Tl carriers. Upon receipt of the alarm messages, rule generator 40 stores them in its associated memory 41 such that alar n messages associated with a specific facility are 10 stored in a respective memory array. When all such messages have been received from database 30 and stored in memory 41, then rule generator 40 processes the messages and generates, in accord with an aspect of the invention, a set of rules which may be applied to identify which of the alarms stored in memory 41 are indicative of respective chronic problems. The rules are then passed to operations system processor 50, which then 15 applies the rules to alarrn messages associated with the same type of facility, but occurring thereafter. Such rules are also passed to processor 60, as discussed below.
The program which controls the operation of rule generator 40 to generate a particular set of rules for the processing of alarm messages is shown in flow chart form in FIG. 2. Specifically, the program is entered at block 2000 responsive to receipt of the aforementioned command. At block 2000, the program proceeds to block 2001 where it forms a request message cont~ining the identity of the facility type and particular period of time, as discussed above. The program then supplies the message to database 30. As mentioned above, database 30, in turn, (a) unloads the alarm messages from its associated memory for the specified facility type and for the specified period of time and (b) passes them to rule generator 40. Rule generator 40, in turn and under control of the program, stores the messages in associated memory 41, as mentioned above. When such alarm messages have been so stored, the program then proceeds to block 2002.
(hereinafter the term "alarm" may also be referred to as an event.) At block 2002, the program processes the stored messages to obtain a number of di~~ l performance measurements for each facility of the specified type. The measurements sllmm~ri~e spatial and temporal information about a facility and associated problems. Such measurements may include the number of (a) different types of alarm messages that occurred for the specific facility, (b) times an alarm (event) occurred CA 02129~10 1998-02-2~
during a unit of time for the facility, (c) units of time or (d) the location of the facility, during which the alarm (event) occurred. As will be discussed below, such measurements are determined for first and second windows, Wl and W2, comprising the aforementioned specified period of time. In addition, each window is partitioned into 5 units of time for the purpose of performing such measurements. (An illustrative list of such performance measurements, or features, is shown in Fig. 3.) Specifically, the program blocks 2002, 2003 and 2004 constitute a "do" loop for processing alarms associated with the identified facility. Once it identifies such messages, then the program generates the di~~ l performance measurements for each 10 window, W1 and W2, and for the units of time associated with those windows. In addition, the program ~ccesses database 20 to determine the circuit layout for the facility within network 200. The circuit layout includes the identities of the other network 200 elements that may be connected directly to the facility. The program then performs the same processing steps for a next facility of the specified type. The program exits the do 15 loop and proceeds to block 2005 upon processing the alarm messages for all of the facilities of the specified type.
At block 2005 the program builds a data file, in which a record of the data filecomprises a number of fields for entering respective performance measurements derived for a specific facility and the circuit layout for the facility. The data file also includes a 20 header which identifies the fields forming a record and identifies the contents of the fields. The program proceeds to block 2006 once it has completed the building of the data file. At block 2006, the program enters a rule generation program, which processes the data file using rule induction methods to generate a set of rules for the specified type of facility. For example, a set of rules of the form "if A is > B, then choose class 1 ", 25 where A may represent a performance measurement, B may represent a particular value and class 1 may represent a chronic problem.
(It is noted that the rule induction program represented by block 2006 follows the rule induction technique disclosed in an article entitled "Reduced Complexity Rule Induction", by S. Weiss and N. Indurkhya and published in the Proceedings of the30 International Joint Conference on Artificial Intelligence (IJCAI), Sydney, Australia, 1991, pages 678-684. Alternatively, a number of rule induction programs are commercially available, such as the C4.5 program available from Morgan EC~ nn Publishers, SanMateo, California, U.S.A.).
CA 02129~10 1998-02-2~
Briefly, rule induction methods attempt to find a compact rule set that completely covers and separates various classes of data, e.g., alarm messages respectively relating to spurious and chronic problems. A covering set of rules is found by heuristically searching for a single best rule that covers dirre~ l cases, e.g., records in 5 the data file, for only one class. Once a best conjunctive rule is found for a particular class, e.g., chronic or nonchronic problems, the rule is then added to the rule set and the cases satisfying it are removed from further consideration. This process is repeated until there are no rem~ining cases to be covered. Once a covering rule set is found that separates the classes, the induced set of rules is further refined by using either pruning or 10 statistical techniques. The initial covering rule set is then scaled back to the most statistically accurate subset of rules using training and testing evaluation methods.
When the program has generated a number of sets of rules, it applies those rule sets to the test data portion of the data file generated at block 2005. In doing so, it prunes the rule sets so that an error rate defining the accuracy of each rule set may be 15 minimi7ed when applied to the test data. That is, the program strives to improve the accuracy of a rule set to predict the occurrence of chronic as well as nonchronic alarms.
Following the foregoing, the program then proceeds to block 2007.
The program at block 2007 selects a rule set from the sets of rules generated atblock 2006, in which the selection is based on a specification provided by an external 20 source, for example, a user or another program. Such a specification may request a rule set having a particular predictive accuracy for a certain class of problems, such as chronic problems. The specification may also include coverage for a particular class of problems. Coverage as defined herein is a measure of the range of particular problems, e.g., chronic problems, that a rule set can predict accurately. For example, the following 25 illustrates one possible rule set:
If (ALRD >x, FBL > x2 SLL > X3 OFL>x4or BVL > X5) then choose chronic;
CA 02129~10 1998-02-2~
where x, through X5 are particular values that are generated by the rule generator. The coverage of the above rule set may be increased by adding, for example, the following rule to form a rule set of 2.
If (CT2 & ALRCT > x6 & ALRH > x,) then choose chronic;
5 where x6 and X7 are also values generated by the rule generator.
It is noted, however, that increasing the coverage of a rule set may lead to a decrease in the predictive accuracy of the rule set. The reason for this is that as the number of cases that are covered increases it is likely that errors will occur.
When the program at block 2007 selects a rule set in accordance with the 10 externally provided specification, it then proceeds to block 2008, where it outputs the rule set to processor 50 and processor 60. The program then proceeds to block 2009 where it checks to see it is done, that is, if the program has been directed to generate a rule set for another type of facility. If that is the case, then the program proceeds to block 2010 where it obtains the identity for the specified facility and then returns to 15 block 2001. Otherwise, the program exits.
Returning to FIG. 1, operations system processor 50 applies the rule set to new alarm messages that it obtains from communications network 200 so that it can accurately predict the occurrence of a particular chronic problem and output a message to that effect via path 51. In this way, processor 50 may direct attention away from 20 spurious problems (nonchronic problems) and focus attention on chronic problems so that such problems may be dealt with before they degrade into a hard fault.
The rule set outputted by rule generator 40 may also be used to predict trends in the occurrence of chronic problems. Quality control processor 60 uses the rule set to determine such trends, and uses the trends in conjunction with other quality control 25 criteria, e.g., the rate at which chronic problems are being repaired, to evaluate the performance of network 200. For example, if a trend indicates that a particular section of network 200 may exceed its number of allowable chronic problems for a given period of time, then processor 60 uses that trend to output a performance value for that section of network 200.
CA 02129~10 1998-02-2~
As mentioned above, rule induction can be applied to other applications as a way of generating rule sets for those applications. For example, one application may be traffic data in which rule induction may be applied to such data to generate a set of rules that may be used to predict particular tends in future traff1c patterns. In particular, traffic 5 data may be collected over a period of time and stored in a database, e.g., database 30.
The above-described program at block 2001 (FIG. 2) may then access that data as well as data indicative of the topology of the network in the manner discussed above. The program may then divide the traffic data into respective windows Wl and W2, and then generate a rule set for the data. The measurements made in Wl are used to predict the 10 measurements in W2. If the prediction is accurate, then the rule set may be used to predict future traffic patterns, in which such patterns may be, for example, predicting applopl;ate levels of blocking for particular telephone routes.
As another example, when a disaster strikes a particular area, e.g., a hurricane, particular controls are activated to automatically block low priority calls so that such 15 calls do not impact the processing and routing of high priority calls. The levels of such controls can be predicted accurately by processing traffic data accumulated during a previous disaster, in accord with the invention, to generate a rule set that can be used to predict such levels of control.
The foregoing is merely iliustrative of the principles of the invention. Those 20 skilled in the art will be able to devise numerous arrangements, which, although not explicitly shown or described herein, nevertheless embody those principles that are within the spirit and scope of the invention.
- 1 ~
ol- ~t;c Tc.~ ,atial Pattern Analysis And I'~"-qc~i~n In A T~lc~ t;on~ Network Using Rule Ind~
Field of the Inv~l~t;~
The invendon relates to IJluCeS~il g data gC-Ihaled by a 5 te1~ cc,-~ -;r~tinn~ network and applying the result of such ~l~f.,~ g to other network data ~o pre~ict the pos~ y of a pard ular event h ~p~ ~li ng in the network.
Back~round of the I~ ..t;~
A large tf 1eco..-...~ ;rln~ network typically Eene~Ps a large number of reports over a short period of time, e.g., twenty-four hours. Such reports include 10 data relating to, for ey~3pl~ telephrJne traffic, circuit alarrns, etc. A Network n~ge...f n~ System ~NMS) ~soci~ d with the telPçomm--n;c~tion~ network ~rCum~ t~s and ploces~s the data to ~ack and manage the p~r - .-~ ce of the network. Such m~n~gem~.nt includes (a) identifying and repailing network problems when they occur, and (b) trends in telephone traffic pa~terns. For eY~npte~ consicler 15 the way in which an NMS processes trouble alarms, which may be c~ ifi~l as either ll, --s'~nt or non~ ;Gnt Non-lsansi~ alarms are usually indicative of a "hard" fault. T~ ansient ala~s, on the o~her hand, are usually indicative of either a ~,OCCU~ g and nonl~occ~ g network problem. A no~ oc..~ g alarm is usually the result of aln in~o.~ f ~~ 1 fault and, ~ .5fo~ may be ignored. The 20 reason for this is that a nonreocc~lrrin~ alarm is most likely the result of some exte~nal con!li tion~ such as, ~or e I ' ~, a s~ce ~ lightnin~, causing 2 m~m~n~loss of a carrier signal at a point in the network. A reoccluSrin~ l",nc~ t alanT~, on the other hand, is usually indica~ive of a chronic problem, for example, a network facility or circuit tha~ fails when particular conflition~ occlar. Such a problem might 2S affect dle ~quali~r of ~ ui-ic~tionc when it ~curs.
Ln most i~ t~ s~ ho~e~,r, it is difficult to rlictingui~h a chronic alan~
from a ~ si~ , non~hronic alarm. Moreover, the Illajc,lily o~ alanns ~a~ are gvn~al~d in a tel~co~ n~ network are of the W.l.~C"., ;.~ ~ype. Because ~ this, an NM~ may spend an inordinate amounlt of ~in~ and n~ vs aLt~,m~lillg to30 identi3~y the sources ~Ic~lJec-liYe no~ ;cul~in~ ala~ns, ~ereby possibly delaying the rep~ of chronic ~luble~Js.
Summary of the In~ention An ~l~n~e..)~,n~ in ~he art of teleco...~ ~1On~ is achiewd by arran~g an operations supp~rt system so that, in a~cord with an aspeet of the 35 invention, it gen. l~t~s a rule set firom data generated by an ~
tclPcol,..l~ ir~ions system and ~hen applies the rule se~ to da~a subseq ~~ntly . . = , . ~ ~
.. . .. . . ..
.
: "
, .~ , CA 02129~10 1998-02-2~
generated by the telecommunicationssystem to predict that a particular event(s) will Iikely occur in the system. In an illustrative embodiment of the invention, a set of predetermined measurements is generated for the data that is initially obtained from the telecommunications system and then supplied to a particular rule induction method which 5 then generates a number of rules as a function of such measurements and selects a best rule set from the generated rules such that the selected rule set is capable of predicting accurately the occurrence of past events characterized by the data. The rule set is then used to predict future occurrences of the events.
In accordance with one aspect of the present invention there is provided a 10 method of processing particular data generated by a telecommunications system, said particular data being indicative of different types of events that occurred in said telecommunications system, said method comprising the steps of: obtaining said particular data from said telecommunications system, segmenting said particular data into a plurality of groups, processing said particular data contained within at least a first 15 group of said plurality of groups and automatically generating values for respective predetermined measurements, independently generating at least one rule set as a function of said predetermined measurements, and processing said particular data in said first group using said at least one rule set to identify said different types of events and to predict measurements in at least a second group.
20 Brief Description of the D~
FIG. 1 shows a broad block diagram of an operations system;
FIG. 2 shows in flow chart form the program which implements the invention in the rule generator of FIG. 1; and FIG. 3 identifies di~lc;l-l types of performance measurements used in the 25 system of FIG. 1.
Detailed Description Turning now to FIG. 1 there is shown a broad block diagram of a conventional telecommunicationsnetwork 200 and operations support system 300. Operations support system 300, more particularly, supports the overall operation of network 200. One 30 aspect of such support deals with tracking the performance of network 200 by accumulating and processing alarm messages generated by individual facilities comprising network 200. An alarm message may be generated by a facility as a result of one of a number of different conditions. Such conditions include (a) a spurious problem CA 02129~10 1998-02-2~
that occurs for just a short period of time, (b) a chronic problem that occurs periodically, but which can cause an associated facility to slowly degrade over time, (c) a "hard"
failure which may cause the associated facility to become inoperable, etc. The number of such alarms that net~vork 200 may generate within a particular period of time, e.g., 24 5 hours, could be very high. Such alarms are accumulated by system 300 and stored in database 30. An alarm message includes information identifying a particular problem and the identity of the network 200 facilities that might be affected by the problem.
The alarm messages that are stored in d~t~b~ce 30 may be accessed by facility type. For example, all alarm messages associated with a particular type of facility, such 10 as echo cancelers, may be accessed by presenting a request identifying the particular type of facility type to database 30. Such access may be refined by also specifying aparticular period of time in the request. In response to the request, database 30 outputs an alarm message that occurred over the specified period of time for the identified type of facility.
One dilemma that an operations support system faces is which of the large number of alarms that are generated on a daily basis should be addressed first. It is appa~ l that alarms that are generated as a result of a "hard" fault are usually easy to identify and are addressed first. However, the majority of alarms that are generated daily are typically due to spurious conditions and chronic problems. Moreover, it is difficult to determine initially whether an alarm occurred as a result of a spurious condition or a chronic problem. As such, much effort is sometimes directed to tracking down the cause of spurious alarms since they comprise the majority of alarms that occur each day.
Accordingly, alarms identifying a chronic problem may not be dealt with until the facility generating such alarms degrades to a hard fault.
Recognizing that problem, we have adapted an operations support system so that it processes alarms to identify chronic problems and prioritize the repair of those problems. To that end, Operations Support System (OSS) 300 includes rule generator 40 and operations system processor 50 which operate in accordance with the invention to identify and prioritize chronic problems. More particularly, rule generator 40, which may be, for example, embodied in a SPARC station 2 available from Sun Microsystems, Inc., responds to receipt of a command from an external source, e.g., terminal T2, by sending an access request identifying a facility type and period of time to d~t~b~ee 30.
As mentioned above, d~t~b~ee 30, in response to such a request, outputs all alarm CA 02129~10 1998-02-2~
messages that occurred over the specified period of time for the type of facility identified - in the request, in which the facility type may be, for example, echo cancelers, Tl carriers, T3 carriers, toll switches, etc. The specified period of time is typically one to several weeks and may be identified by a starting date and an ending date.
Assuming that the requested period of time equals two weeks and the facility type is Tl carriers, then the number of alarm messages of the requested type that database 30 outputs could be very large if network 200 employs an appreciable number of Tl carriers. Upon receipt of the alarm messages, rule generator 40 stores them in its associated memory 41 such that alar n messages associated with a specific facility are 10 stored in a respective memory array. When all such messages have been received from database 30 and stored in memory 41, then rule generator 40 processes the messages and generates, in accord with an aspect of the invention, a set of rules which may be applied to identify which of the alarms stored in memory 41 are indicative of respective chronic problems. The rules are then passed to operations system processor 50, which then 15 applies the rules to alarrn messages associated with the same type of facility, but occurring thereafter. Such rules are also passed to processor 60, as discussed below.
The program which controls the operation of rule generator 40 to generate a particular set of rules for the processing of alarm messages is shown in flow chart form in FIG. 2. Specifically, the program is entered at block 2000 responsive to receipt of the aforementioned command. At block 2000, the program proceeds to block 2001 where it forms a request message cont~ining the identity of the facility type and particular period of time, as discussed above. The program then supplies the message to database 30. As mentioned above, database 30, in turn, (a) unloads the alarm messages from its associated memory for the specified facility type and for the specified period of time and (b) passes them to rule generator 40. Rule generator 40, in turn and under control of the program, stores the messages in associated memory 41, as mentioned above. When such alarm messages have been so stored, the program then proceeds to block 2002.
(hereinafter the term "alarm" may also be referred to as an event.) At block 2002, the program processes the stored messages to obtain a number of di~~ l performance measurements for each facility of the specified type. The measurements sllmm~ri~e spatial and temporal information about a facility and associated problems. Such measurements may include the number of (a) different types of alarm messages that occurred for the specific facility, (b) times an alarm (event) occurred CA 02129~10 1998-02-2~
during a unit of time for the facility, (c) units of time or (d) the location of the facility, during which the alarm (event) occurred. As will be discussed below, such measurements are determined for first and second windows, Wl and W2, comprising the aforementioned specified period of time. In addition, each window is partitioned into 5 units of time for the purpose of performing such measurements. (An illustrative list of such performance measurements, or features, is shown in Fig. 3.) Specifically, the program blocks 2002, 2003 and 2004 constitute a "do" loop for processing alarms associated with the identified facility. Once it identifies such messages, then the program generates the di~~ l performance measurements for each 10 window, W1 and W2, and for the units of time associated with those windows. In addition, the program ~ccesses database 20 to determine the circuit layout for the facility within network 200. The circuit layout includes the identities of the other network 200 elements that may be connected directly to the facility. The program then performs the same processing steps for a next facility of the specified type. The program exits the do 15 loop and proceeds to block 2005 upon processing the alarm messages for all of the facilities of the specified type.
At block 2005 the program builds a data file, in which a record of the data filecomprises a number of fields for entering respective performance measurements derived for a specific facility and the circuit layout for the facility. The data file also includes a 20 header which identifies the fields forming a record and identifies the contents of the fields. The program proceeds to block 2006 once it has completed the building of the data file. At block 2006, the program enters a rule generation program, which processes the data file using rule induction methods to generate a set of rules for the specified type of facility. For example, a set of rules of the form "if A is > B, then choose class 1 ", 25 where A may represent a performance measurement, B may represent a particular value and class 1 may represent a chronic problem.
(It is noted that the rule induction program represented by block 2006 follows the rule induction technique disclosed in an article entitled "Reduced Complexity Rule Induction", by S. Weiss and N. Indurkhya and published in the Proceedings of the30 International Joint Conference on Artificial Intelligence (IJCAI), Sydney, Australia, 1991, pages 678-684. Alternatively, a number of rule induction programs are commercially available, such as the C4.5 program available from Morgan EC~ nn Publishers, SanMateo, California, U.S.A.).
CA 02129~10 1998-02-2~
Briefly, rule induction methods attempt to find a compact rule set that completely covers and separates various classes of data, e.g., alarm messages respectively relating to spurious and chronic problems. A covering set of rules is found by heuristically searching for a single best rule that covers dirre~ l cases, e.g., records in 5 the data file, for only one class. Once a best conjunctive rule is found for a particular class, e.g., chronic or nonchronic problems, the rule is then added to the rule set and the cases satisfying it are removed from further consideration. This process is repeated until there are no rem~ining cases to be covered. Once a covering rule set is found that separates the classes, the induced set of rules is further refined by using either pruning or 10 statistical techniques. The initial covering rule set is then scaled back to the most statistically accurate subset of rules using training and testing evaluation methods.
When the program has generated a number of sets of rules, it applies those rule sets to the test data portion of the data file generated at block 2005. In doing so, it prunes the rule sets so that an error rate defining the accuracy of each rule set may be 15 minimi7ed when applied to the test data. That is, the program strives to improve the accuracy of a rule set to predict the occurrence of chronic as well as nonchronic alarms.
Following the foregoing, the program then proceeds to block 2007.
The program at block 2007 selects a rule set from the sets of rules generated atblock 2006, in which the selection is based on a specification provided by an external 20 source, for example, a user or another program. Such a specification may request a rule set having a particular predictive accuracy for a certain class of problems, such as chronic problems. The specification may also include coverage for a particular class of problems. Coverage as defined herein is a measure of the range of particular problems, e.g., chronic problems, that a rule set can predict accurately. For example, the following 25 illustrates one possible rule set:
If (ALRD >x, FBL > x2 SLL > X3 OFL>x4or BVL > X5) then choose chronic;
CA 02129~10 1998-02-2~
where x, through X5 are particular values that are generated by the rule generator. The coverage of the above rule set may be increased by adding, for example, the following rule to form a rule set of 2.
If (CT2 & ALRCT > x6 & ALRH > x,) then choose chronic;
5 where x6 and X7 are also values generated by the rule generator.
It is noted, however, that increasing the coverage of a rule set may lead to a decrease in the predictive accuracy of the rule set. The reason for this is that as the number of cases that are covered increases it is likely that errors will occur.
When the program at block 2007 selects a rule set in accordance with the 10 externally provided specification, it then proceeds to block 2008, where it outputs the rule set to processor 50 and processor 60. The program then proceeds to block 2009 where it checks to see it is done, that is, if the program has been directed to generate a rule set for another type of facility. If that is the case, then the program proceeds to block 2010 where it obtains the identity for the specified facility and then returns to 15 block 2001. Otherwise, the program exits.
Returning to FIG. 1, operations system processor 50 applies the rule set to new alarm messages that it obtains from communications network 200 so that it can accurately predict the occurrence of a particular chronic problem and output a message to that effect via path 51. In this way, processor 50 may direct attention away from 20 spurious problems (nonchronic problems) and focus attention on chronic problems so that such problems may be dealt with before they degrade into a hard fault.
The rule set outputted by rule generator 40 may also be used to predict trends in the occurrence of chronic problems. Quality control processor 60 uses the rule set to determine such trends, and uses the trends in conjunction with other quality control 25 criteria, e.g., the rate at which chronic problems are being repaired, to evaluate the performance of network 200. For example, if a trend indicates that a particular section of network 200 may exceed its number of allowable chronic problems for a given period of time, then processor 60 uses that trend to output a performance value for that section of network 200.
CA 02129~10 1998-02-2~
As mentioned above, rule induction can be applied to other applications as a way of generating rule sets for those applications. For example, one application may be traffic data in which rule induction may be applied to such data to generate a set of rules that may be used to predict particular tends in future traff1c patterns. In particular, traffic 5 data may be collected over a period of time and stored in a database, e.g., database 30.
The above-described program at block 2001 (FIG. 2) may then access that data as well as data indicative of the topology of the network in the manner discussed above. The program may then divide the traffic data into respective windows Wl and W2, and then generate a rule set for the data. The measurements made in Wl are used to predict the 10 measurements in W2. If the prediction is accurate, then the rule set may be used to predict future traffic patterns, in which such patterns may be, for example, predicting applopl;ate levels of blocking for particular telephone routes.
As another example, when a disaster strikes a particular area, e.g., a hurricane, particular controls are activated to automatically block low priority calls so that such 15 calls do not impact the processing and routing of high priority calls. The levels of such controls can be predicted accurately by processing traffic data accumulated during a previous disaster, in accord with the invention, to generate a rule set that can be used to predict such levels of control.
The foregoing is merely iliustrative of the principles of the invention. Those 20 skilled in the art will be able to devise numerous arrangements, which, although not explicitly shown or described herein, nevertheless embody those principles that are within the spirit and scope of the invention.
Claims (10)
1. A method of processing particular data generated by a telecommunications system, said particular data being indicative of different types of events that occurred in said telecommunications system, said method comprising the steps of:
obtaining said particular data from said telecommunications system, segmenting said particular data into a plurality of groups, processing said particular data contained within at least a first group of said plurality of groups and automatically generating values for respective predetermined measurements, independently generating at least one rule set as a function of said predetermined measurements, and processing said particular data in said first group using said at least one rule set to identify said different types of events and to predict measurements in at least a second group.
obtaining said particular data from said telecommunications system, segmenting said particular data into a plurality of groups, processing said particular data contained within at least a first group of said plurality of groups and automatically generating values for respective predetermined measurements, independently generating at least one rule set as a function of said predetermined measurements, and processing said particular data in said first group using said at least one rule set to identify said different types of events and to predict measurements in at least a second group.
2. The method set forth in claim 1 further comprising the step of applying said at least one rule set to data currently generated by said telecommunications system to determine if individual ones of said different types of events are likely to occur subsequently.
3. The method set forth in claim 1 wherein said step of generating includes the step of generating a plurality of rule sets and selecting as said at least one rule set that one of said plurality of rule sets which meets a predetermined specification.
4. The method set forth in claim 3 wherein said predetermined specification includes a predetermined level of accuracy and a predetermined level of coverage.
5. The method set forth in claim 3 wherein said step of generating includes the steps of generating a data file comprising said predetermined measurements, and supplying said data file to a rule induction generator to generate said plurality of rule sets.
6. The method set forth in claim 1 wherein said different types of events include chronic problems identified by respective alarm conditions that occurred in said telecommunications system.
7. The method set forth in claim 1 wherein said particular data defines different traffic patterns within said telecommunications system.
8. The method set forth in claim 1 wherein said step of generating includes the step of supplying said at least one rule set to an operations system processor so that said operations system processor may use said at least one rule set to identify chronic problems when they occur in said communications system.
9. The method set forth in claim 1 wherein said step of generating includes the step of supplying said at least one rule set to a quality control processor so that saidquality control processor may predict trends in the occurrence of said different types of events.
10. The method set forth in claim 1 wherein said particular data is service provisioning information associated with a particular service feature provided by said telecommunications system.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14078093A | 1993-10-21 | 1993-10-21 | |
US140,780 | 1993-10-21 |
Publications (2)
Publication Number | Publication Date |
---|---|
CA2129510A1 CA2129510A1 (en) | 1995-04-22 |
CA2129510C true CA2129510C (en) | 1999-04-13 |
Family
ID=22492762
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CA002129510A Expired - Fee Related CA2129510C (en) | 1993-10-21 | 1994-08-04 | Automatic temporospatial pattern analysis and prediction in a telecommunications network using rule induction |
Country Status (4)
Country | Link |
---|---|
US (1) | US5724263A (en) |
EP (1) | EP0650302A3 (en) |
JP (1) | JPH07183948A (en) |
CA (1) | CA2129510C (en) |
Families Citing this family (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7107185B1 (en) * | 1994-05-25 | 2006-09-12 | Emc Corporation | Apparatus and method for event correlation and problem reporting |
US5768284A (en) * | 1995-12-29 | 1998-06-16 | At&T Corp | Monitoring of periodic patterns |
JP3191707B2 (en) * | 1996-12-20 | 2001-07-23 | 日本電気株式会社 | Path switching device |
US6336109B2 (en) * | 1997-04-15 | 2002-01-01 | Cerebrus Solutions Limited | Method and apparatus for inducing rules from data classifiers |
GB2337903B (en) * | 1998-05-28 | 2000-06-07 | 3Com Corp | Methods and apparatus for collecting storing processing and using network traffic data |
GB2361327A (en) | 2000-04-11 | 2001-10-17 | Mitel Corp | Automatically executing instruction sets in response to network error logs |
GB2361382A (en) | 2000-04-12 | 2001-10-17 | Mitel Corp | Tree hierarchy and description for generated logs |
GB0008952D0 (en) * | 2000-04-12 | 2000-05-31 | Mitel Corp | Dynamic rule sets for generated logs |
US6870900B1 (en) | 2000-06-16 | 2005-03-22 | Bellsouth Intellectual Property Corporation | Proactive maintenance application |
US6771739B1 (en) | 2000-06-16 | 2004-08-03 | Bellsouth Intellectual Property Corporation | Pressure alarms and reports system module for proactive maintenance application |
US7050547B1 (en) | 2000-06-16 | 2006-05-23 | Bellsouth Intellectual Property Corporation | Digital loop carrier module for proactive maintenance application |
US7409356B1 (en) | 2000-06-21 | 2008-08-05 | Applied Systems Intelligence, Inc. | Method and system for intelligent supply chain collaboration |
AU2001271351A1 (en) * | 2000-06-21 | 2002-01-02 | Applied Systems Intelligence, Inc. | Method and system for providing an intelligent goal-oriented user interface to data and services |
US6892192B1 (en) | 2000-06-22 | 2005-05-10 | Applied Systems Intelligence, Inc. | Method and system for dynamic business process management using a partial order planner |
US6751661B1 (en) * | 2000-06-22 | 2004-06-15 | Applied Systems Intelligence, Inc. | Method and system for providing intelligent network management |
WO2002045393A2 (en) * | 2000-11-30 | 2002-06-06 | Bellsouth Intellectual Property Corporation | Proactive maintenance application |
US6838986B2 (en) * | 2001-09-25 | 2005-01-04 | Koninklijke Philips Electronics N.V. | Notification of use of network-enabled device |
US7899893B2 (en) * | 2002-05-01 | 2011-03-01 | At&T Intellectual Property I, L.P. | System and method for proactive management of a communication network through monitoring a user network interface |
US6983401B2 (en) * | 2002-05-01 | 2006-01-03 | Bellsouth Intellectual Property Corporation | System and method for generating a chronic circuit report for use in proactive maintenance of a communication network |
CN100393048C (en) * | 2006-01-13 | 2008-06-04 | 武汉大学 | Method for building network fault diagnosis rule base |
US8205116B2 (en) * | 2009-02-18 | 2012-06-19 | At&T Intellectual Property I, L.P. | Common chronics resolution management |
US9122995B2 (en) | 2011-03-15 | 2015-09-01 | Microsoft Technology Licensing, Llc | Classification of stream-based data using machine learning |
EP3058679B1 (en) | 2013-10-18 | 2018-10-03 | Telefonaktiebolaget LM Ericsson (publ) | Alarm prediction in a telecommunication network |
CN106202420A (en) * | 2016-07-12 | 2016-12-07 | 中国石油大学(华东) | A kind of method for building up of outcrop knowledge base |
Family Cites Families (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4769761A (en) * | 1986-10-09 | 1988-09-06 | International Business Machines Corporation | Apparatus and method for isolating and predicting errors in a local area network |
US4872122A (en) * | 1987-06-19 | 1989-10-03 | University Of Pennsylvania | Interactive statistical system and method for predicting expert decisions |
US5261086A (en) * | 1987-10-26 | 1993-11-09 | Nec Corporation | Performance analyzing and diagnosing system for computer systems |
US5179634A (en) * | 1989-04-14 | 1993-01-12 | Omron Corporation | System for synthesizing new fuzzy rule sets from existing fuzzy rule sets |
JP2507593B2 (en) * | 1989-04-25 | 1996-06-12 | 株式会社東芝 | Inference method |
US5107497A (en) * | 1989-07-28 | 1992-04-21 | At&T Bell Laboratories | Technique for producing an expert system for system fault diagnosis |
JP2978184B2 (en) * | 1989-10-06 | 1999-11-15 | 株式会社日立製作所 | Control rule creation device |
US5063523A (en) * | 1989-11-16 | 1991-11-05 | Racal Data Communications Inc. | Network management system with event rule handling |
US5159685A (en) * | 1989-12-06 | 1992-10-27 | Racal Data Communications Inc. | Expert system for communications network |
US5282261A (en) * | 1990-08-03 | 1994-01-25 | E. I. Du Pont De Nemours And Co., Inc. | Neural network process measurement and control |
US5243543A (en) * | 1991-01-17 | 1993-09-07 | Hewlett-Packard Company | Remote LAN segment traffic monitor |
DE69230968D1 (en) * | 1991-03-04 | 2000-05-31 | Inference Corp | CASE-BASED DEDUCTIVE SYSTEM |
DE69225822T2 (en) * | 1991-03-12 | 1998-10-08 | Hewlett Packard Co | Diagnostic method of data communication networks based on hypotheses and conclusions |
EP0506038A3 (en) * | 1991-03-27 | 1995-02-15 | Toyoda Machine Works Ltd | Processing apparatus having rule base |
JP3076399B2 (en) * | 1991-06-12 | 2000-08-14 | 松下電器産業株式会社 | Automatic generation of fuzzy inference rules |
JP3184243B2 (en) * | 1991-06-20 | 2001-07-09 | 株式会社日立製作所 | Knowledge acquisition support system in expert system |
JP2605505B2 (en) * | 1991-07-04 | 1997-04-30 | 株式会社日立製作所 | Process operation support rule acquisition system, process operation support system, process operation control system, and process operation support rule acquisition method |
-
1994
- 1994-08-04 CA CA002129510A patent/CA2129510C/en not_active Expired - Fee Related
- 1994-10-12 EP EP94307460A patent/EP0650302A3/en not_active Withdrawn
- 1994-10-21 JP JP6281554A patent/JPH07183948A/en active Pending
-
1997
- 1997-05-30 US US08/866,047 patent/US5724263A/en not_active Expired - Lifetime
Also Published As
Publication number | Publication date |
---|---|
EP0650302A2 (en) | 1995-04-26 |
JPH07183948A (en) | 1995-07-21 |
EP0650302A3 (en) | 1997-10-01 |
CA2129510A1 (en) | 1995-04-22 |
US5724263A (en) | 1998-03-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CA2129510C (en) | Automatic temporospatial pattern analysis and prediction in a telecommunications network using rule induction | |
US5748098A (en) | Event correlation | |
US5734697A (en) | Method and apparatus for improving telecommunications system performance | |
US5568399A (en) | Method and apparatus for power outage determination using distribution system information | |
JP2984684B2 (en) | Traffic control method | |
US6650731B1 (en) | Simulator for simulating an intelligent network | |
US5771274A (en) | Topology-based fault analysis in telecommunications networks | |
Sasisekharan et al. | Data mining and forecasting in large-scale telecommunication networks | |
JPH08307524A (en) | Method and equipment for discriminating risk in abnormal conditions of constitutional element of communication network | |
US7894361B1 (en) | System and method for network capacity engineering | |
US6636486B1 (en) | System, method and apparatus for monitoring and analyzing traffic data from manual reporting switches | |
WO2000025527A2 (en) | Alarm correlation in a large communications network | |
JPH08506946A (en) | Event correlation in communication networks | |
US5923742A (en) | System and method for detecting mass addressing events | |
US5450468A (en) | Method of rapidly assessing damage to outside loop plant | |
CA2486780C (en) | A method and system for network wide fault isolation in an optical network | |
US6373825B1 (en) | Telecommunication network and state propagation method | |
US5734811A (en) | Segment substitution/swap for network restoration pre-plans | |
CN108363529A (en) | The determination method and device of the front stage operation time of Application Program Interface | |
EP0420528B1 (en) | Dynamic transport facility identification | |
Davis et al. | No. 4 ESS: Performance objectives and service experience | |
Ali | Digital switching systems: system reliability and analysis | |
US5805681A (en) | Systems and methods for estimating a blocking probability | |
Suzuki et al. | A method for automatic evaluation of fault effects in the advanced Intelligent Network | |
Wolf | Advanced techniques for managing telecommunications networks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
EEER | Examination request | ||
MKLA | Lapsed |