Data processing method for controlling a network
Patterns of change in system operating parameters are identified which may be used to identify unexpected operational conditions and to trigger an appropriate alert or action. A network is controlled using operating data for the network. Network operating data is received and divided into sequential time periods. A structural description such as an association rule is determined for the received data in each time period and a change pattern in the determined structural description is identified over the time periods. The network is then controlled using a predetermined action corresponding to a predetermined structural description change pattern in response to the identified structural description change pattern matching the predetermined structural description change pattern.
1. A method for controlling a network using operating data for the network, the method comprising:
receiving network operating data;
dividing the received network operating data into sequential time periods and, for each of said sequential time periods:
(i) determining support and/or confidence of an association rule for the data received in each of said sequential time periods, and
(ii) identifying a trend or stability in the respective support and/or confidence history of the association rule over a plurality of said sequential time periods to identify an association rule change pattern, and
controlling the network using a predetermined action corresponding to a predetermined association rule change pattern in response to the identified association rule change pattern matching the predetermined association rule change pattern.
2. A method according to claim 1 , further comprising outputting the identified change pattern to a network operator.
3. A method according to claim 1 , further comprising filtering association rules which are temporally redundant compared with other association rules.
4. A method according to claim 1 , wherein identifying change patterns comprises applying a trend statistical test to each determined association rule in order to identify any trends, and applying a stability statistical test to each determined pattern in order to identify any stabilities.
5. A method according to claim 1 , further comprising allocating an interestingness parameter to the identified one or more association rule change patterns depending on one or more statistical measures of a respective association rule change pattern and outputting the identified change patterns depending on their respective interestingness parameters.
6. A method according to claim 5 , further comprising receiving an interestingness rating for an identified association rule change pattern from a user, and wherein allocating an interestingness parameter to the identified change patterns further comprises adjusting the parameter dependent said interestingness rating.
7. A method according to claim 5 , further comprising:
receiving an interestingness rating for a first identified association rule change pattern from a user;
adjusting the interestingness parameter for a second identified association rule in response to determining that the second identified association rule is similar to the first identified change pattern.
8. A method according to claim 5 , wherein allocating an interestingness parameter comprises calculating and combining a number of statistical measures for the respective association rule.
9. A method according to claim 8 , wherein the statistical measures test for the following statistical properties: clarity; pronouncedness; dynamic; homogeneity.
10. A processor code product comprising non-transitory digital storage media carrying processor code which, when executed on a processor, causes the processor to carry out a method according to claim 1 .
11. A data processing apparatus for controlling a network using operating data for the network, the apparatus comprising:
an input for receiving network operating data;
a processor configured to:
divide the received data into sequential time periods,
determine support and/or confidence of an association rule for the received data in each time period;
identify a trend or stability in the respective support and/or confidence history of the association rule over a plurality of said sequential time periods to identify an association rule change pattern, and to control the network using a predetermined action corresponding to a predetermined association rule change pattern in response to the identified association rule change pattern matching the predetermined association rule change pattern.