Error detection device, error detection method, and error detection program
An object is to make it possible to accurately detect abnormality of event data. A training unit ( 105 ) trains a parameter of a model based on a plurality of event series that are event data in a time series and labels that indicate abnormality or normality with respect to event data of each of the plurality of event series, the model outputting a degree of abnormality of a target event series when the target event series is input, the target event series being an event series of which the degree of abnormality is to be predicted, the parameter being trained to optimize an objective function that represents a relationship between a probability of occurrence of an event at each time point in the time series and a degree of abnormality of each of the plurality of event series.
1 . An abnormality detection device comprising circuitry configured to execute a method comprising:
training a parameter of a model based on a plurality of event series that are event data in a time series and labels that indicate abnormality or normality with respect to event data of each of the plurality of event series, the model outputting a degree of abnormality of a target event series when the target event series is input, the target event series being an event series of which the degree of abnormality is to be predicted, the parameter being trained to optimize an objective function that represents a relationship between a probability of occurrence of an event at each time point in the time series and a degree of abnormality of each of the plurality of event series,
wherein the probability of occurrence is expressed with an intensity function of a point process,
the objective function is expressed with a linear regression model that is expressed using a likelihood of the point process with respect to each of the plurality of event series, wherein the objective function is expressed as:
ℒ
=
∑
i
=
1
n
D
(
y
i
❘
"\[LeftBracketingBar]"
f
(
Z
i
;
β
)
)
,
and
the method further comprising:
training the parameter of the model to maximize the value of the objective function, wherein the parameter is trained based on a plurality of event series in which the plurality of event series comprises event data that indicates abnormality with respect to the event data of each of the plurality of event series.
2 . The abnormality detection device according to claim 1 , the method further comprising:
accepting input of the target event series; and
calculating the degree of abnormality of the target event series based on the target event series, the model, and the trained parameter.
3 . The abnormality detection device according to claim 1 , wherein the event data includes a record of transactions made in a financial market, and wherein the labels indicate whether a stock value indicator fluctuated.
4 . The abnormality detection device according to claim 1 , wherein the event data includes a history of boarding and alighting a taxi, and wherein the labels indicate whether a congestion occurred at a time of boarding or alighting.
5 . The abnormality detection device according to claim 1 , wherein the event data indicates a search log of a route search application, and wherein the labels indicate whether an area on a route was congested.
6 . The abnormality detection device according to claim 1 , the circuitry configured to execute the method further comprising:
accepting input of the target event series; and
calculating the degree of abnormality of the target event series based on the target event series, the model, and the trained parameter.
7 . An abnormality detection method comprising:
training a parameter of a model based on a plurality of event series that are event data in a time series and labels that indicate abnormality or normality with respect to event data of each of the plurality of event series, the model outputting a degree of abnormality of a target event series when the target event series is input, the target event series being an event series of which the degree of abnormality is to be predicted, the parameter being trained to optimize an objective function that represents a relationship between a probability of occurrence of an event at each time point in the time series and a degree of abnormality of each of the plurality of event series,
wherein the probability of occurrence is expressed with an intensity function of a point process,
the objective function is expressed with a linear regression model that is expressed using a likelihood of the point process with respect to each of the plurality of event series, wherein the objective function is expressed as:
ℒ
=
∑
i
=
1
n
D
(
y
i
❘
"\[LeftBracketingBar]"
f
(
Z
i
;
β
)
)
,
and
the method further comprising:
training the parameter of the model to maximize the value of the objective function, wherein the parameter is trained based on a plurality of event series in which the plurality of event series comprises event data that indicates abnormality with respect to the event data of each of the plurality of event series.
8 . The abnormality detection method according to claim 7 , the method further comprising:
accepting input of the target event series; and
calculating the degree of abnormality of the target event series based on the target event series, the model, and the trained parameter.
9 . The abnormality detection method according to claim 7 , wherein the event data includes a record of transactions made in a financial market, and wherein the labels indicate whether a stock value indicator fluctuated.
10 . The abnormality detection method according to claim 7 , wherein the event data includes a history of boarding and alighting a taxi, and wherein the labels indicate whether a congestion occurred at a time of boarding or alighting.
11 . The abnormality detection method according to claim 7 , wherein the event data indicates a search log of a route search application, and wherein the labels indicate whether an area on a route was congested.
12 . The abnormality detection method according to claim 7 , the method further comprising:
accepting input of the target event series; and
calculating the degree of abnormality of the target event series based on the target event series, the model, and the trained parameter.
13 . A computer-readable non-transitory recording medium storing computer-executable an abnormality detection program instructions that when executed by a processor cause a computer system to execute a method comprising:
training a parameter of a model based on a plurality of event series that are event data in a time series and labels that indicate abnormality or normality with respect to event data of each of the plurality of event series, the model outputting a degree of abnormality of a target event series when the target event series is input, the target event series being an event series of which the degree of abnormality is to be predicted, the parameter being trained to optimize an objective function that represents a relationship between a probability of occurrence of an event at each time point in the time series and a degree of abnormality of each of the plurality of event series,
wherein the probability of occurrence is expressed with an intensity function of a point process,
the objective function is expressed with a linear regression model that is expressed using a likelihood of the point process with respect to each of the plurality of event series, wherein the objective function is expressed as:
ℒ
=
∑
i
=
1
n
D
(
y
i
❘
"\[LeftBracketingBar]"
f
(
Z
i
;
β
)
)
,
and
the method further comprising:
training the parameter of the model to maximize the value of the objective function, wherein the parameter is trained based on a plurality of event series in which the plurality of event series comprises event data that indicates abnormality with respect to the event data of each of the plurality of event series.
14 . The computer-readable non-transitory recording medium according to claim 13 , wherein the event data includes a record of transactions made in a financial market, and wherein the labels indicate whether a stock value indicator fluctuated.
15 . The computer-readable non-transitory recording medium according to claim 13 , wherein the event data includes a history of boarding and alighting a taxi, and wherein the labels indicate whether a congestion occurred at a time of boarding or alighting.
16 . The computer-readable non-transitory recording medium according to claim 13 , the method further comprising:
accepting input of the target event series; and
calculating the degree of abnormality of the target event series based on the target event series, the model, and the trained parameter.
17 . The computer-readable non-transitory recording medium according to claim 13 , the method further comprising:
accepting input of the target event series; and
calculating the degree of abnormality of the target event series based on the target event series, the model, and the trained parameter.