Information processing apparatus, information processing method, and recording medium
An information processing apparatus 4 includes: a transformation unit 423 configured to transform a plurality of tuples of feature vector data XV in a presentation space into a plurality of tuples of latent vector data ZV in a latent space; an extraction unit 423 configured to extract, based on the plurality of tuples of latent vector data, feature vector data having a shorter distance from query data DQ in the latent space than the other feature vector data, as neighbor data DN i from among the plurality of tuples of feature vector data; a generation unit 424 configured to generate, based on the neighbor data, a local model LM that outputs an estimation value dp i of a latent distance d i when difference information V i is inputted, the latent distance being a distance between the query data and the neighbor data in the latent space, the difference information being related to a difference, for each element of the features, between the query data and the neighbor data in the presentation space; and a calculation unit 425 configured to calculate, based on the local model and the difference information, an element contribution degree c i,f representing a magnitude of an effect that each element of the features of the neighbor data exerts on the latent distance.
1. An information processing apparatus comprising:
at least one memory storing configured to store instructions; and
at least one processor configured to execute the instructions to:
transform a plurality of tuples of feature vector data into a plurality of tuples of latent vector data, the plurality of tuples of feature vector data representing features of a plurality of sample data groups, respectively, in a presentation space, the plurality of tuples of latent vector data representing features of the plurality of sample data groups, respectively, in a latent space that is different from the presentation space, wherein the plurality of sample data groups are log data groups indicating an archival record of a communication relayed by a proxy server;
extract, based on the plurality of tuples of latent vector data, the plurality of tuples of feature vector data as a plurality of neighbor data, the extracted plurality of tuples of feature vector data each having a shorter distance from desired query data in the latent space than the plurality of tuples of feature vector data that were not extracted;
generate, based on the plurality of neighbor data, a local model that outputs an estimation value of a latent distance when difference information is inputted thereto, the latent distance being a distance between the query data and the plurality of neighbor data in the latent space, the difference information being related to a difference, for each element of the features, between the query data and the plurality of neighbor data in the presentation space;
calculate, based on the local model and the difference information, an element contribution degree representing a magnitude of an effect that each element of the features of each of the plurality of neighbor data exerts on the latent distance; and
output a list of the element contribution degrees for the plurality of neighbor data.
2. The information processing apparatus according to 1 , wherein
at least one processor configured to execute the instructions to calculate the element contribution degree, based on a parameter defining the local model, and the difference information.
3. The information processing apparatus according to claim 2 , wherein
the local model includes a linear regression model that uses the difference information for an explanatory variable, and uses the latent distance for an objective variable, and
the parameter includes a weight by which the explanatory variable is multiplied.
4. The information processing apparatus according to claim 3 , wherein
the linear regression model does not include a bias term.
5. The information processing apparatus according to claim 3 , wherein
the weight is equal to or more than zero.
6. The information processing apparatus according to claim 1 , wherein
at least one processor configured to execute the instructions to generate the linear regression model represented by expression 1 where the number of elements of the features in each of the plurality of neighbor data and the query data is F (where F is a constant indicating an integer equal to or more than one), the difference information corresponding to a difference between a feature corresponding to an f-th (where f is a variable indicating an integer that satisfies 1≤f≤F) element in the neighbor data and a feature corresponding to an f-th element in the query data is V i,f a weight by which the difference information V i,f is multiplied is w f , and the estimation value of the latent distance outputted by the local model is dp i , and
at least one processor configured to execute the instructions to calculate the element contribution degree by using expression 2 where the element contribution degree representing a magnitude of an effect that the feature corresponding to the f-th element in the neighbor data exerts on the latent distance is c i,f ,
dp
i
3
’
∑
f
=
1
F
w
f
×
v
i
,
f
[
Expression
1
]
c
i
,
f
=
w
f
×
v
i
,
f
dp
i
=
w
f
×
v
i
,
f
∑
f
′
=
1
F
w
f
′
×
v
i
,
f
′
.
[
Expression
2
]
7. The information processing apparatus according to claim 1 , wherein
each element of the features indicated by the feature vector data is allowed to belong to at least one of a plurality of different feature groups, and
at least one processor configured to execute the instructions to calculate, based on the element contribution degrees, a group contribution degree representing a magnitude of an effect that each feature group exerts on the latent distance.
8. The information processing apparatus according to claim 7 , wherein
the plurality of feature groups respectively correspond to a plurality of types of sample information included in the sample data groups,
the information processing apparatus further comprises a display configured to display the sample data group, in a display form in which at least a part of the plurality of types of sample information included in the sample data group for which each of the plurality of neighbor data indicates the features is associated with the group contribution degrees.
9. An information processing method performed by a computer and comprising:
transforming a plurality of tuples of feature vector data into a plurality of tuples of latent vector data, the plurality of tuples of feature vector data representing features of a plurality of sample data groups, respectively, in a presentation space, the plurality of tuples of latent vector data representing features of the plurality of sample data groups, respectively, in a latent space that is different from the presentation space, wherein the plurality of sample data groups are log data groups indicating an archival record of a communication relayed by a proxy server;
extracting, based on the plurality of tuples of latent vector data, the plurality of tuples of feature vector data as a plurality of neighbor data, the at least one extracted plurality of tuples of feature vector data each having a shorter distance from desired query data in the latent space than the plurality of tuples of feature vector data that were not extracted;
generating, based on the plurality of neighbor data, a local model that outputs an estimation value of a latent distance when difference information is inputted thereto, the latent distance being a distance between the query data and the plurality of neighbor data in the latent space, the difference information being related to a difference, for each element of the features, between the query data and the plurality of neighbor data in the presentation space;
calculating, based on the local model and the difference information, an element contribution degree representing a magnitude of an effect that each element of the features of each of the plurality of neighbor data exerts on the latent distance; and
outputting a list of the element contribution degrees for the plurality of neighbor data.
10. A non-transitory recording medium storing a computer program that executable by a computer to perform an information processing method comprising:
transforming a plurality of tuples of feature vector data into a plurality of tuples of latent vector data, the plurality of tuples of feature vector data representing features of a plurality of sample data groups, respectively, in a presentation space, the plurality of tuples of latent vector data representing features of the plurality of sample data groups, respectively, in a latent space that is different from the presentation space, wherein the plurality of sample data groups are log data groups indicating an archival record of a communication relayed by a proxy server;
extracting, based on the plurality of tuples of latent vector data, the plurality of tuples of feature vector data as a plurality of neighbor data, the at least one extracted plurality of tuples of feature vector data each having a shorter distance from desired query data in the latent space than the plurality of tuples of feature vector data that were not extracted;
generating, based on the plurality of neighbor data, a local model that outputs an estimation value of a latent distance when difference information is inputted thereto, the latent distance being a distance between the query data and the plurality of neighbor data in the latent space, the difference information being related to a difference, for each element of the features, between the query data and the plurality of neighbor data in the presentation space;
calculating, based on the local model and the difference information, an element contribution degree representing a magnitude of an effect that each element of the features of each of the plurality of neighbor data exerts on the latent distance; and
outputting a list of the element contribution degrees for the plurality of neighbor data.