Computer and graph data generation method

Patent metadata

Field	Value
Publication number	US-9846679-B2
Application number	US-201415109320-A
Country	US
Kind code	B2
Filing date	Feb 3, 2014
Priority date	Feb 3, 2014
Publication date	Dec 19, 2017
Grant date	Dec 19, 2017

Abstract

A computer utilizes correlation matrix data including a value indicating correlation between indicators and comprises a graph process unit that includes a control factor calculation unit configured to calculate a control factor for generating compressed graph data based on a constraint condition on a process; and a graph generation unit configured to generate graph data from the correlation matrix data by using the control factor. The control factor calculation unit obtains a target process time as the constraint condition; calculates a maximum of a number of edges to be included in the graph data; calculates a threshold for extracting the elements converted from the correlation matrix data to the graph data based on the maximum of the number of edges; and outputs the threshold as the control factor. The graph generation unit extracts the elements from the correlation matrix data based on the threshold, and generates the graph data.

First claim

What is claimed is:

1. A computer comprising a processor and a memory coupled to the processor, configured to execute a process by using correlation matrix data which includes a plurality of elements each of which is a value indicating correlation between a plurality of indicators,
wherein the computer comprises a graph generation unit configured to generate graph data from the correlation matrix data obtained from a storage apparatus, the graph data includes a plurality of vertexes each of which corresponds to an indicator, a plurality of edges each of which connects two of vertexes having correlation, and a plurality of weights of the plurality of edges each of which indicates a value of a element,
wherein the graph process unit includes:
a control factor calculation unit configured to calculate a control factor for generating compressed graph data based on a constraint condition on a process by using the correlation matrix data; and
a graph generation unit configured to generate the graph data from the correlation matrix data by using the control factor,
wherein the control factor calculation unit is configured to:
obtain a target process time which is a completion time for the process by using the correlation matrix data, as the constraint condition;
calculate a maximum of a number of edges to be included in the graph data to complete a process within the target process time;
calculate a threshold for extracting the plurality of elements converted from the correlation matrix data to the graph data based on the maximum of the number of edges; and
output the threshold as the control factor to the graph data generation unit, and
wherein the graph generation unit is configured to:
extract at least one of the plurality of elements from the correlation matrix data based on the threshold, and
generate the graph data by generating information of the plurality of vertexes, the plurality of edges, and the plurality of weights of the plurality of edges based on a combination of the plurality of indicators for identifying the extracted at least one of the plurality of elements and a value of the at least one of the plurality of elements.

2. The computer according to claim 1 ,
wherein the graph data generation unit is configured to:
set a value of the element of the correlation matrix data whose size is smaller than the threshold as zero, and
extract the at least one of the plurality of elements whose value is not zero from the correlation matrix data.

3. The computer according to claim 2 ,
wherein the graph data generation unit is configured to extract the at least one of the plurality of elements whose value is greater than the threshold from the correlation matrix data.

4. The computer according to claim 3 ,
wherein the graph process unit includes an edge information amount calculation unit configured to generate a frequency distribution of a plurality of values of the plurality of elements of the correlation matrix data by analyzing the plurality of elements of the correlation matrix data, and
wherein the control factor calculation unit is configured to:
calculate a function indicating a relation between the number of edges and a completion time for the process by using the correlation matrix data;
calculate the maximum of the number of edges based on the function and the target process time; and
calculate the threshold based on the frequency distribution and the maximum of the number of edges.

5. The computer according to claim 4 ,
wherein the control factor calculation unit is further configured to:
obtain a limitation of usage of the memory as the constraint condition;
determine, based on the limitation of usage of the memory, a number of representation bit for each of the plurality of weights of the plurality of edges so that a size of the graph data is smaller than the limitation of usage of the memory; and
output the threshold and the number of representation bit for the each of the plurality of weights of the plurality of edges as the control factor, and
wherein the graph data generation unit is configured to round the plurality of weights of the plurality of edges to the determined number of representation bit.

6. A graph data generation method for a computer including a processor and a memory coupled to the processor, configured to execute a process by using correlation matrix data which includes a plurality of elements each of which is a value indicating correlation between a plurality of indicators,
wherein the computer includes a graph generation unit configured to generate graph data from the correlation matrix data obtained from a storage apparatus, the graph data including a plurality of vertexes each of which corresponds to an indicator, a plurality of edges each of which connects two of the vertexes having correlation, a plurality of weights of the plurality of edges each of which includes a value of a element,
wherein the graph process unit includes a control factor calculation unit configured to calculate a control factor for generating compressed graph data based on a constraint condition on a process by using the correlation matrix data,
wherein the graph data generation method includes:
a first step of obtaining, by the control factor calculation unit, a target process time which is a completion time for the process by using the correlation matrix data as the constraint condition;
a second step of calculating, by the control factor calculation unit, a maximum of a number of edges to be included in the graph data to complete a process within the target process time;
a third step of calculating, by the control factor calculation unit, a threshold for extracting the plurality of elements converted from the correlation matrix data to the graph data based on the maximum of the number of edges; and
a fourth step of outputting, by the control factor calculation unit, the threshold as the control factor to the graph data generation unit;
a fifth step of extracting, by the graph generation unit, at least one of the plurality of elements from the correlation matrix data based on the threshold; and
a sixth step of generating, by the graph generation unit, the graph data by generating information of the plurality of vertexes, the plurality of edges, and the plurality of weights of the plurality of edges based on a combination of the plurality of indicators for identifying the extracted at least one of the plurality of elements and a value of the at least one of the plurality of elements.

7. The graph data generation method according to claim 6 ,
wherein the fifth step includes steps of:
setting a value of the element of the correlation matrix data whose size is smaller than the threshold as zero; and
extracting the at least one of the plurality of elements whose value is zero from the correlation matrix data.

8. The graph data generation method according to claim 7 ,
wherein the fifth step includes a step of extracting the at least one of the plurality of elements whose value is greater than the threshold from the correlation matrix data.

9. The graph data generation method according to claim 8 ,
wherein the graph process unit includes an edge information amount calculation unit configured to generate a frequency distribution of a plurality of values of the plurality of elements of the correlation matrix data by analyzing the plurality of elements of the correlation matrix data,
wherein
the first step includes a step of calculating a function indicating a relation between the number of edges and a completion time for the process by using the correlation matrix data;
the second step includes a step of calculating the maximum of the number of edges based on the function and th

Assignees

• Hitachi Ltd

Inventors

• Miyamoto Atsushi
• Kudo Yasuyuki
• Miyakoshi Junichi

Classifications

• G06F17/15 Primary

  Correlation function computation {including computation of convolution operations (arithmetic circuits for sum of products per se, e.g. multiply-accumulators G06F7/5443; digital filters, e.g. FIR, IIR, adaptive filters H03H17/00)} · CPC title

• G06F17/10 Primary

  Complex mathematical operations {(function generation by table look-up G06F1/03; evaluation of elementary functions by calculation G06F7/544)} · CPC title

• G06F18/211

  Selection of the most significant subset of features · CPC title

• G06K9/6228

  Physics · mapped topic