Method and system for predicting resource requirements for service engagements

• US 20070094061A1
• Filed: 10/12/2005
• Published: 04/26/2007
• Est. Priority Date: 10/12/2005
• Status: Abandoned Application

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Hu, Jianying, Ray, Bonnie
• Application Number:
  US11/247,197
• Publication Number:
  US 20070094061A1
• Time in Patent Office: Days
• US Class Current: 705/7.13
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
• CPC Class Codes: G06Q 10/06311 Scheduling, planning or task assignment for a person or group
• CPC Class Codes: G06Q 30/0204 Market segmentation

Abstract

A method and system for predicting resource requirements of a current service engagement by modeling records of past service engagements to create and classify templates of service resource usage. This is done by clustering past engagements into groups having similar time series requirements for service resources. A service resource template for the current service engagement is generated from a classified template by using characteristics of the current service engagement to select a group of which the current service engagement is a likely member. The corresponding template is then customized to fit the characteristics of the current service engagement. The invention may be implemented using Hidden Markov Models. An aspect of the invention is use of dynamic time warping to quantify dissimilarity between engagement sequences prior to fitting Hidden Markov Models. Another aspect of the invention is removal of outliers from the clustered groups.

Claims

1. A method for predicting resource requirements of a service engagement, comprising:

modeling records of past service engagements to create and classify one or more templates of service resource usage; and

generating a service resource plan for said service engagement using a template created by said modeling, wherein characteristics of said service engagement are used to select a created template by its classification and wherein input attributes of said service engagement are used to customize the selected template.

2. A method according to claim 1, wherein said modeling further comprises:

clustering the past service engagements into groups having similar service resource requirements;

creating a staffing template for each clustered group reflecting service resource patterns typical for engagements in the group; and

identifying characteristics of each group by which a service engagement may be classified as belonging to the group.

3. A method according to claim 1, wherein generating a service resource plan further comprises:

inputting identifying characteristics and input attributes of the service engagement;

assigning the service engagement to a clustered group, based on said identifying characteristics; and

adapting the template of the assigned clustered group to fit the input attributes of the service engagement to generate a plan.

4. A method according to claim 2, wherein the clustering is done by computing the dissimilarity between any two past engagements and then applying hierarchical clustering.

5. A method according to claim 2, wherein the clustering is done by computing the dissimilarity between any two past engagements using dynamic time warping.

6. A method according to claim 2, wherein the clustering is done by fitting Hidden Markov Models to the past engagements.

7. A method according to claim 6, wherein dynamic time warping is used to quantify dissimilarity between engagement sequences prior to fitting Hidden Markov Models.

8. A method according to claim 7, wherein a Baysian Information Criterion (BIC) is used to measure goodness of fit of the Hidden Markov Models, the BIC measures being normalized by the length of said engagement sequences.

9. A method according to claim 6, wherein an engagement sequence whose likelihood is below a threshold value is rejected by an HMM model, and wherein an engagement sequence rejected by all current HMM models is not modeled.

18. A system according to claim 6, wherein an engagement sequence whose likelihood is below a threshold value is rejected by an HMM model, and wherein an engagement sequence rejected by all current HMM models is not modeled.

10. A system for predicting resource requirements of a service engagement, comprising:

means for modeling records of past service engagements to create and classify one or more templates of service resource usage; and

means for generating a service resource plan for said service engagement using a template created by said modeling, wherein characteristics of said service engagement are used to select a created template by its classification and wherein input attributes of said service engagement are used to customize the selected template.

11. A system according to claim 10, wherein said means for modeling further comprises:

means for clustering the past service engagements into groups having similar service resource requirements;

means for creating a staffing template for each clustered group reflecting service resource patterns typical for engagements in the group; and

means for identifying characteristics of each group by which a service engagement may be classified as belonging to the group.

12. A system according to claim 10, wherein means for generating a service resource plan further comprises:

means for inputting identifying characteristics and input attributes of the service engagement;

means for assigning the service engagement to a clustered group, based on said identifying characteristics; and

means for adapting the template of the assigned clustered group to fit the input attributes of the service engagement to generate a plan.

13. A system according to claim 11, wherein the clustering is done by computing the dissimilarity between any two past engagements and then applying hierarchical clustering.

14. A system according to claim 11, wherein the clustering is done by computing the dissimilarity between any two past engagements using dynamic time warping.

15. A system according to claim 11, wherein the clustering is done by fitting Hidden Markov Models (HMMs) to the past engagements.

16. A system according to claim 15, wherein dynamic time warping is used to quantify dissimilarity between engagement sequences prior to fitting Hidden Markov Models.

17. A system according to claim 16, wherein a Baysian Information Criterion (BIC) is used to measure goodness of fit of the Hidden Markov Models, the BIC measures being normalized by the length of said engagement sequences.

19. A computer implemented system for predicting resource requirements of a service engagement, comprising:

first computer code for modeling records of past service engagements to create and classify one or more templates of service resource usage; and

second computer code for generating a service resource plan for said service engagement using a template created by said modeling, wherein characteristics of said service engagement are used to select a created template by its classification and wherein input attributes of said service engagement are used to customize the selected template.

20. A computer implemented system according to claim 19, wherein said first computer code for modeling further comprises:

third computer code for clustering the past service engagements into groups having similar service resource requirements;

fourth computer code for creating a staffing template for each clustered group reflecting service resource patterns over time typical for engagements in the group; and

fifth computer code for identifying characteristics of each group by which a service engagement may be classified as belonging to the group.