Quality prediction

Patent metadata

Field	Value
Publication number	US-2016307133-A1
Application number	US-201514688646-A
Country	US
Kind code	A1
Filing date	Apr 16, 2015
Priority date	Apr 16, 2015
Publication date	Oct 20, 2016
Grant date	—

Abstract

A set of predicted binary quality indexes is created from a sample set of application lifecycle information and customer encountered defects (CED) for each module id and revision (rev) pair for each application. Normalized effort and quality related factors are extracted for each module id and rev pair of each application. A binary quality index is created based on a set of weighted CED ratings for each module id and rev pair of each application. A prediction model for the binary quality index is created by training a decision tree-based classifier with the sample set to create a set of prediction weights for each effort and quality factor. The set of prediction weights is applied to the effort and quality related factors to each module id and rev pair of an application under-development to create the set of predicted binary quality indexes.

First claim

1 . A method to create a set of predicted binary quality indexes from a sample set of application lifecycle information and customer encountered defects (CED) for each module id and revision (rev) pair for each application, comprising:
extracting from the sample set normalized effort and quality related factors for each module id and rev pair of each application; creating a binary quality index based on a set of weighted CED severity ratings extracted from the sample set for each module id and rev pair of each application; creating a prediction model for the binary quality index by training a decision tree-based classifier with the sample set and respective binary quality indexes to create a set of prediction weights for each normalized effort and quality factor; applying the set of prediction weights to the normalized effort and quality related factors for each module id and rev pair of an application under-development to create the set of predicted binary quality indexes; and creating a list of recommendations for each module id and rev pair of the application under-development based on a shortest path from one binary quality index state to the other binary quality index state.

2 . The method of claim 1 , wherein the decision tree-based classifier uses a bootstrap aggregating classifier with a Bagging algorithm using a first sub-set of the extracted sample set to create the prediction model and a second sub-set of the extracted sample set to test the model.

3 . The method of claim 1 wherein creating the sample set of lifecycle information includes using synchronization tools to import data from two or more different application lifecycle management systems to a single database.

4 . (canceled)

5 . The method of claim 1 further comprising:
presenting a predicted release state with a return on investment (ROI) configuration of resources, a list of problematic modules, and a first estimated cost to fix problematic modules with the configuration of resources;
accepting a change in the configuration of resources; and
presenting an optimized release state with an updated list of problematic modules and a second estimated cost.

6 . The method of claim 1 , further comprising:
presenting a list of recommendations for the under-development application by applying the set of prediction weights to the respective effort and quality related factors of each module of the under-development application to create a set of predicted Good/Bad quality indexes for each module of the under-development application; and determining the shortest path from bad to good of the quality index state for each module of the under-development application.

7 . A system to predict quality, comprising:
a processor; a non-transient computer readable memory (CRM) connected to the processor and including instructions for performing tasks on the processor, the CRM including program modules and data structures comprising, a data extraction module to create a sample data set (S ij ) indexed by application module revision(i) and identification(j) from application lifecycle information in a historical database including customer encountered defects (CEDs), the sample dataset (S ij ) having, a set of normalized effort and quality related factors for each S ij of each sampled application, and a binary quality index indicating a Good/Bad indication for each S ij based on the number of CEDs and weighted by the respective severity level of the CEDs; a prediction model training module to predict the binary quality index Good/Bad indication for an under-development application configured to use a decision tree-based classifier with the sample data set and respective binary quality indexes to create a set of prediction weights for each normalized effort and quality factor and to apply the set of prediction weights to each respective normalized effort and quality factor for each S ij of the application under-development to create a set of predicted Good/Bad binary quality indexes for each S ij ; and a presentation module to present a list of recommendations for the under-development application by applying the set of prediction weights to the respective normalized effort and quality related factors of each Sit of the under-development application to create a set of predicted Good/Bad binary quality indexes for each Sij, wherein the presentation module is configured to determine the shortest path from bad to good of the binary quality index state for each Sij.

8 . (canceled)

9 . (canceled)

10 . The system of claim 7 wherein the presentation module is configured to:
present a predicted release state with a return on investment (ROI) configuration of resources, a list of problematic modules, and a first estimated cost to fix problematic modules with the configuration of resources; and
present an optimized release state with an updated list of problematic modules and a second estimated cost based on a change in the configuration of resources.

11 . The system of claim 7 , wherein the decision tree-based classifier uses a bootstrap aggregating classifier using a first sub-set of the sample data set to create the prediction model and a second sub-set of the sample data set to test the model.

12 . The system of claim 7 further comprising a recommendation module to create a list of recommendations for each S ij of the under-development application based on the shortest path from bad to good for the binary quality index state.

13 . The system of claim 7 wherein the recommendation module is configured to present a predicted release state, an optimized release state, and a return on investment.

14 . A non-transitory computer readable medium (CRM) comprising computer readable instructions that when executed by a processor cause the processor to:
extract data from a historical database for a plurality of applications that includes lifecycle information and customer encountered defects (CEDs) (S ij ) for each application indexed by each application module revision(i) and identification(j) to create a sample data set having,
a set of normalized effort and quality related factors for each S ij , and
a binary quality index indicating a Good/Bad indication for each S ij based on a set of weighted CEDs severity ratings for each S ij ;
predict the quality index for an under-development application using a decision tree-based classifier with the sample data set and respective binary quality indexes to create a set of prediction weights for each normalized effort and quality factor; present a set of recommendations for the under-development application by applying the set of prediction weights to the respective normalized effort and quality related factors of each Sij of the under-development application to create a set of predicted Good/Bad binary quality indexes for each Sij; and determine the shortest path from bad to good of the binary quality index state for each Sij.

15 . (canceled)

16 . The CRM of claim 14 wherein the set of recommendations includes a predicted release state, an optimized release state, and a return on investment.

17 . The CRM of claim 16 wherein at least one of the predicted release state and the optimized release state is configured to display a problematic module list.

18 . (canceled)

19 . The CRM of claim 14 , wherein the decision tree-based classifier is a bootstrap aggregating classifier with a Bagging algorithm that uses a first sub-set of the sample data set to create the prediction model and a second sub-set of the sample da

Assignees

• Hewlett Packard Development Co Lp

Inventors

• Kour George
• Toplian Yaniv
• Vinik Alon

Classifications

• G06Q10/06313 Primary

  Resource planning in a project environment · CPC title

• G06F8/20

  Software design · CPC title