1. Kim Kaminsky Gary D. Boetticher Department of Computer Science
Building a Genetically Engineerable Evolvable Program (GEEP) Using Breadth-Based Explicit Knowledge for Predicting Software Defects
Kim Kaminsky Gary D. Boetticher
Department of Computer Science
University of Houston - Clear Lake
Houston, Texas, USA

2. Key Concept: Duality in Research

3. A Genetically Engineerable Evolvable Program (GEEP)
Genetic Program Process - 1
Fitness Value =
Model performance
on data.
2 (of many) Chromosomes
Data + A B * - 3 D
888 out of 1000
913 out of 1000

4. A Genetically Engineerable Evolvable Program (GEEP)
Genetic Program Process - 2
Mutation
2 Chromosomes
Crossover + B - 3 D * A + A B + B - D
3.1 * A - 3 D

5. A Genetically Engineerable Evolvable Program (GEEP)
GEEP uses Domain Specific Equations
Pythag. Thm.
Cosine

6. Breadth-Based Explicit Knowledge
Akiyama D = * L
D = 0.12 * C – 0.84
Compton D = *L +
*L2
Gaffney D = *(L)4/3
Halstead D = Volume / 3,000
Lipow D = L * ( ( )
ln2* L + ( )ln2L)

7. The GEEP Experiments Akiyama Compton Gaffney Halstead No External All
Lipow
All
Models
No External
Knowledge

8. } } NASA Data Description NASA Defect Data (KC2 Dataset) 379 Unique
tuples }
Input: Product Metrics (Size, Complexity, Vocaulary) }
Output: Defect Count

9. Experiment Configuration
30 Trials
128 Generations Max.
512 Chromosomes
Initial Maximum Tree Height = 6
Fitness = 1 - Standard error

10. Experimental Results Akiyama Compton Gaffney Halstead No External All
Lipow
All
Models
0.0169
0.0009
0.0012
0.0003
0.0823
0.1626
No External
Knowledge
0.0006
8.023E-06
0.0246
1.733E-06

11. Discussion

12. Conclusions Feasibility of leveraging explicit knowledge
Data Mining Algorithm Mining

13. Future Directions Apply to other NASA Defect Datasets
Equation Reduction
((((Lipow((Compton((9 + 8) ^ ( iv(g) - 9)) + d )) + (AkiyamaLoc((Compton(Gaffney(Lipow(9) ^ lOBlank )) ^ Lipow(b) ^ Lipow(((6 * AkiyamaLoc ((Gaffney((6 * AkiyamaLoc((Lipow(9)) + Lipow((Compton(AkiyamaComp((( Halstead((Compton(( Halstead(Lipow(9) + AkiyamaLoc( AkiyamaLoc(l)) - Halstead(AkiyamaComp( iv(g) ) * Halstead(Compton( AkiyamaComp( AkiyamaLoc(Compton(3)) ^ AkiyamaLoc(Halstead( AkiyamaComp( iv(g) ) ^ Gaffney((lOComment ^ (Gaffney(Lipow(2)) ^ (Compton(6) ^ AkiyamaLoc((6 - 10)) + Compton(Gaffney(9) * Compton(Lipow(iv(g))) ^ Lipow(((6 * AkiyamaLoc((Gaffney(Lipow(((6 * AkiyamaLoc((Gaffney( Lipow((Compton(AkiyamaComp(AkiyamaLoc(Compton(AkiyamaLoc(l) ^ Gaffney(((Halstead((AkiyamaLoc(AkiyamaLoc((3 ^ lOBlank)) - 7)) * Halstead(AkiyamaComp( iv(g))^ (Gaffney(Lipow(2)) ^ (Compton(6) ^ AkiyamaLoc((6 - 10)) + Compton(3) ^ Gaffney(10))) + Compton(l) * Compton(l )))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
Performance Improvement