39th International Conference on Software Engineering May 26, 2017 | Buenos Aires, Argentina Impact-Driven Process Model Repair (Journal-First ACM TOSEM Paper, DOI: 10.1145/2980764) Artem Polyvyanyy 1,*, Wil M. P. van der Aalst 1 Queensland 2 2,1, Arthur H. M. ter Hofstede 1,2, and Moe T. Wynn 1 University of Technology, Brisbane, Australia Eindhoven University of Technology, Eindhoven, The Netherlands * Presenting/corresponding author Deviances… Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 2 Conformance Checking perfect alignments t2 b p1 t1 real-world traces t3 p3 p2 c p5 a t4 p4 d t5 a b c γ1 = a b c a d e γ2 = a d e t1 t2 t3 t1 t4 t 5 e designed traces move on trace a b x e≫ γ3 = a b ≫ ≫ c t1 t2 t3 0 + 0 +1 + 1 + 2 = 4 synchronous move a b x≫e γ4 = a b ≫ c ≫ t1 t2 move on model a≫b x e γ5 = a d ≫ ≫ e t3 0 + 0 +1 +2 +1 =4 t1 t4 t5 0 +3 +1 +1 + 0 = 5 cost(γ3)=cost(γ4)=4 and cost(γ5)=5; γ3 and γ4 are optimal alignments! Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 3 Process Model Repair a≫e a b x≫e γ4 = a b ≫ c ≫ γ6 = a d e x t6 t2 b p1 t1 p2 t3 p3 a c t1 t2 p5 t7 t4 d p4 t3 t1 t4 t5 0 0 1 2 1 t5 e insert 0 3 0 skip =7 R=({x},{d}) Note the problem with x! a b x≫e γ7 = a b x c ≫ t1 t2 t6 t3 0+0 +0 +2 +1 = 3 a≫b x e γ9 = a τ ≫ ≫ e t1 t7 t5 a≫e γ8 = a τ e t 1 t7 t5 0+ 0 +1 +1+ 0 = 2 0+ 0 + 0 = 0 γ8 and γ9 are optimal alignments! = 2 Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 4 Towards Optimal Repair simulating repair via “free” moves t2 b p1 t1 p2 t3 c p3 p5 a≫e a≫b x e γ10 = a d ≫ ≫ e γ11= a d e t5 t1 t4 t5 t 1 t4 t5 e 0 0 1 0 0 0 0 0 x t6 R=({x},{d}) =1 a≫b x e γ12 = a τ ≫ x e a≫e γ8 = a τ e a t7 t4 d p4 t1 t7 t6 t5 t1 t7 t5 0+ 0 +1 +0+ 0 = 1 0+ 0 + 0 = 0 γ8 and γ12 are optimal alignments! =1 Note no problem with x! Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 5 Optimal Repair (Theorem 5.3) R=({x},{d}) t2 b p1 t1 p2 a There exist optimal alignments between the traces and the repaired model which: t3 c p3 p5 t7 t4 d p4 t5 (ii) “Fulfill” the repair recommendation, e.g., γ8 and γ12 contain no “bad” moves (x,≫) and (≫,d). e x t6 a≫b x e γ12 = a τ ≫ x e t1 t7 (i) Demonstrate that the traces fit the repaired model “at least as good” as they fit the original model; t6 t5 a≫e γ8 = a τ e t1 t7 t5 0+ 0 +1 +0+ 0 = 1 0+ 0 + 0 = 0 γ8 and γ12 are optimal alignments! =1 Note no problem with x! Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 6 One More Example t2 p7 b p9 f t11 t7 p3 g p2 p1 t8 t5 t3 t1 a p6 p8 t9 p10 p11 d p4 t10 p5 c t4 h t6 e 9x 10 x 7x 9x 2x 6x All deviations cost one Peso! 2x Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 = 120 7 A Possible Repair t2 t13 b f t15 p9 p7 p3 p2 p1 t5 t3 t1 a t7 p8 d t4 p5 p4 p10 p11 e x t17 t6 t10 e h R = ({f,e,x},{c,f,g}) insert t11 t9 p6 t16 t14 f g c t12 t8 skip Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 = 47 8 Another Possible Repair t2 t12 t13 b e f p7 p1 t1 a d p2 t15 t3 p3 t4 p5 p4 t8 p9 f t5 t7 p8 t9 p10 t11 g t17 c p11 p6 t14 e t18 h t16 t13 t6 t10 e R = ({f,e},{g,d,e,c}) insert skip Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 = 33 9 Towards Optimal Repair Recommendation t12 t2 x b p3 p1 a t4 p4 p7 d p2 t15 t3 t1 t5 t8 f t7 p8 t9 p10 g p5 R = ({f,x},{d,e,c,h}) p9 t11 t14 f t17 c insert p6 p11 t18 = 25 f t16 implemented heuristics t13 t6 t10 e h skip costs of repair recommendation numbers of discovered recommendations numbers of alignment computations Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 deviation costs 10 Conclusion and Future Work In this work, we … Defined, and proposed a solution to, the optimal process model repair problem Defined, and proposed several (approximate) solutions to, the optimal repair recommendation problem Developed a publicly available tool that implements all the invented techniques, https://svn.win.tue.nl/repos/prom/Packages/SelectivePRepair/Trunk/release/ Reported on lessons learned from experiments with real-world data, e.g., for small repairs proposed heuristics perform just as well as brute-force Possible avenues for future work include: Case studies to better understand requirements of process model repair Aesthetic process model repair Repair heuristics with guarantees Repairs based on other conformance data structures Repairs that go beyond trace fitness Let’s work together: Engage into research collaborations with us Write your master and/or PhD thesis with us (scholarships available) Artem Polyvyanyy | May 26, 2017 | Impact-Driven Process Model Repair | ICSE 2017 11
© Copyright 2025 Paperzz