Optimization of programming teams on compatibility of programmers

Abstract

The programming team formation problem has been solved using different optimization criteria: programmer and programming team competences; required set of skills, productivity of teams, etc. This paper formulates the problem of optimizing programming teams accounting for pairwise compatibility of programmers described by a matrix whose elements are changes of the programmer and team runtimes when two programmers are included in the same team. When the matrix element is positive the runtime increases, when it is negative the runtime decreases. The problem is formulated as to partition a set of programmers into a set of teams in such a way that the overall teams’ runtime is minimal. The graph clique partitioning problem is related to the team formation problem. It maximizes the overall sum of constant weights of edges located within the cliques. The team formation problem differs because it searches for a solution by changing the graph edge weights. Both problems are NP-hard. The paper proposes a greedy algorithm of stepwise pairwise merge of programming teams and provides a software for team optimization. Experimental results show that the algorithm finds partitions of large sets of programmers and generates teams which reduce the runtime by up to 36 % compared to the one-programmer teams and the single team.