Conclusion

Cluster utility was shown to be an index that reflects with high fidelity the quality of clustering result judged by given class structure. When the relationship with quality index, QI was considered, it had very high R-value and correlation coefficient . Well-known relative indices from general clustering applications have been adapted to sequence clustering domain. CU far exceeded them in terms of degree of correlation and  consistency. The fact its high linearity invariably holds across all test data sets demonstrates its robustness and integrity as a relative index. The usual unavailability of class labels in clustering in general and a very large space spanned by algorithm parameters greatly calls for a guiding index that tracks the quality of clustering results. CU aids parameter setting towards optimum including edge cutoff value due to its high correlation with the quality and its consistency.