K-Means Associated Terrorism Clustering Avoiding Noise
Terrorism has become an ever increasing menace globally, especially in the Indian- subcontinent with its diverse terrain and inherent threats. Terrorism is Spatio-Temporal- Cultural in nature. It can be eradicated through accurate intelligence regarding location, pattern, timing, type of attack and terror outfit responsible, derived from clustering data mining techniques. The significant research problem of profiling the terror outfits, to infer their signature timing, location and type of attack, can be solved through clustering. Partition based clustering’s characteristic limitation is that it includes noisy data but gives good coverage.Density based clustering fails when locational data is spread uniformly but is immune to noise. KATSCAN, the hybrid technique proposed in this paper, leverages advantage of both partition and density based clustering techniques and outperforms them. On real world terrorism data for varied Indian terrain, KATSCAN gives promising results which will be a boon for foot soldiers in fighting terrorism.