A density based algorithm has been applied for clustering of cell-hits in the Photon Multiplicity Detector(PMD) installed in the ALICE experiment at CERN. This approach is shown to produce better clustering and thus better correlation among the cell-hits on the two planes of PMD and high energy primary photons. Sixteen features are ex- tracted from the clusters and three multivariate techniques, namely Boosted Decision Trees, Support Vector Machines and Bayesian Neural Networks are used to obtain a classification of hits as either photons or hadrons. The final result is shown to have both a better efficiency as well as better purity of particle identification than the cut based scheme. This has important implications in the search for the Disoriented Chiral Condensates (DCC) conjectured to be formed in relativistic collisions. This is primarily because the cut based analysis relies on high energy cuts on the energy of the photons to obtain reasonable efficiency and purity. This would eliminate low energy photons resulting from the decay of off-shell pions. The work in important as it would be a direct signature of chiral phase transition in these collisions.
|Centralised submission by Collaboration||Presenter name already specified|