Additive manufacturing with metals is garnering interest for its potential in producing complex parts through near-net shape fabrication, with the potential to significantly reduce waste, cost, and lead-time. Use of the laser hot wire process for additive manufacturing offers relatively superior products when compared with other techniques. However, increasing the speed and material deposition rates, tends to induce unwanted arcs between the welding head and molten pool. These arcs produce spatter and irregularities in an otherwise relatively smooth, uninterrupted bead. Presented is a classifier that is able to predict the onset of these arcs on the scale of milliseconds before they occur, based only on voltage and current measurements. This classifier is developed to later be implemented as part of a LHW controller that can readily refute any arc generation.