Dynamic optimization of service part inventory control policy through applied data mining and simulation.

This research defines a novel approach for associating inventory item behavior, focusing initially on demand patterns, with an optimal inventory control policy. This method relies upon the definition of typical service part inventory demand patterns and the ability of data mining algorithms to classify inventory transaction data into one of these defined demand patterns. To facilitate this data mining effort, a simulation which creates archetypal inventory demand time series is proposed as the training data source for the data mining task. Actual service part inventory transactions thus classified will be used in a separate service part inventory simulation, modeling a multi-item inventory controlled using a set of common stochastic inventory control policies. Through simulation optimization, using simultaneous perturbation stochastic approximation (SPSA), an optimal demand-pattern to control-policy pairing is sought. The resulting set of optimal pairings will then be used to determine the optimal policy which should be applied to actual service part inventory items after performing demand classification data mining of the actual inventory transaction time series. Improving the efficiency of inventory management within the maintenance and repair service business area holds great promise for reducing inventory investment and improving customer service. Ideally, application of this research could enable an inventory management system which supports the use of multiple concurrent and dynamic inventory management policies focused on reducing inventory cost and increasing customer service and complex equipment availability