ABC analysis is a useful method to classify inventory items into different categories that can be specifically managed and controlled. Traditional ABC analysis classifies inventory items into three categories according to the annual dollar usage of an inventory item. Multi criteria inventory classification models have been proposed by researchers in order to take into account other important criteria. From these models we considered in this paper the R- model and the Ng-model in order to improve them. Illustrative examples are presented to compare the improved models and the initial ones. Keywords: ABC classification, Multi criteria inventory classification models, R-model, Ng-model. I. Literature review Companies often manage a large number of items. It is difficult to give all these items, the same level of monitoring and control. For this reason, managers should set priority levels for each category of items. Once this hierarchy established, we can assign specific business rules to each family of items. In this field, the ABC classification is mostly used by managers for inventory control. The classical ABC classification is based on the Pareto principle and according to the criterion of the annual use value. In fact, Class A items are relatively few in number (between 10% and 20%) but they represent between 70% and 80% of the total annual use value. The Class A items represents a great annual use value. For this reason, control of items of this class must be rigorous and tight. Class B contains between 30% and 40% of items that represent 15% to 20% of the total annual use value. The inventory control of class B items can be less tight than that of the first class A. Finally class C items can reach 50% of the items but represent only 5% to 10% in terms of value. Managers can reduce their effort in controlling these items because they are numerous but too small in terms of total annual use value. When handling inventory classification, managers need to consider at the same time many important classification factors. Several Methodologies have been proposed in the literature for the multi- criteria inventory classification problem (MCIC), but we will focus on the weighted optimization models. These models used a weighted additive function to aggregate the performance of an inventory item in terms of different criteria to a single score, called the inventory score of an item. Ramanathan (2006) proposed a linear model which gives each item an opportunity to choose a set of weights favorable to the item itself when maximizing its score. To improve this model proposed by Ramanathan (2006), Zhou and Fan (2007) suggested a new classification based on a composite index. This score or index is a combination of a good index (maximizing score) and a bad index (minimizing score). The items classification will be then based on the calculated composite index. Ng (2007) proposed a linear programming model for the MCIC using a scalar measuring for all the criteria and with the main contribution of giving an importance order to the criteria. Hadi-Vencheh (2010) presented a nonlinear programming model as an extension version of the Ng-model. This model changed in the linear programming the constraint of the criterion weights by considering squared coefficients. Finally, Chen (2011) developed an improved approach to MCIC by which all items are peer-estimated. This new approach first determines two performance scores for each inventory item, then we determine a weight for each score in order to calculate a final aggregated index that will be used for the classification. In this paper we will suggest firstly, a modified variant of R-model proposed by Ramanathan (2006) for MCIC. In this version, called R*-model, we consider the normalization step of the data to measure and analyze the impact of this pre-processing on the initial model. In the second part of this work, we will propose a new weighted linear optimization model which is an improvement of the Ng- model developed by Ng (2007). This new model will be referred to as M-Ng model.
Read full abstract