Machine learning for multi-criteria inventory classification applied to intermittent demand

Multi-criteria inventory classification groups inventory items into classes, each of which is managed by a specific re-order policy according to its priority. However, the tasks of inventory classification and control are not carried out jointly if the classification criteria and the classification approach are not robustly established from an inventory-cost perspective. Exhaustive simulations at the single item level of the inventory system would directly solve this issue by searching for the best re-order policy per item, thus achieving the subsequent optimal classification without resorting to any multi-criteria classification method. However, this would be very time-consuming in real settings, where a large number of items need to be managed simultaneously.

In this article, a reduction in simulation effort is achieved by extracting from the population of items a sample on which to perform an exhaustive search of best re-order policies per item; the lowest cost classification of in-sample items is, therefore, achieved. Then, in line with the increasing need for ICT tools in the production management of Industry 4.0 systems, supervised classifiers from the machine learning research field (i.e. support vector machines with a Gaussian kernel and deep neural networks) are trained on these in-sample items to learn to classify the out-of-sample items solely based on the values they show on the features (i.e. classification criteria). The inventory system adopted here is suitable for intermittent demands, but it may also suit non-intermittent demands, thus providing great flexibility. The experimental analysis of two large datasets showed an excellent accuracy, which suggests that machine learning classifiers could be implemented in advanced inventory classification systems.     

Robust multi-criteria sorting with the outranking preference model and characteristic profiles

We present a new multiple criteria sorting approach that uses characteristic profiles for defining the classes and outranking relation as the preference model, similarly to the Electre Tri-C method. We reformulate the conditions for the worst and best class assignments of Electre Tri-C to increase comprehensibility of the method and interpretability of the results it delivers. Then, we present a disaggregation procedure for inferring the set of outranking models compatible with the given preference information, and use the set in deriving, for each decision alternative, the necessary and possible assignments. Furthermore, we introduce simplified assignment procedures and prove that they maintain a no class jumps-property in the possible assignments. Application of the proposed approach is demonstrated by classifying 40 land zones in 4 classes representing different risk levels.     

Class‐Based Storage with a Finite Number of Items: Using More Classes is not Always Better

Class‐based storage is widely studied in the literature and applied in practice. It divides all stored items into a number of classes according to their turnover. A class of items with higher turnover is allocated to a region closer to the warehouse depot. In the literature, it has been shown that the use of more storage classes leads to a shorter travel time for storing and retrieving items. A basic assumption in this literature is that the required storage space for all items equals their average inventory level, which is valid only if an infinite number of items can be stored in each storage region. This study revisits class‐based storage by considering each storage space to contain only a finite number of items. We develop a travel time model and an algorithm that can be used for determining the optimal number and boundaries of storage classes in warehouses. Different from the conventional research, our findings illustrate that commonly a small number of classes is optimal. In addition, we find the travel time is fairly insensitive to the number of storage classes in a wide range around the optimum. This suggests that a manager can select a near‐optimal number of storage classes in an easy way and need not be worried about the impact of storage‐class reconfigurations. We validate our findings for various cases, including different ABC‐demand curves, space‐sharing factors, number of items, storage rack shapes, discrete storage locations, and stochastic item demand.     

Interval-based extensions of two outranking methods for multi-criteria ordinal classification

We present two new interval-based extensions of ELECTRE TRI-nB and ELECTRE TRI-nC, in which the preference relations are built by using the interval outranking approach. The boundaries between adjacent classes (or categories) are described by a set of limiting profiles, whose criteria could take on values which are interval numbers. Compared to INTERCLASS (a recently proposed interval-based ordinal classification method), the assignments suggested by our first extension come from enhanced preference relations between the actions and limiting boundaries and are likely to be more appropriate. In addition, it can suggest assignments from a pseudo-conjunctive logic. In our second extension, each class is characterized by a set of representative actions, again allowing interval numbers as criterion scores. The outranking relations between actions and classes are enhanced by having several representative actions in each category. Both new extensions can handle imperfect knowledge of the model parameters (weights, veto thresholds, credibility threshold), even missing criterion scores. Some fundamental properties and results are proved to guarantee the consistency of the methods, which are also illustrated with some numerical examples.     

Shelf Space Management When Demand Depends on the Inventory Level

Two factors that their influence on the demand has been investigated in many papers are (i) the shelf space allocated to a product and to its complement or supplement products and (ii) the instantaneous inventory level seen by customers. Here we analyze the joint shelf space allocation and inventory decisions for multiple items with demand that depends on both factors. The traditional approach to solve inventory models with a state‐dependent demand rate uses a time domain approach. However, this approach often does not lead to closed‐form expressions for the profit rate with both dependencies. We analyze the problem in the inventory domain via level crossing theory. This approach leads to closed‐form expressions for a large set of demand rate functions exhibiting both dependencies. These closed‐form expressions substantially simplify the search for optimal solutions; thus we use them to solve the joint inventory control and shelf space allocation problem. We consider examples with two products to investigate the significance of capturing both demand dependencies. We show that in some settings it is important to capture both dependencies. We consider two heuristics, each one of them ignores one of the two dependencies. Using these heuristics it seems that ignoring the dependency on the shelf space might be less harmful than ignoring the dependency on the inventory level, which, based on computational results, can lead to profit losses of more than 6%. We demonstrate that retailers should use their operational control, e.g., reorder point, to promote higher demand products.     

PRODUCT LINE AND INVENTORY STRATEGY

The problem of optimal variety in inventory heretofore has been approached on an individual item basis. This paper demonstrates the analysis of aggregate inventory through the use of the log-normal distribution. The theory and application of this distribution to inventory problems are reviewed. This distribution is then applied to the problem of optimal variety, measuring the impact on sales and cost of deletion of slow-moving items in order to determine the optimal assortment for profitability. This then leads to analysis of optimal margin determination for slow-moving items. The final part of the discussion deals with inventory characteristics of the distribution channel, utilizing the characteristics of the log-normal distribution.     

Coordinated Replenishments from a Common Supplier

Coordinated replenishment strategies may be implemented by jointly ordering multiple items from a common supplier. A major benefit of coordinated replenishment is that it increases the size of shipments, permitting the buyer to enjoy transportation economies without a major increase in average inventory levels. The coordinated replenishment problem is complex because side constraints govern the attainment of transportation rate breaks. The problem is further complicated by the presence of purchase quantity discount opportunities. Thus, the buyer must decide which items to order independently, which items to include in a group order, and the order quantities of each item, governed by the frequency of independent or group orders. We present a mathematical model and a heuristic solution procedure that provide analytical support to the buyer seeking to minimize total costs of replenishing multiple items from a common supplier. The relevant costs are purchase prices, ordering costs, holding costs, and transportation costs. Coordinated replenishment provides nearly a 30 percent reduction in controllable costs relative to independent control. Experimentation with the heuristic has yielded optimal solutions over 88 percent of the time. When optimality was not obtained, the mean penalty was much less than one percent. The average heuristic search was more than two orders of magnitude faster than branch and bound, even for small problems, and possessed a much tighter distribution around the mean search time.     

ABC Classification: Service Levels and Inventory Costs

ABC inventory classifications are widely used in practice, with demand value and demand volume as the most common ranking criteria. The standard approach in ABC applications is to set the same service level for all stock keeping units (SKUs) in a class. In this paper, we show (for three large real life datasets) that the application of both demand value and demand volume as ABC ranking criteria, with fixed service levels per class, leads to solutions that are far from cost optimal. An alternative criterion proposed by Zhang et al. performs much better, but is still considerably outperformed by a new criterion proposed in this paper. The new criterion is also more general in that it can take criticality of SKUs into account. Managerial insights are obtained into what class should have the highest/lowest service level, a topic that has been disputed in the literature.     

Clustering Binary Oligonucleotide Fingerprint Vectors for DNA Clone Classification Analysis

We considered the problem of clustering binarized oligonucleotide fingerprints that attempts to identify clusters. Oligonucleotide fingerprinting is a powerful DNA array based method to characterize cDNA and rRNA libraries and has many applications including gene expression profiling and DNA clone classification. DNA clone classification is the main application for the problem considered in this paper. Most of the existing approaches for clustering use normalized real intensity values and thus do not treat positive and negative hybridization signals equally. This is demonstrated in a series of recent publications where a discrete approach typically useful in the classification of microbial rRNA clones has been proposed. In the discrete approach, hybridization intensities are normalized and thresholds are set such that a value of 1 represents hybridization, a value of 0 represents no hybridization, and an N represents unknown, which is also called a missing value. A combinatorial optimization problem is then formulated attempting to cluster the fingerprints and resolve the missing values simultaneously. It has been examined that missing values cause much difficulty in clustering analysis and most clustering methods are very sensitive to them. In this paper, we turned a little back to the traditional clustering problem, which takes in no missing values but with the revised goal to stabilize the number of clusters and maintain the clustering quality. We adopted the binarizing scheme used in the discrete approach as it is shown to be typically useful for the clone classifications. We formulated such a problem into another combinatorial optimization problem. The computational complexity of this new clustering problem and its relationships to the discrete approach and the traditional clustering problem were studied. We have designed an exact algorithm for the new clustering problem, which is an A* search algorithm for finding a minimum number of clusters. The experimental results on two commonly tested real datasets demonstrated that the A* search algorithm runs fast and performs better than some popular hierarchical clustering methods, in terms of separating clones that have different characteristics with respect to the given oligonucleotide probes.Supported by NSERC and CFI.Supported by NSERC.Supported partially by NSERC, CFI, and NNSF Grant 60373012.     

A production model for deteriorating items with stochastic preventive maintenance time and rework process with FIFO rule

Due to unreliable production facility and stochastic preventive maintenance, deriving an optimal production inventory decision in practice is very complicated. In this paper, we develop a production model for deteriorating items with stochastic preventive maintenance time and rework using the first in first out (FIFO) rule. From our literature search, no study has been done on the above problem. The problem is solved using a simple search procedure; this makes it more practical for use by industries. Two case examples using uniform and exponential distribution preventive maintenance time are applied. Examples and sensitivity analysis are conducted for each case. The results show that rework and preventive maintenance time have significant affected the total cost and the optimal production time. This provides helpful managerial insights to help management in making smart decisions.     

An extension of the Electre I method for group decision-making under a fuzzy environment

Many real-world decision problems involve conflicting systems of criteria, uncertainty and imprecise information. Some also involve a group of decision makers (DMs) where a reduction of different individual preferences on a given set to a single collective preference is required. Multi-criteria decision analysis (MCDA) is a widely used decision methodology that can improve the quality of group multiple criteria decisions by making the process more explicit, rational and efficient. One family of MCDA models uses what is known as “outranking relations” to rank a set of actions. The Electre method and its derivatives are prominent outranking methods in MCDA. In this study, we propose an alternative fuzzy outranking method by extending the Electre I method to take into account the uncertain, imprecise and linguistic assessments provided by a group of DMs. The contribution of this paper is fivefold: (1) we address the gap in the Electre literature for problems involving conflicting systems of criteria, uncertainty and imprecise information; (2) we extend the Electre I method to take into account the uncertain, imprecise and linguistic assessments; (3) we define outranking relations by pairwise comparisons and use decision graphs to determine which action is preferable, incomparable or indifferent in the fuzzy environment; (4) we show that contrary to the TOPSIS rankings, the Electre approach reveals more useful information including the incomparability among the actions; and (5) we provide a numerical example to elucidate the details of the proposed method.     

A production model for deteriorating items with stochastic preventive maintenance time and rework process with FIFO rule

Due to unreliable production facility and stochastic preventive maintenance, deriving an optimal production inventory decision in practice is very complicated. In this paper, we develop a production model for deteriorating items with stochastic preventive maintenance time and rework using the first in first out (FIFO) rule. From our literature search, no study has been done on the above problem. The problem is solved using a simple search procedure; this makes it more practical for use by industries. Two case examples using uniform and exponential distribution preventive maintenance time are applied. Examples and sensitivity analysis are conducted for each case. The results show that rework and preventive maintenance time have significant affected the total cost and the optimal production time. This provides helpful managerial insights to help management in making smart decisions.     

多车场带时间窗车辆路径问题的变邻域搜索算法

多车场带时间窗车辆路径问题是车辆路径问题集合中的一个极为复杂、且仍未得到较好解决的问题。针对这一问题,建立了它的整数规划数学模型,提出了一种改进型变邻域搜索算法。该算法在初始解的构造阶段采用聚类方法完成客户的分配,运用混合算子进行局部搜索,通过后优化过程增强寻优效果,引入模拟退火模型对新解的接受进行控制。最后,在Cordeau提出的标准用例上对改进型变邻域算法进行了实验,实验结果更新了大部分目前该问题的最优解,并在算法的稳定性和求解时间上体现出一定优势。实验表明,该算法是一种求解多车场带时间窗车辆路径问题的有效方法。     

航空公司机上周转品多基地库存优化模型

在民航业日益激烈的市场竞争环境下，航空公司对成本控制的精细化管理显得尤为重要。本文基于航空公司的实际背景，针对为旅客提供的机上周转品，建立了一个多基地库存系统的库存优化模型。该模型以系统的总库存成本最小为目标，在考虑需求、回收、调运、安全库存等因素下，决策一个订货周期内的订货及调运方案。基于最优解的分析，在成本参数满足一定条件下，可以得到系统最优的期初订货量；在此基础上，可以将该模型转化为一个网络流模型，参考最小费用最大流算法，设计了一个多项式求解算法来求解该网络流模型，进而得到各基地之间的最优调运量，并证明该算法的最优性；此外，根据期初订货量可行解的范围，设计了一个求解原库存模型的启发式算法。通过实例分析，验证了该模型与两种算法的有效性。     

A Heuristic Algorithm for the Capacitated Multiple Supplier Inventory Grouping Problem

This paper presents and solves a model for the multiple supplier inventory grouping problem, which involves the minimization of logistics costs for a firm that has multiple suppliers with capacity limitations. The costs included in the model are purchasing, transportation, ordering, and inventory holding, while the firm's objective is to determine the optimal flows and groups of commodities from each supplier. We present an algorithm, which combines subgradient optimization and a primal heuristic, to quickly solve the multiple supplier inventory grouping problem. Our algorithm is tested extensively on problems of various sizes and structures, and its performance is compared to that of OSL, a state-of-the-art integer programming code. The computational results indicate that our approach is extremely efficient for solving the multiple supplier inventory grouping problem.     

A Multicriteria Outranking Modeling Approach for Credit Rating

Rating models are widely used by credit institutions to obtain estimates for the probabilities of default for their clients (firms, organizations, individuals) and to assess the risk of credit portfolios. Several statistical and data mining methods are used to develop such models. In this article, the potential of an outranking multicriteria decision‐aiding approach is explored. An evolutionary algorithm is used to fit a credit rating model on the basis of the ELimination Et Choix Traduisant la REalité trichotomique method. The methodology is applied to a large sample of Greek firms. The results indicate that outranking models are well suited to credit rating, providing good classification results and useful insight on the relative importance of the evaluation criteria.     

A heuristic for inventory management of slow-moving items

Abstract. The traditional approach in the study of inventory systems is to give more importance to items whose demands are either large or very difficult to forecast, discounting the importance of inventory items with small demand. The objective of this paper is to present a simple and practical heuristic to the slow-moving items problem. First we provide insight into previous work in this area and also give a brief outline of classical models in stochastic inventory theory. Then we present the proposed approach and detail the heuristic. We summarize the paper with an analysis of the simulation results and finish with the conclusions that are drawn from the results.     

Inventory Rationing for Multiple Class Demand under Continuous Review

We consider how a firm should ration inventory to multiple classes in a stochastic demand environment with partial, class‐dependent backlogging where the firm incurs a fixed setup cost when ordering from its supplier. We present an infinite‐horizon, average cost criterion Markov decision problem formulation for the case with zero lead times. We provide an algorithm that determines the optimal rationing policy, and show how to find the optimal base‐stock reorder policy. Numerical studies indicate that the optimal policy is similar to that given by the equivalent deterministic problem and relies on tracking both the current inventory and the rate that backorder costs are accumulating. Our study of the case of non‐zero lead time shows that a heuristic combining the optimal, zero lead time policy with an allocation policy based on a single‐period profit management problem is effective.     

Inventory Rationing in a Make‐to‐Stock System with Batch Production and Lost Sales

We address an inventory rationing problem in a lost sales make‐to‐stock (MTS) production system with batch ordering and multiple demand classes. Each production order contains a single batch of a fixed lot size and the processing time of each batch is random. Assuming that there is at most one order outstanding at any point in time, we first address the case with the general production time distribution. We show that the optimal order policy is characterized by a reorder point and the optimal rationing policy is characterized by time‐dependent rationing levels. We then approximate the production time distribution with a phase‐type distribution and show that the optimal policy can be characterized by a reorder point and state‐dependent rationing levels. Using the Erlang production time distribution, we generalize the model to a tandem MTS system in which there may be multiple outstanding orders. We introduce a state‐transformation approach to perform the structural analysis and show that both the reorder point and rationing levels are state dependent. We show the monotonicity of the optimal reorder point and rationing levels for the outstanding orders, and generate new theoretical and managerial insights from the research findings.     

The resource constrained multi-item inventory problem with price discount: a geometric programming approach

In this paper we present a geometric programming approach for determining the inventory policy for multiple items having price discount and a limit on the total average inventory of all the items. An example is solved to illustrate the method.