Meta-Prediction Models for Bullwhip Effect Prediction of a Supply Chain Using Regression Analysis

Meta-Prediction Models for Bullwhip Effect Prediction of a Supply Chain Using Regression Analysis

Navee Chiadamrong, Nont Sarnrak
DOI: 10.4018/IJISSCM.2021100103
OnDemand:
(Individual Articles)
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

In this study, the main factors that can cause the bullwhip effect and stock amplification are investigated using a simulation-based optimization approach and regression analysis. A two-echelon supply chain with uncertain customer demand and delivery lead time operating with the periodic-review reorder cycle policy is studied. The parameters of smoothing inventory replenishment and forecasting methods are required. These parameters are optimized in terms of minimizing the Total Stage Variance Ratios (TSVRs) of both echelons. The results show that even though all factors of interest have an impact on the bullwhip effect, using smoothing proportional controllers can reduce TSVRs (the sum of the order varaince ratio and net stock amplification). The meta-prediction models can effectively help predict the amount of the bullwhip effect of a chain under various situations with an average MAPE of less than 11%. The results can assist decision makers in the management of a supply chain to realize, benchmark with the optimal results, and reduce the TSVRs under an uncertain environment.
Article Preview
Top

1. Introduction

The bullwhip effect is one of the major causes of supply chain shortages and over-supplied inventory. Supply chain managers experience this variance amplification in inventory levels and orders (Hassanzadeh et al., 2014). This phenomenon refers to the amplification of demand or inventory variability as it moves up the supply chain. It can be defined as the distortion of demand information moving upstream in the supply chain, causing severe inefficiencies in the whole supply chain. Many researchers have claimed that the bullwhip effect is a result of many combined reasons such as inaccurate demand forecasting, uncertainty of the lead time, inventory replenishment policy, fluctuation of end-customer demand, and batch size of stock ordering (Chatfield et al., 2004).

The inventory replenishment policy is the process by which policy takes the necessary actions to bring products down the supply chain methodically to maintain stock levels. Typically, there are two types of inventory replenishment of interest: the reorder level policy and the reorder cycle policy. The difference is at what time and what quantity of products to deliver to the downstream echelon that can satisfy the customer demand. The reorder level policy orders only when the inventory level falls below the predetermined minimum reorder level while the reorder cycle policy defines a pre-determined “reorder cycle period “and orders at every period. The actual order quantity can be determined by calculating the difference between the maximum target stock level and the stock on hand at the review period (daily, weekly, and monthly).

Traditionally, the performance in a supply chain and the so-called “Bullwhip Effect” are evaluated by the order variance and the inventory variance. More researchers have studied the order variance ratio as the performance measure than the inventory variance ratio. However, our study combines both of them into one objective function called the “Total Stage Variance Ratio” (TSVR). Similarly, the studies of Wang and Shalaby (2016) and Costantino et al. (2014) also used the TSVR to determine the overall performance of a supply chain. The order variance ratio increases the cost at the upstream echelon while the inventory variance ratio increases the holding and shortage costs. Hence, it is worthwhile to determine the TSVR, considering both factors as equally important at each echelon (Disney and Lambrecht, 2008). By minimizing the TSVR, the overall performance is improved in a supply chain. There have also been a few researchers who studied the influence of smoothing order policies on inventory performance (Disney and Lambrecht, 2008; Costantino et al., 2016). They also reported the trade-off between order smoothing and inventory performance where order smoothing can give a lower customer service level. In addition, a majority of the studies have focused on a single echelon chain but our study uses a two-echelon supply chain so different mechanisms (between the centralized and decentralized controlling policies) can be explored.

Regarding the two-echelon supply chain in this study, the simulation-based optimization model is constructed and run to minimize the TSVRs by the OptQuest optimization tool. All smoothing parameters of the replenishment inventory policy and the forecasting method’s parameters are searched for their optimal settings, leading to an optimal ordering size for each order in each review period. Based on the results, the meta-prediction models can be constructed to determine the best level of the TSVRs for a two-echelon supply chain under the ROC policy with the exponential smoothing forecasting technique under fluctuating lead time and end-customer demand. In addition, this two-echelon supply chain is modeled with both the decentralized and centralized controlling policies, with and without proportional controllers (smoothing parameters). The major contribution of this study is to assist decision makers in predicting and realizing the amount of the TSVRs in the chain under both controlling situations, including with and without optimizing proportional controllers. They can then prepare and benchmark their supply chain system’s performance to the optimal results, obtained from the meta-prediction models. The ordering quantity in each review period is optimized in terms of the lowest TSVRs by adjusting two proportional controllers. With known TSVRs and causes, a proper alleviation plan to reduce such effects can be made.

Complete Article List

Search this Journal:
Reset
Volume 17: 1 Issue (2024)
Volume 16: 1 Issue (2023)
Volume 15: 7 Issues (2022): 6 Released, 1 Forthcoming
Volume 14: 4 Issues (2021)
Volume 13: 4 Issues (2020)
Volume 12: 4 Issues (2019)
Volume 11: 4 Issues (2018)
Volume 10: 4 Issues (2017)
Volume 9: 4 Issues (2016)
Volume 8: 4 Issues (2015)
Volume 7: 4 Issues (2014)
Volume 6: 4 Issues (2013)
Volume 5: 4 Issues (2012)
Volume 4: 4 Issues (2011)
Volume 3: 4 Issues (2010)
Volume 2: 4 Issues (2009)
Volume 1: 4 Issues (2008)
View Complete Journal Contents Listing