Prediction of B2C e-commerce order arrival using hybrid autoregressive-adaptive neuro-fuzzy inference system (AR-ANFIS) for managing fluctuation of throughput in e-fulfilment centres. (15th November 2019)

• Record Type: Journal Article
• Authors: Leung, K.H.; Choy, K.L.; Ho, G.T.S.; Lee, Carman K.M.; Lam, H.Y.; Luk, C.C.
• Is Part Of: Expert systems with applications. Volume 134(2019)
• Page Start: 304
• Page End: 324
• Publication Date: 2019-11-15
• Subjects: E-commerce logistics -- Order arrival prediction -- Warehouse postponement applications -- Adaptive neuro-fuzzy inference system (ANFIS) -- Autoregressive (AR) model; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2019.05.027
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: A new study that forecasts the daily arrival pattern of e-commerce orders. An AR-ANFIS model for e-order arrival prediction is proposed. A two-stage model performance validation is introduced. Experimental results indicate that the proposed model outperforms ARIMA model. Abstract: The complexity of today's e-commerce logistics environment compels practitioners to achieve a higher level of operating efficiency. As it is infeasible for operators to process a large number of discrete e-orders individually, warehouse postponement, that is, delaying the execution of a logistics process until the last possible moment, is essential. Yet the question remains as to how one can accurately identify the timing for consolidating e-orders, and subsequently releasing the grouped e-orders for batch order picking. This is a subject, lacking previous research, but is fundamentally crucial in today's e-commerce logistics environment. This paper introduces an integrated autoregressive-adaptive neuro-fuzzy inference system (AR-ANFIS) approach for forecasting e-commerce order arrivals. Two AR-ANFIS models are built for evaluating their prediction ability against ARIMA models. The experimental results confirm the suitability of the hybrid model for forecasting e-order arrivals. To make use of the model output, an algorithm is formulated to convert e-order arrival figures into cut-off time of order grouping. In this sense, this total solution, packaged as a decision support system, namely the E-order arrival prediction system, assists logistics practitioners in judging when to release the grouped e-orders for batch processing, and essentially improves their order handling capability.