Abstract: The storage location recommendation process and the order picking are the two important aspects of reducing warehouse operation cost. The recommendation process employs the storage policy to allocate the incoming products in the warehouse to specific storage locations, which strategically determines the ordering picking efficiency and the cost of warehouse operations to some extent, and at the same time determines the efficiency of the warehouse operations directly responding to the demand quickly and efficiently. However, based on our investigation in practice, it is found that ecommerce selling different goods usually have different order structure, and usually use different storage policies, like the Class-based random storage and dedicated storage policy that are widely used. Therefore, this motivates us to investigate the operation efficiency of different industries under the Class-Based random storage and dedicated storage policy.
In our paper, the comparison of warehouse operation efficiency in a planar warehouse under the Class-Based random storage and dedicated storage strategy is considered, namely, the comparison of the total cost of location recommend process and the picking process in the industries of the fast-selling goods industry, the book industry and the pharmaceutical industry. A mathematical model based on the location recommend process and the picking process is built. And the Class-Based random storage and the dedicated storage strategy under the same warehouse background are analyzed. Based on the characteristics of the order structure under the fast-selling goods industry, the book industry and the pharmaceutical industry, three typical enterprises under their respective industries are selected for analysis. Their order structure of the three industries is 9-12 items, 2-3 items and 1-2 items. According to the order structure obtained from the actual investigation and the actual sales volume of goods, the order generation is simulated. Then a reasonable simulation of the order structure based on the warehouse operation and design an algorithm is made to solve the model.
Discrete differential evolution algorithm is employed to test the different types of order structures underthe Class-Based random storage and the dedicated storage strategy respectively. The results show that: (a) In those three industries, the total cost of the dedicated storage is lower than the total cost of the Class-Based random storage strategy if the proportion of the cost of its location recommend process to the total cost is less than 0.25. (b) If the proportion is less than 0.26 but greater than 0.25, Both the fast-selling goods industry and the pharmaceutical industry have lower total cost of dedicated storage strategies, while the book industry has lower cost for the Class-based random storage. (c) And if the proportion is less than 0.27 but greater than 0.26, the total cost of the dedicated storage strategy is lower in the fast-selling goods industry while higher in the book industry and the pharmaceutical industry. (d) The total cost of the Class-Based random storage is lower than that of the dedicated storage strategy if the proportion is higher than 0.27 in the three industries.
When the location recommend process is as important as the picking process, under the two storage strategies, the cost of the location recommend process of the respective strategies is not so much different. But the cost of the picking process is gradually reduced from the fast-selling goods industry and the book industry to the pharmaceutical industry. And the total cost of the Class-Based random storage is 7%-8% lower than the total cost of the dedicated storage strategy. Besides, the total cost of its location recommend process is 25.2% lower than that of the dedicated storage strategy. But for the picking process, the picking cost of the Class-Based random storage is 7.2% higher that of the dedicated storage strategy. Research shows that our paper can provide a certain reference value for the storage strategy choice of different order structure. In the actual warehouse operation, the order data can be analyzed according to the actual sales ratio of goods, and the appropriate storage mode can be selected based on the emphasis of the location recommend process and the picking process.

Key words: class-based random storage, dedicated storage, order structure, discrete differential evolution algorithm