Research On Inventory Optimization Problem Based On Machine Learning

With the continuous advancement of economic globalization and the rapid development of Internet technology,the scope of demand of enterprises has been widely expanded through traditional channels and online channels.The involvement of e-commerce has intensified competition among companies,and companies must constantly adopt strategies that focus on adjusting product features to meet customer needs and preferences.In order to further attract consumers,many companies are committed to improving the level of logistics services such as warehousing and distribution.High-level logistics services make the daily inventory cost of enterprises increase day by day.In addition,the application of advanced data collection systems allows enterprises to collect and transmit data from almost all processes.The amount of data faced by enterprises is increasing exponentially,and the demand is becoming more and more fluctuating and random.In the face of massive supply chain data,companies cannot dig out effective information,resulting in the inability to respond quickly to changes in demand.In the end,the only way to avoid shortages is to maintain inventory at a high level,which also increases inventory costs.The promotion of computer technologies such as big data and machine learning has brought new opportunities for enterprises.Machine learning is able to draw information from data,make decisions based on changes in environmental factors,and continuously improve performance as data is fed.Therefore,using machine learning algorithms to optimize inventory management has become a viable option for companies.Based on this,this paper firstly sorts out the relevant literature on inventory optimization problems and machine learning applications,summarizes the theoretical basis of inventory optimization,and introduces the algorithms used in this paper.Starting from the safety stock setting and demand forecasting problems in the inventory optimization problem,and using machine learning algorithms to build a model framework to solve them respectively.Taking enterprise A as an example,the inventory optimization is carried out,and the corresponding inventory management suggestions are put forward.The main work of this paper is as follows:(1)A three-level supply chain is constructed by using simulation modeling technology.The wholesaler in the middle position is used as the decision-making object,and the safety stock setting problem of commodities under different environmental changes is considered,and the random forest algorithm is used to determine the optimal product.The results show that the algorithm can select the safety stock factor with the lowest inventory cost in 79.57%of the cases.In order to enhance the interpretability of the model,we give a ranking of key influencing factors that affect the setting of safety stock based on the feature importance of the tree model,and the visualization of the tree model further explains the algorithm.The results show that the setting of the optimal safety stock mainly depends on the lead time of the node and the correlation coefficient of inventory cost.Finally,the algorithm-based dynamic safety stock setting strategy is compared with the static safety stock setting strategy.The simulation results show that the algorithm can reduce the inventory cost by up to 25.86%.These results show the great potential of this method in the practical application of supply chain.(2)Describe the problems existing in company A’s safety stock setting,use simulation modeling technology to model the inventory operation process of company A,and apply the random forest algorithm of machine learning to determine the best warehouse of company A.Safety stock.Taking the lowest inventory cost as the evaluation index,the results show that the algorithm achieves an accuracy of 79.57%.In order to enhance the interpretability of the model,based on the feature importance of the tree model,this paper gives the ranking of key influencing factors that affect the setting of safety stock,and visualizes the tree model to further explain the algorithm.The results show that the setting of optimal safety stock mainly depends on the warehouse lead time and the correlation coefficient of inventory cost.Finally,the algorithm-based dynamic safety stock setting strategy is compared with the static safety stock setting strategy.The simulation results show that the inventory cost can be reduced by up to 25.86% through the algorithm,and these results illustrate the possibility of this method being practically applied in enterprise A.Finally,the corresponding inventory management suggestions are put forward for A company.