Research On Demand Forecasting And Multi-level Inventory Cost Optimization Based On Large Fmcg Manufacturing Enterprises

At present,the inventory structure of manufacturing enterprises in our country is constantly being optimized,and scientific and reasonable inventory management methods have become an important means for manufacturing enterprises to maintain their competitive advantages and improve efficiency.Despite the impact of the Internet and the impact of the epidemic,manufacturing companies still have great potential and continue to innovate technologically under the dividends of digital transformation.Of course,this also puts forward higher requirements for manufacturing companies,which need to pay more attention to market demand feedback and realize intelligent inventory management.However,in the process of development,manufacturing enterprises still face some obstacles.On the one hand,due to the complex factors of consumer demand,it is difficult to predict the demand of enterprises.At present,enterprises generally use factors such as historical sales data,market trends,and product characteristics to make forecasts.However,due to fierce market competition and diversified consumer needs,existing demand forecasts still have certain limitations and uncertainties.On the one hand,enterprises generally use some common inventory management control methods,such as ABC classification,safety stock,economic batch and other indicators to control inventory.However,these methods lack scientificity and accuracy,which can easily lead to excessively high or low inventory,increasing the difficulty of inventory for enterprises,thus affecting the benefits of the upstream and downstream of the supply chain.Therefore,this paper conducts research on the market demand forecast and inventory control optimization of a large FMCG manufacturer Z.The research contents are as follows:(1)A demand forecasting model integrating self-attention mechanism and SARIMA-LSTM is proposed.First,process the experimental data,and comprehensively consider the influence of various factors on the demand,carry out feature construction,use the variance filtering method and the Spearman rank correlation coefficient method to carry out combined screening,and select relevant factors as optional demand influencing factors.Model.Secondly,the SARIMA model is suitable for fitting the linear relationship in the sequence,and the LSTM neural network is suitable for mining the nonlinear relationship in the sequence,making combined predictions,and integrating the prediction results in the neural network with the self-attention mechanism module,The self-attention mechanism can automatically learn the correlation between input elements,so as to obtain more accurate prediction values.Finally,the model is also verified in experiments.The results show that the demand forecasting model that integrates the self-attention mechanism and SARIMA-LSTM can significantly improve the accuracy and stability of market demand forecasting,achieve good forecasting results,and provide a data basis for inventory optimization.(2)A multi-level inventory cost optimization model based on the improved whale algorithm is proposed.Firstly,a mathematical model of multi-level inventory cost is constructed,and the cost structure and related parameter settings of the supply chain are analyzed in detail,so as to quantify and calculate various costs under constraints.Secondly,an improved whale algorithm is proposed,which introduces elite reverse learning mechanism,optimizes the convergence factor a,adjusts adaptive inertia weight and other strategies to improve the convergence speed of the algorithm and balance the local and global search capabilities.Finally,a comparative experiment is carried out,and the accuracy and effectiveness of the improved algorithm are verified by combining the traditional whale algorithm and the particle swarm algorithm.The results show that the improved whale algorithm has higher solution efficiency and practical value in multi-level inventory cost optimization problems,and the total cost of supply chain operations in one cycle is reduced by 10.9%,which is a feasible optimization method for the supply chain Management provides strong support and assistance.