Transfer Learning-guided Demand Forecasting Study For Fabric Spare Parts

The overall environment of the apparel industry is complex.In order to cope with changing market and product demands,apparel companies should strive to improve their fabric demand forecasting capabilities and inventory management.As the primary element of garment production,fabric demand problems are mainly reflected in the instability of inventory and uncertainty of demand: fabric shortages can lead to production lines coming to a standstill,which in turn affects the supply of finished products and corporate sales,harming the economic development of the business.At the same time,due to the short lifecycle and highly variable nature of garments,excess fabric spares can increase cost pressures and obsolete fabrics can lead to waste,increasing the financial risk to the company.The uncertainty of fabric demand therefore poses a huge challenge for apparel companies.Moreover,the complex composition of apparel fabrics and the different sizes and types of fabrics demanded by companies at different times make it difficult to solve this problem with traditional manual forecasts and single intelligent forecasting models.In order to cope with the difficult problem of forecasting demand for various sizes and types of fabrics in apparel production,and in response to the uncertainty and difficulty in forecasting fabric demand due to the short product life cycle of apparel enterprises and the changing business environment,the following are the main research contents of this topic for constructing a demand forecasting model applicable to apparel fabric spare parts:1.Forecasting demand for fabric spare parts is an important task,but the accuracy of forecasting may be limited due to the fact that demand for fabric spare parts is affected by many factors,such as equipment life,production schedule,and market demand.At the same time,there is a wide variety of spare parts,and predicting the demand for each spare part requires the creation of a predictive model,which increases the computational cost.In addition,the special fabrics of clothing manufacturing enterprises have the problem of less spare parts data,resulting in insufficient training data to support high-precision prediction,while the training of neural networks requires a lot of computing power,and the increase in the number of iterations of the predictive model will bring higher time costs.In order to cope with these problems,an ordered long-term short-term memory network prediction model based on transfer learning guidance is proposed.The model can learn additional knowledge from neuron sequencing and play an important role in small sample data.Then,the knowledge learned by the BI ON-LSTM network in spare parts data(source domain)similar to the target domain is transferred to the prediction process of other spare parts data(target domain)through model migration,so as to improve the prediction performance.This method effectively reduces the cost of repeated training,improves prediction accuracy and iteration speed.2.With the development of technologies related to intelligent manufacturing,the requirements for fabric spare parts demand prediction accuracy are getting higher and higher.With the continuous development in the fields of big data technology,artificial intelligence and machine learning,many new methods and algorithms for time series forecasting have been applied in practice to improve the forecasting accuracy and precision.Therefore,based on Chapter 3,a composite model forecasting method based on the attention mechanism and BI ON-LSTM is proposed,which integrates the attention mechanism with the BI ON-LSTM model to assign high probabilities to key information and improve forecasting accuracy,and then combines the trend function to mine the hidden trend information in the information,and then uses the tensor migration technique to obtain more reasonable field of related The initial values of the parameters,applying this knowledge to the model,can reduce the risk of overfitting of the fabric spare parts demand forecasting model and over-initialisation of the model,while improving the forecasting accuracy and stability of the forecasting model.After several rounds of experimental comparison,the results show that the prediction accuracy of the model is better than that of the commonly used time series prediction models such as RNN,LSTM and GRU,and it has important application significance for the fabric spare parts demand prediction of apparel enterprises.In summary,this topic investigates two forecasting models to address the short product life cycle of apparel companies and the frequent changes in the business environment that lead to uncertainty and difficulty in forecasting fabric demand.Both models aim to improve the accuracy and stability of fabric demand forecasts for apparel companies,thereby reducing cost pressures and minimising the risks associated with overstocking or wasting fabric materials.