Research On Key Issues Of Time Series Forecasting Based On LASSO And Recommendation System And Their Application In Supply Chain Management

Under the background of the digital era,digital technology has brought new forms to the development of the supply chain,playing an increasingly important role in supply chain management decisions.Obtaining effective information from massive data and making scientific management decisions are important technical methods to optimize the supply chain.Time series based on least absolute shrinkage and selection operator(LASSO)technology,as a prediction tool that can effectively process high-dimensional data,is widely used in the supply chain management.This dissertation focuses on the Vector autoregression-LASSO(VAR-LASSO)models and proposes prediction models using different sparse regular optimizations(Tikhonov,Ivanov)to efficiently process high-dimensional data.Meanwhile,the dissertation designs a first-order sparse optimization algorithm,suitable for solving high-dimensional data.What's more,with experimental analysis of short-term solar power,the results show that the forecasting capability of VAR-LASSO models is good and models with Ivanov can be applied in engineering.Matrix factorization is widely studied in academia.It is widely used in recommendation systems of e-commerce,logistics and other fields.This dissertation focuses on the collaborative filtering model based on matrix factorization technology.The dissertation proposes a collaborative filtering model with a minimax concave penalty(MCP),using Variable Bregman Stochastic Coordinate Descent method(VBSCD)which can achieve parallel computing to solve the difficulty of algorithm parallelization in the development of recommendation technology.With experimental analysis of Amazon rating data,the results show that the MCP-CF model performs better than the L2-CF model.Parallel algorithms will play an important role in the development of recommendation technology for solving non-convex and non-smooth collaborative filtering problems.