Materials inventory optimization using various forecasting techniques and purchasing quantity in packaging industry

Abstract

Purpose: This paper studies the problem that occurs on material purchase quantity in price uncertainty situation. Larger buying quantity when the price at high will increase the purchase amount while smaller buying quantity could risk the inventory level. The decision on the purchase quantity of a cycle takes future price as input from price prediction output.Design/methodology/approach: This paper examines five price prediction models, Classification and Regression Tree (CART), Random Forest Regressor (RFR), Support Vector Regressor (SVR), Seasonal Autoregressive Integrated Moving Average (SARIMA) and Long Short Term Memory (LSTM) to predict four Petrochemical products, Linear Low Density Polyethylene (LLDPE), Low Density Polyethylene (LDPE), Biaxially Oriented Polypropylene (BOPP) and Cast Polypropylene (CPP) using dataset built from weekly datapoints from January 2020 to June 2023. The most performing model is validated with data from July 2023 to September 2023 where the prediction result is fed into Linear Programming Simplex method to minimize the amount of purchase by making advanced or postpone orders.Findings: Result that RFR performs higher at most products tested, while SVR performs higher in LDPE product. The fitting of RFR and SVR models prediction, as predicted price to Linear Programming that decides optimum purchase quantity, delivers a total 2.2% of purchase amount reduction compared to original purchase quantity reflecting base scenario issued by the planner.Research limitations/implications: This study does not include additional prediction factors such as freight cost and the hyperparameters tuning studies on the existing factors.Originality/value: The novelty of this paper is prediction value is followed up by an optimization model that would guide the Procurement team decisions for future anticipation because imported raw materials should be purchased ahead of time. This research will provide a scientific approach input that would counterbalance or strengthen decision making that is typically made by individuals owning years of experience. This combined approach is rarely researched and has not been done to polymer products.