Inventory forecasting model using genetic programming and Holt-Winter's exponential smoothing method
Ramakant Soni, D. Srikanth · 2017
Accurate and reliable inventory forecasting can save an organization from overstock, under-stock and no stock/stock-out situation of inventory. Overstocking leads to high cost of storage and its maintenance, whereas under-stocking leads to failure to meet the demand and losing profit and customers, similarly stock-out leads to complete halt of production or sale activities. Inventory transactions generate data, which is a time-series data having characteristic volume, speed, range and regularity. The inventory level of an item depends on many factors namely, current stock, stock-on-order, lead-time, annual/monthly target. In this paper, we present a perspective of treating Inventory management as a problem of Genetic Programming based on inventory transactions data. A Genetic Programming - Symbolic Regression (GP-SR) based mathematical model is developed and subsequently used to make forecasts using Holt-Winters Exponential Smoothing method for time-series modeling. The GP-SR model evolves based on RMSE as the fitness function. The performance of the model is measured in terms of RMSE and MAE. The estimated values of item demand from the GP-SR model is finally used to simulate a time-series and forecasts are generated for inventory required on a monthly time horizon.