Demand forecasting using the Monte Carlo method

Abstract

With the constant market’s change and the competition that exists, having a robust forecasting model becomes essential for the companies to be able to compete against other companies and apply it to the Planification and Purchasing Departments to satisfy the customers trying to anticipate what will happen in the market. In this project, a demand forecasting model has been developed using the ARIMA method. The model input is the company's historical sales data and future forecasts are obtained for the selected period. After the demand forecasting, the Monte Carlo simulation has been designed and applied due to obtaining the predictions for products' containers, products' quantity, etc. This has been done using the company's previous years' frequencies. At the same time, three different programs have been developed; an algorithm to find relationships between orders, a program to segment and classify customers, and, finally, an algorithm to group by similarity, in the project’s case, customers, but it can be applied to other aspects. To verify the ARIMA model and reduce the pandemic’s effect, three different scenarios have been planned, and their errors have been calculated for each of them. The results obtained are very promising giving low errors meaning that the model is adequate and can be applied in the company to do little predictions. Monte Carlo's simulation results are not very favorable due to the high errors obtained. In future projects, the company will study the possibility to add more parameters to the algorithm, such as neural networks (artificial intelligence) to improve the project and apply it