An Explainable Machine Learning Model to Optimize Demand Forecasting in Company DEOS - Tesis

Aug 5, 2026

Upload by Finn

Accurate demand forecasting enables CompanyDEOS to plan resources efficiently, increasing productivity and supporting better market understanding. This research applies machine learning to optimize demand forecasting, training on collected data while using cross-validation to reduce the risk of overfitting given the available data rate. A time-series approach is used to predict future sales for the first trimester of 2021. The study evaluates the ML tool’s impact on forecast deviation using accuracy and bias indicators, quantifying both performance and systematic tendencies.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages14

Reading Time35 min

AccessFree to download

Introduction

Objectives

Literature Review

Universidad de Lima Facultad de Ingeniería Carrera de Ingeniería Industrial AN EXPLAINABLE MACHINE LEARNING MODEL TO OPTIMIZE DEMAND FORECASTING IN COMPANY DEOS Tesis para optar el Título Profesional de Ingeniero Industrial Gianella Valeria, Cabrera Feijoo Código 20170231 Jimena Mariana, Germana Valverde Código 20170640 Asesor Yvan Jesus, Garcia-Lopez Lima – Perú Abril de 2023 An explainable machine learning model to optimize demand forecasting in CompanyDEOS Cabrera Feijoo Gianella, Germana Valverde Jimena Facultad de Ingeniería y Arquitectura Carrera de Ingeniería Industrial Universidad de Lima Lima, Perú [20170231@aloe.ulima.edu.pe](20170231@aloe.ulima.edu.pe), [20170640@aloe.ulima.edu.pe](20170640@aloe.ulima.edu.pe) Garcia-Lopez Yvan Jesus Facultad de Ingeniería y Arquitectura Carrera de Ingeniería Industrial Universidad de Lima Lima, Perú [ygarcia@ulima.edu.pe](ygarcia@ulima.edu.pe) Abstract Nowadays, having an accurate demand forecast is extremely important as it allows the company to manage resources in an optimal way and thus achieve greater productivity. There is a large demand for accurate forecasting, and utilizing artificial intelligence can help companies gain a better understanding of their market. In this research presentation, Machine Learning (ML) is used to optimize demand forecasting. The data collected was trained and due to the available data rate, the Cross-Validation technique was used to avoid overfitting. Using time-series, it will be possible to predict future sales for the first trimester of 2021. Finally, the impact of the ML tool on the deviation of the company's demand forecast was evaluated using indicators of accuracy (forecast accuracy) and bias (forecast bias) . Keywords Demand Forecast; Machine Learning; Forecast Accuracy; Forecast Bias; Consumers Good Company 1. Introduction Due to the importance of an accurate demand forecast, the former Head of Demand Planning at the company was consulted. His explanation was given and will be utilized for the case study. The real name of the company will be omitted for the study due to confidentiality issues. Instead, the name CompanyDEOS will be used to refer to the company, and DemandChief will be used to refer to the former Head of the Demand Planning team. DemandChief (2021), indicates that: "demand forecasting is the mind of the business. It is the analytical portion of the solution. The forecast shows is believed to be what can be done and sold, for justified reasons, taking into account what has happened or what is known to occur, the forecast is made" . In the case ofa consumer goods company (CGC), these are affected by having an excessive number of products that do not rotate or a shortage of products due to high rotation. As Aktepeet al. (2021) mention: "Although inventories appear to be owned by the company, they have no return unless they are sold. Oversupplied raw materials for a production company occupy machinery and workers, creating additional labor costs as well as purchase costs." Excess inventory translates into a loss. Either due to excess or lack of inventories, it results in expired products and reduced sales. The main causes for an inaccurate demand forecast are the lack of data, unreliable data, peaks in demand, and the failure of taking into account life cycles. In addition, there is a great dependence on distributors because the vast majority ofCGC sells their products to them so they can then sell it to the final consumer. 1.1 Objectives The objective of this research is to improve the accuracy of the forecast of the monthly demand for deodorants in a CGC through the implementation of an alternative forecasting model and ML. 2. Literature Review The present research work was carried out with the need to find agile and precise methods to execute a demand forecast, for this case, in the category of deodorants. Currently, there are 5 types of demand forecasts

FAQs

Why is accurate demand forecasting critical for CompanyDEOS?

Accurate forecasting supports optimal resource management and productivity. It also helps anticipate sales reliably for planning decisions.

How does the research train and validate the machine learning model?

Collected data is used for training, and cross-validation is applied to avoid overfitting based on the data rate.

What metrics evaluate the impact of the machine learning tool?

Forecast accuracy and forecast bias indicators are used to measure both prediction performance and systematic deviation.

Aug 5, 2026

Upload by Finn