Track: Machine Learning

Abstract

In this work, it is necessary to analyze the increase of Back Order in the attention of crossdocking orders in the attention of Homecenter customers due to the lack of definition of purchase planning processes, resulting in logistics costs, fill rate charges and low service level. Thus, it is intended the companies that handle high volumes of inventory and constant orders should have a forecast plan to cover possible stock-outs.

The main purpose of the research is to explain a way to prevent stock-outs using an artificial intelligence model, based on historical sales data of a medium-sized company that manages inventories, as well as to determine the machine earning model to predict and reduce backorders.

For the data analysis, the Orange software was used, where the data was trained with different artificial intelligence models such as Decision Tree, Support Vector Machine, Random Forest, and neural networks. The most accurate model was defined according to numerical indicators such as the confusion matrix, the area under the curve (AUC) and the ROC curve analysis. Thus, we opted for the neural network model, which presented the most accurate data.

Finally, the results are presented and a suggestion is made at the management level regarding decision making in the supply process. For this purpose, itis considered pertinent to delve into the subject of the variables that influence the accumulation of backorders.