Implementación de Machine Learning para predicción de costos de camarón hacia China con exportadores de Guayaquil.

Abstract

Exports are currently a logistical challenge for many companies. Shrimp exporters must adapt to the new challenges they face and companies that handle these exports are always facing cost variations in their operations. These cost increases force companies to raise their prices and can even hinder exports by cutting financial resources for shrimp farmers. Within the study of data, being able to know and implement Machine Learning tools to predict the costs of shrimp exports to China becomes a very important resource. With the data provided by a forwarder in Guayaquil. The investigation has determined which algorithm was the most appropriate for its application, and then applied the K Means and KNN algorithms in order to segment customers into groups with similar characteristics and predict the logistics costs associated with exports. The results showed a segmentation between “Premium Customers” and “Normal Customers”, which will allow the development of differentiated strategies. In addition, a predictive model with a low root mean square error (RMSE) was developed, suggesting that the model is accurate in estimating export costs. With this, it is possible to anticipate costs and know that sales costs can be reduced to obtain higher profits. Finally, future improvements were proposed, such as the inclusion of additional variables and the integration of other Machine Learning algorithms, to further optimize the model. This study contributes to the optimization of logistics in shrimp exports and also logistical improvements for both the forwarder and the exporter. The forwarder will receive a service adequate to the costs he is paying by being considered to the group of assigned clients and the forwarder will be able to reduce his costs through different strategies applied by knowing the predictions of the model.