The Stock-flow Classification Model Using Ensemble Machine Learning Approach for Particle Board Inventory

This work used multiple machine learning algorithms, namely, Decision Tree, Naïve Bayes, and k-Nearest Neighbours to develop an ensemble data-driven classification model for the stock flow of panel board inventory categorization. The ensemble approach is when the model employs multiple machine learning algorithms to perform a certain prediction task. In this case, the ensemble approach was performed, along with the stand-alone approach using the aforementioned single machine learning algorithm against one another. The inventory data from the anonymous panel board manufacturer in southern Thailand were used to construct models. The 5-folds cross-validation technique was used to validate the performance of all models. Additionally, the handpicked rule of thumb had been used to select the set of hyperparameters for each algorithm. Then, the grid search was applied to select the best parameters for each predictor within the ensemble model. The result shows that using the ensemble approach provides better prediction accuracy compared to the stand-alone approach.