Extraction of Toast Packaging Design Elements Using Long Short Term Memory-Neural Network with Kansei Engineering Approach

Packaging is one of the factors that can increase consumer satisfaction. Packaging performance can be realized through the right combination of design elements based on design concepts that are in accordance with consumer preferences. The process of extracting design elements quantitatively is an important thing to do in minimizing subjectivity. The purpose of this research is to determine the optimal packaging design elements quantitatively with the Kansei Engineering approach. Artificial Neural Network (ANN) and Long-Short Term Memory Neural Network (LSTM-NN) ore some method that is able to predict packaging design elements. The novelty of this research is comparing the accuracy of the two methods so that it can be a reference in selecting the appropriate element extraction method. Design elements prediction was carried out by determining factors such as packaging shape (X1), packaging material (X2), packaging features (X3), image elements (X4), design style (X5), design surface (X6), and lock opening (X7). Design elements determining using ANN method for Practical-Unique concept is obtained X1.2 is Horizontal Beam, X2.5 is Paper and Polymer, X3.2 is Window, X4.2 is None, X5.1 is Trendy, X6.1 is Direct, X7.4 is Perforation. While the design elements for the Practical-Unique concept using the LSTM-NN method obtained X1.2 is Horizontal Beam, X2.2 is Brown Kraft, X3.4 is Cutlery & Handle, X4.2 is None, X5.1 is Trendy, X6.3 is None, X7.3 is Die cut. It is concluded that the results of the LSTM-NN method show better element design due to high training accuracy and having complete variables compared to ANN.