Facial Expression in Tourism Destinations Using A Deep Learning Approach

The development of the world of tourism is currently supported by information technology. New tourist attractions are introduced through social media. Apart from being easy and cheap to reach for visitors, tourist attractions must also be responsive to improved facilities and services. The step is through regular visitor surveys. One way that can be done is through visitors' facial expressions with the help of deep learning. In this research, the proposed contribution is using a convolutional neural network with 34 layers. The goal is to get good accuracy, but the computational burden of the training process is light. The image data used comes from secondary datasets to identify angry, disgusted, scared, happy, neutral, sad., and surprised classes with a total of 5,250 images. Data Augmentation technique is used to overcome the class imbalance. The results showed that the system could recognize facial expressions with an average accuracy of 99.38%. The average computational time for the training process to get the model is 25 minutes 23 seconds, with a testing time of 1–2 seconds. Experimental data testing results above 98%: error MSE 0.0445, RMSE 0.2110, and MAE 0.0150.