A Shoreline Change Prediction Technique Combining Physics and Data-driven Model

In modern engineering, Artificial Intelligence (AI) and several data analysis techniques are frequently used and developed in various fields. These quantitative approaches, however, are somewhat focused on the assumption that sensor data properly expresses the physical phenomenon. Besides they still have limitations such as nonlinearity, different environmental condition and complexity of response. Another issue is that the data can be obtained through experiments, but due to the constraints of time and cost of experiments, obtaining a large amount of data that may be able to fully explain diverse natural occurrences is impossible. To deal with the aforementioned issues, we propose shoreline prediction techniques using a combination of physics and data analysis models. The physical coefficients of the existing differential equation are optimized through a genetic algorithm and approximate solution is obtained through the Euler method. This was used as prior knowledge and combined with a data analysis model to predict the shoreline position. As a result of the experiment, when there was enough training data, the performance of data analysis model was better than that of the proposed method, but the performance of the proposed method was better in situations where the training data was insufficient.