<![CDATA[Prediction or forecasting is an action that aims to find out future events based on indicators that influence an event. Consumer goods are products or goods purchased by people or households that are intended for direct consumption in the sense that they are not for further production purposes. Based on this, serious handling is needed to maintain the state of the Indonesian economy, especially in the industrial sector. Predicting the value of consumer goods imports is a step in finding out the value of consumer goods imports in the next period so that the government has a reference in determining policies. In this study, the prediction of the value of consumer goods imports was carried out based on factors that influence the value of consumer goods imports based on research in the field of economics. This study uses the Radial Basis Function Neural Network (RBFNN) method using a combination of clustering methods, namely Fuzzy C-Means Clustering to improve method performance. The RBFNN method is the best method used in predicting future data based on previous research and the FCM method is a clustering method that is able to overcome ambiguity in the prediction process. This study proves that the Fuzzy C-Means method is effective in optimizing the performance of the Radial Basis Function Neural Network method with a comparison of MAPE values in each combination, namely RBFNN - FCM 15.73%, RBFNN - K-Means 16.87% and RBFNN - Random centroid 17.70%. The learning rate parameter is directly proportional to the RBFNN - FCM model where the greater the learning rate, the better the model performance, indicating that the model does not need to do in-depth learning to recognize data patterns. In contrast to the fuzzification parameter which increases accuracy when the fuzzification value is lowered, indicating that the model does not require a very vague approach to recognize data patterns. The best architecture is 8 - 4 - 1 with a fuzzification parameter value of 1.5, a learning rate of 0.3 and a threshold error of 0.3 produced by a combination of RBFNN and FCM.]]>
Copyrights © 2025
