Time-Series Data Forecasting and Approximation with Smoothing Technique

Data mining process on weather forecasting in recent years has taken a great contribution toward human life, every aspect of it can be taken as consideration based on what kind of problem to be solved. One of its usages is to predict or forecast how much rain that will happen in some certain area just by studying the historical data of the precipitation, temperature, relative humidity and so on. Another example of weather forecasting benefit is to know what kind of crop should be planted on that time into how much profit that can be gathered from the crop harvest, it's more like economical optimization to apparently. The purpose of this research is to find out whether the Centroid Decomposition approximation approach is going to be a good alternative or not in terms of enhancing the forecasting results produced by the initial forecasting method and to prove whether the smoothing technique addition to it makes it a better procedure or not. The approximation results compared to several forecasting methods namely Linear Extrapolation, Triple Exponential Smoothing, and Moving Average. Several scenarios of forecasting interval have been made to make a fair comparison and deeper analysis of each method used in this study. The result shows that generally, Centroid Decomposition (CD) approximation process can enhance the initial forecasting method's result yet still unstable in terms of the RMSE score. But the smoothing technique adds to the CD procedure showing a great result as the RMSE successfully lowered in all given scenarios.