Skripsi
KAJIAN PEMODELAN DAN SIMULASI PREDIKTOR DRY RUBBER CONTENT (DRC) MENGGUNAKAN ANALISIS REGRESI
Dry Rubber Content (DRC) is a fundamental aspect of rubber processing materials (BOKAR) for which, to date, no accurate and rapid method of analysis has been developed; a predictive model using regression analysis could be a solution to this issue. This study aimed to develop a mathematical model capable of predicting DRC based on sample age under conditions consistent with the established research parameters and to evaluate the model’s performance when applied to samples of varying weights. Linear regression and polynomial regression models from the 2nd to the 6th order were constructed and evaluated using k-fold cross-validation, encompassing the entire calibration dataset and validation datasets derived from sample weights of 200 g, 500 g, and 1000 g. Model performance was assessed using R² (R-square), RMSE (Root Mean Square Error), MRD (Mean Relative Deviation), RPD (Residual Predictive Deviation), and SMAPE (Symmetric Mean Absolute Percentage Error), complemented by paired sample t-tests and equivalence paired sample t-tests. The 3rd order polynomial model demonstrated high accuracy and stability; however, it was not suitable for predicting DRC in samples exceeding 200 g. In k-fold cross-validation, the model achieved mean values of R² = 0.998637, RMSE = 2.521977, MRD = 0.038647, RPD = 8.125425, and SMAPE = 0.038461, with corresponding standard deviations of 0.000432, 0.373841, 0.011188, 1.90344, and 0.010256. For the overall calibration data, the model yielded R² = 0.988505, RMSE = 1.874113, MRD = 0.027123, RPD = 9.654489, and SMAPE = 0.027114, while validation results showed R² = 0.997302, RMSE = 0.826966, MRD = 0.012496, RPD = 23.580174, and SMAPE = 0.012602. Statistical tests indicated no significant differences and confirmed equivalence, supporting the robustness of the model.
No other version available