Moisture Content Prediction of Dried Longan Aril from Dielectric Constant Using Multilayer Perceptrons and Support Vector Regression | Science Publications

American Journal of Applied Sciences

Nipon Theera-Umpon¹  Kittichai Wantanajittikul¹  Sansanee Auephanwiriyakul¹  Sanong Amaroek¹ 

关键词: Dried longan aril;    dielectric constant;    moisture content;    bulk density;    regression models;    polynomial regression;    linear regression;    artificial neural network;    multilayer perceptrons;    support vector regression;    cross validation;    Dimocarpus longan;   

DOI  :  10.3844/ajassp.2010.1387.1392

学科分类：自然科学（综合）

来源: Science Publications

【 摘 要 】

Problem statement: Estimation of moisture contents of dried food products from theirdielectric constants was an important step in moisture measurement systems. The regression modelsthat provide good prediction performance are desirable. Approach: The Multilayer Perceptrons (MLP)and Support Vector Regression (SVR) were applied in this research to predict the moisture contents ofdried longan arils from their dielectric constants. The data set was collected from 1500 samples ofdried longan aril with five different moisture contents of 10, 14, 18, 22 and 25% Wet basis (Wb.)Dielectric constant of dried longan aril was measured by using our previously proposed electricalcapacitance-based system. The results from the MLP and SVR models were compared to that from thelinear regression and polynomial regression models. To take into account the generalization of themodels, the four-fold cross validation was applied. Results: For the training sets, the average meanabsolute errors over three bulk densities of 1.30, 1.45 and 1.60 g cm^-3 were 1.7578, 0.6157, 0.3812,0.3113, 0.0103 and 0.0044% Wb for the linear regression, second-, third-, fourth-order polynomialregression, MLP and SVR models, respectively. For the validation sets, the average mean absoluteerrors over the three bulk densities were 1.7616, 0.6192, 0.3844, 0.3146, 0.0126 and 0.0093% Wb forthe linear regression, 2nd, 3rd and 4th-order polynomial regression, MLP and SVR models,respectively. Conclusion: The regression models based on MLP and SVR yielded better performancesthan the models based on linear regression and polynomial regression on both training and validationsets. The models based on MLP and SVR also provided robustness to the variation of bulk density. Notonly for dried longan aril, the proposed models can also be adapted and applied to other materials ordried food products.

【 授权许可】

Unknown