Design of detection technique of lard in beef based on electrical properties

Abstract

Sensitive and accurate detection technology of lard contaminating tallow and beef is important to assure the halalness of food products, that needed by consumers. Therefore, it is important to develop reliable detection techniques of lard contamination. The using of electrical properties for detecting food product’s quality, data preprocessing and classification techniques are potential for designing lard detection technique. The objectives of this study was to obtain the frequency range for the measurement of electrical properties of lard, tallow and beef as experiment materials, that are significant for the detection of lard contamination in beef; and to develop a detection technique of lard contamination in beef based on the electrical properties and data mining, as well as to test its accuracy. In this study, a parallel plate made of copper by dimension of 20 mm x 10 mm separated at 5 mm distance was used. Na2SO4 anhydrous was used to reduce the water content in the sample. The measurement of electrical properties was done in room temperature (26-27oC) with frequency range of 0,100-0,999 MHz, 1,000-3,799 MHz and 3,80-5,00 MHz. The measured electrical properties were conductance, impedance, and capacitance. The test of fatty acid profile by using gas chromatography was used as a comparison in this study. The preprocessing techniques that used were mean, median, smoothing, normalization and discretization. Techniques of classification that were used to differentiate lard from tallow and palm oil were naive bayes, decision tree C4.5, simple logistic, support vector machine (SVM) and multilayer perceptron (MLP). In order to predict lard contamination in beef, the classification techniques including linear regression, SMO regression, support vector regression (SVR), and MLP were employed. The results showed that the parallel plate from copper by dimension of 20 mm x 10 mm with 5 mm distance was effective to acquire electrical properties data. The use of Na2SO4 anhydrous to absorb water from fat samples and frequency at 3,80-5,00 MHz to measure electrical properties resulted valid data. The order of electrical properties that influence to lard detection were impedance, capacitance and conductance. The use of preprocessing techniques normalization and classification technique MLP were able to differentiate lard from tallow and palm oil with correctly classification of 87,52% and root mean squared error (RMSE) of 0,2394. The use of preprocessing normalization and classification technique MLP were able to predict the contamination of lard in beef with correlation coefficient (r) of 0,9985 and RMSE of 1,9708. The future of the combination of materials electrical properties with selected preprocessing and classification techniques can be used to develop a detection of lard contamination.