PERBANDINGAN EKSTRAKSI CIRI DENGAN METODE SUPPORT VECTOR MACHINE PADA ELECTRONIC NOSE UNTUK IDENTIFIKASI TAHU MURNI DAN TAHU BERFORMALIN; COMPARISON FEATURE EXTRACTION WITH SUPPORT VECTOR MACHINE METHOD IN ELECTRONIC NOSE FOR IDENTIFYING PURE AND FORMALINE TOFU

Nurbaeti, Laely, Danang Lelono

2015 | Skripsi | FMIPA

Tofu is a traditional food ingredients that easily obtained with hight protein. High content of water and protein in tofu, make tofu susceptible to decomposition. This makes some rogue trader to add formalin in tofu. Tofu has a distinctive aroma which is formed of several chemical compounds. Formaline is a chemical compound that is volatile and has a scent. Electronic nose has the ability to analyze samples with complex compositions thus characterizing and qualitative analysis of the sample. After the discovery of e-nose, the aroma can be measured using a sensor array contained in the e-nose and further processed using pattern recognition. Scent pattern is then used for the process of feature extraction using integral, fractional, and relative. After the feature extraction, then the variable reduction performed by PCA and used for the training process. Optimized parameters are then tested on a random sample. Based on test results using 3 methods of feature extraction that is integral, relatively, fractional produce different accuracy. Data samples with integral feature extraction yield 73,3 % accuracy, relative yield 80 % accuracy, and fractional yield 85% accuracy. Whereas if the combined third feature extraction produces an accuracy yield 83.3%.

Kata Kunci : Support Vector Machine (SVM); linearly classifier; electronic nose; tahu.