Application of Clustering Analysis with Unsupervised Technique on Fish Oil Samples
Abstract
Fish oil contains many fatty acids (FAs) important for human health. Each fish oil tends to have a different fatty acid content. This study aims to group the fish oil content profile in several samples, such as keting fish oil (KFO), catfish oil (CFO), and pomfret fish oil (PFO), based on GC-MS analysis data with unsupervised techniques. GC-MS is a method that can be used to identify the fatty acid content in fish oil. Fish oil is extracted using the dry rendering method combined with a hydraulic press to obtain the oil, then derivatised before being analysed on the GC-MS instrument. Due to the multivariate data from GC-MS, multivariate statistical techniques are required to effectively group the fish oil samples based on their Fatty Acid profiles. Principal Component Analysis (PCA) and Cluster Analysis (CA) chemometrics are unsupervised techniques that can group multivariate data by displaying plot scores and dendrograms of sample analysis results. The fatty acid content of keting fish, catfish, and pomfret fish oil has three dominant compounds in sequence, namely oleic acid, palmitic acid, and stearic acid, with different percentages. The
grouping profile of fish oil was successfully determined by PCA, total variance explained by the first four components (PC4) of 99.4%, and CA, which produced three groups based on the fatty acid content of fish oil.