Abstract: This study proposes two texture development models for Cod surimi gel. Dimensionality of the training data sets (12 patterns) of surimi gel strength are reduced to four eigen-gel patterns using an unsupervised method, the PCA method. Then we obtain an eigen-gel pattern for each cluster. Two texture models, consecutive and competitive-consecutive first order reactions are developed based on an eigen-gel pattern for each cluster. The correlation coefficient method is introduced to achieve a good identification rate of similarity between the two proposed methods and the eigen-gel pattern for each cluster.