Discovery Of Critical Control Points In The Processing Of Agricultural Product By Support Vector Machine

Support vector machine (SVM) is a frontier technology in artificial intelligence. The intelligent decision for agricultural product processing based on Hazard Analysis and Critical Control Point (HACCP) is still at the starting stage. The present study firstly applied artificial intelligence technology to solve the problem of discovering critical control point in the processing of agricultural product, and SVM was applied to discover the critical control point in HACCP. The SVM algorithm was meliorated and optimized, and the SVM model optimized by genetic algorithm and an incremental SVM model were developed. The present provided some valuable results for the intelligent discovery of critical control point in HACCP.The main content and innovations were as follows:1. SVM was firstly applied for the intelligent discovery of critical control point in HACCP. The support vector classification model was constructed by weighted SVM algorithm, and the experiment results for the classification of critical control point were analyzed and evaluated. (see Chapter 3)2. Some innovative studies were conducted in allusion to the problem of the selection SVM parameters. The selection of penalty factor C and kernel function parameters were lack of theoretical guide, therefore they were selected mostly by man-made experiment, which were easily lead to local optimization. The present study applied genetic algorithm to obtain global optimized penalty factor C and kernel function parameters, and it was validated for the discovery of critical control point in the processing of fresh-cut vegetables. (see Chapter 4)3. A new incremental SVM method was developed aimed at solving the problem that the SVM training speed decreased when adding new samples. The calculation of distance between two samples in kernel space was simplified by conditional positive definite kernel function, and the relative boundary vectors set was considered as sample training set, and then the incremental SVM algorithm was achieved. The developed method was validated. It improved the model training speed and assured the model precise as well. (see Chapter 5)4. Fuzzy fault tree and artificial neural network (ANN) were used to classify the critical control points, and the results were compared with that obtained by SVM.The evaluation of different classification models was discussed. (see Chapter 6) The present study applied SVM to solve the problem of discovering critical control points in HACCP. It provided contributions for the intelligent control technology. In addition, it enriched and expanded the application scale of SVM. As spot of light, the current research findings had integration application in the first Chinese self-owned HACCP software easyHACCP^?.