Association Rule Mining In Food Safety Supervision

A lot of food safety detection data are delivered in the daily food safetysupervising. Potential useful information can be gained by data analysis and thus canprovide reference information to the decision making in the food safety supervising.According to the characteristic and needs of food safety supervision, this thesisintroduced data mining techniques into food safety detection data analysis. The maincontributions of this thesis are as follows:1. Preprocessing methods including data selection, cleaning, discretization andhierarchy were introduced based on food safety detection data characteristic, andmade preparation for next mining step.2. A new association mining method based on multidimensional, multi-hierarchyand multivalued food safety detection data was introduced and applied. Theresults showed that the algorithm was effective and it could provide usefulreference information to food safety supervision.3. A new integrative filter approach for finding interesting association rules wasintroduced. It took rule restrict redundancy, negative correlation redundancy,inner-dimension redundancy and multi-hierarchy redundancy into considerationand provided solutions. Applying it to the food association rules, the resultsshowed that the approach was effective.