Research On Classification Methods Of Mechanical Products System Based On Similarity Measures

In this dissertation, the problems of pattern classification of mechanical products system is researched based on similarity measures. After the weightiness of product attributes being researched, a new method is established to calculate the weight of attributes. The method is applied in Weighted Distance Coefficient and System Similarity to measure similarity degree. It enhances the objectivity and accuracy of the similarity measures. The Main Component Analytical Method is applied to choose the attributes and simplify the complexity of calculation.Cluster Analysis is argued to process based on Weighted Distance Coefficient and System Similarity methods to enhance the validity of the clustering result. Classification and decision-making function is established to solve the product instance. The Minimum Weighted Distance Classifier and the Maximum System Similarity Classifier is designed to enhance the precision of classification. The Minimum Risk Decision-making based on system similarity is founded to minimize the loss brought by decision-making.The research improves the objectivity and accuracy of the classification and decision-making of mechanical products system; it also enhances the automatization level of classification.