Research On Visual Classification Based On Multi_MAP Graphics

With the development of information society, the massive information has emerged, which has brought magnanimous complex multi-dimensional data. We need to class these data to achieve to analyze and observe the information. Therefore, the classification of the multi-dimensional data which receives more attention becomes the most important researching direction. Now, it has become the most important status in the pattern recognition domain. The visualized classification based on the multi-map graphics is the most popular method. The simple graphics and pictures are used to express the eigenvector of the data and information. So it is easy for us to find out the rule of the multi-dimensional data, and offer the tool for people to talk with the data.First,this paper studied the traditional method, and then pay more attention on the new classification of visualization. Introduced the theory and studied the application, it has offered the foundation for the next work. Parallel coordinate is a kind of technology used extensively in data visualization. For strengthen its visibility, we present the size of area, the length of cordite, the slope of cordite, the scale of each little area, the direction of the cordite and so on. The forth part of this paper expounded the classical technique based on the feature of multi-map. First, we expatiated the traditional arithmetic of classification and the classifier of simple model and the nearest graphic. We introduced the arithmetic principle of these two classifiers. Then, we advanced the conception of the multi-map of distance of different sorts and studied the feature choice of this multi-map.At last, we use this method to prove to class the classical data in the UCI dataset. In the breast tissue data, we use the multi-map of sorts distance to elect the feature of graphics, then take the area of the whole map as the feature for classing. The classification of this paper achieved ideal result. Each result of this classification are all on 80%. This experiment proved that our method is practicality. In the wine classification data, we use the subarea and the slope of multi-map based on distance of sorts to elect feature and choice the best feature for classification. Each result of this classification are all on 90%. The two experiment proved that our method of classification has the little computation which based on the same precision.