| Under the Internet era,advances in technology have contributed to the abundance of data sources and reduced the cost of data collection,resulting in the expansion of databases and the explosion of data dimensions.Various types of data,such as commercial transaction data,clinical medical data,Web documents,multimedia data,and user comment data,have dimensions of as few as tens of dimensions,as many as thousands of dimensions,and even higher.The widespread and continuous increase of high-dimensional data makes the high-dimensional data mining has important significance and prospects.However,on the one hand,due to the existence of dimensionality disaster,the difficulty of high-dimensional data mining is increased,and data must be processed by other means.On the other hand,the limitation of human perception limits the recognition of data by data scientists and experts know.The combination of data visualization and dimensionality reduction method makes it possible to intuitively convey the key features of data through graphical representation and further understand the sparse and complex high-dimensional datasets.However,visual analysis of high-dimensional datasets requires special charts,understanding and implementation of dimensionality reduction algorithms,which greatly increases the threshold for visual analysis of high-dimensional data.At the same time,the analysis of high-dimensional data in the process of analysis of the uncertainty and repeated parameter adjustment process also increased the difficulty of visual analysis of high-dimensional data.Under this background,this paper designs and implements a high-dimensional data visualization platform based on nonlinear dimensionality reduction.The back-end framework adopts Django,the front-end framework uses bootstrap,and the implementation of parallel coordinate and scatter matrix based on D3.js,the dimension reduction algorithm includes PCA,t-sne and Large Vis.Data Storage Use the JSON object array to store uploaded data samples.On this basis,the platform helps users to analyze the correlation between dimensions of different dimensions of data by means of interactive visualization experience,parameter adjustment function and descriptive statistics analysis,simplifying the operation flow of users,reducing the learning cost and breaking High-dimensional data visualization threshold.In the last chapter of the dissertation,we test the function of the platform and test the usability and practicability of the high-dimensional data visualization platform by using different types of high-dimensional data sets to ensure that the platform meets the user’s needs and achieves the designed goals. |
