A high-dimensional data subspace visualization exploration and analysis method and system
Through the visualization method of map metaphor, the characteristics of high-dimensional data subspace are extracted and the map is constructed, which solves the complexity and redundancy of the subspace analysis of high-dimensional data, and realizes efficient subspace exploration and dimensional decision-making guidance.
Patent Information
- Application Number
- CN202111041998.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-09-07
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2041-09-07
AI Technical Summary
High-dimensional data subspace analysis faces the problems of huge exploration space, complex dimensions and data patterns, and lack of effective exploration directions. The output of existing algorithms is prone to redundancy and lacks guidance on dimensional selection, which is especially suitable for two-dimensional subspaces.
Using the visual exploration and analysis method of high-dimensional data subspace visualization, subspace features such as subspace similarity, clustering, data mode and dimensional mode are extracted through the visualization form of map metaphor, and subspace map map is constructed, using grid tiling and map metaphor Determine the layout and provide interactive exploration toolkit D3 to draw subspace maps and auxiliary views.
Support users to efficiently explore high-dimensional data subspaces, understand subspace relationships, reveal the interaction between dimensions and data patterns, provide guidance for dimensional decision-making, and help users quickly identify representative subspaces and their differences.
Smart Images

Figure CN114048369B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of visualization and human-computer interaction, and in particular to a high-dimensional data subspace visualization exploration and analysis method and system. Background Art
[0002] As an important data type for recording abstract information, high-dimensional data is very common in daily life. Its data samples have multiple attributes, such as environmental monitoring data containing multiple indicators, personal files containing multiple information, etc. Subspace refers to any low-dimensional space in a high-dimensional space. In order to avoid the interference of irrelevant dimensions and redundant dimensions and the influence of dimensionality disaster, it is necessary to choose a suitable subspace when analyzing high-dimensional data. Subspace analysis is an important part of high-dimensional data visualization. However, subspace analysis faces problems such as huge exploration space, complex interactions between dimensions and data patterns, and lack of exploration direction. Although researchers have proposed various algorithms to find subspaces containing data clusters, the outputs of these algorithms are prone to redundancy and need to be further organized with the help of visualization. In addition, they do not provide guidance on dimension selection. There are also some methods that aim to guide users in subspace exploration. However, they either rely on inefficient manual planning or are only applicable to two-dimensional subspaces. In general, although subspace analysis has been widely explored, the above three problems have not been well solved. Summary of the invention
[0003] In view of the defects existing in the prior art, the purpose of the present invention is to provide a method and system for visual exploration and analysis of high-dimensional data subspaces, which supports users to efficiently complete the exploration and analysis of high-dimensional data subspaces in a visual form of map metaphor.
[0004] To achieve the above purpose, the technical solution adopted by the present invention is as follows:
[0005] A high-dimensional data subspace visualization exploration and analysis method, comprising:
[0006] S100, searching for valuable subspaces from a high-dimensional data set and extracting their features, wherein the features include subspace similarity, subspace clustering, data patterns, dimensional patterns, and representative subspaces;
[0007] S200, constructing a subspace map based on the extracted features, including determining anchor point locations and subspace traversal order, determining a layout by grid tiling, and adding additional map metaphors;
[0008] S300: Draw the subspace map and other auxiliary views, and complete subspace exploration through interaction with the user.
[0009] Further, in the method as described above, S100 includes:
[0010] Based on the subspace clustering algorithm, the subspace with non-uniform distance distribution pattern of data items is searched out as a valuable subspace.
[0011] Further, in the method as described above, S100 includes:
[0012] (1) Extracting subspace similarity: Calculate the k-NN list of each data item in each subspace obtained by the search, take the similarity between the k-NN lists of two subspaces corresponding to a certain data item as the similarity of the two subspaces on the data item, and average the similarities of the two subspaces on all data items to obtain the similarity of the two subspaces;
[0013] (2) Extract subspace clusters: Use the DBSCAN clustering algorithm to perform two-layer clustering on each subspace obtained by the search, and the clusters in the first layer serve as the algorithm input for the second layer;
[0014] (3) Extracting data patterns: Evaluate data items with similar k-NN lists in all subspaces of the same cluster as the data patterns of the cluster;
[0015] (4) Extracting dimensional patterns: Calculate the number of times each dimension appears in all subspaces of the same cluster. Dimensions whose number of appearances in a certain cluster reaches the corresponding high preset number and whose number of appearances is lower than the corresponding low preset number are taken as the dimensional pattern of the cluster. The number of times a dimension appears in the neighboring subspace of a certain subspace is taken as the dimensional pattern of the subspace.
[0016] (5) Extract representative subspace: Select the subspace with the highest average similarity to other member subspaces as the representative subspace.
[0017] Further, in the above method, in S200, determining the anchor point position and the subspace traversal order includes:
[0018] (1) Determine the anchor point location: The reasonable locations of different countries and islands are obtained by drawing the dimension reduction projection of all subspaces. For each country, its capital is regarded as the anchor point. For each island, it is regarded as the anchor point and its location on the hexagonal grid is determined.
[0019] (2) Determine the subspace traversal order: Determine the order of all countries’ capitals and islands at the first level, and determine the order of all cities in the same country at the second level. That is, create an empty list and add the city with the smallest average dissimilarity to other cities in the list to the list until all cities in the same country are included. Different countries are traversed separately.
[0020] Further, in the above method, in S200, the layout is determined by grid tiling, including:
[0021] (1) After placing the anchor points according to the determined anchor point positions, the cities in the same country are placed according to the determined subspace traversal order, that is, a position queue is established, and each subspace passed into the hexagonal grid is accommodated by the first grid in the position queue, and the queue is continuously updated. After a grid is used, adjacent grids that meet predetermined conditions are added to the queue in a clockwise order, and the predetermined conditions include not being occupied by other subspaces and not being a disabled grid;
[0022] (2) If two adjacent subspaces belong to the same cluster, the current position queue is updated; otherwise, the current position queue is cleared and a new position queue is established for the new cluster;
[0023] (3) After determining the layout of all countries and islands, the layout is made more compact by narrowing the gaps between them and the remaining grids are enlarged in the center to fill the entire view space, with the subspace similarity reflected by color.
[0024] Furthermore, in the above method, in S200, an additional map metaphor is added, including:
[0025] The representative subspace of each cluster is displayed as the capital or provincial capital. For each city / region, its natural factor is the dimensional pattern and its landscape is the data pattern. After the user selects the starting and ending cities, the corresponding land and sea routes are obtained by calculating the routes with the least dissimilarity through the cities. The route network is made more concise through the minimum spanning tree algorithm.
[0026] Further, in the method as described above, S300 includes:
[0027] The subspace map and other auxiliary views are drawn based on the visualization toolkit D3, and a subspace list view and a map detail view are provided to assist users in completing exploration and analysis. The subspace list view is used to display the dimensional information of the subspace, and the map detail view is used to display the dimensional pattern and data pattern of the subspace; at the country or province level, the map detail view is used to display the data pattern and dimensional pattern of the cluster, and at the city level, the map detail view is used to display the data pattern and dimensional pattern of the selected subspace.
[0028] A high-dimensional data subspace visualization exploration and analysis system, comprising:
[0029] An extraction module, used for searching valuable subspaces from high-dimensional data sets and extracting their features, wherein the features include subspace similarity, subspace clustering, data patterns, dimensional patterns and representative subspaces;
[0030] The construction module is used to construct the subspace map based on the extracted features, including determining the anchor point location and subspace traversal order, determining the layout by grid tiling, and adding additional map metaphors;
[0031] The drawing module is used to draw the subspace map and other auxiliary views, and complete the subspace exploration by interacting with the user.
[0032] A storage medium stores a computer program, wherein the computer program is configured to execute a high-dimensional data subspace visualization exploration and analysis method when running.
[0033] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a high-dimensional data subspace visualization exploration and analysis method.
[0034] The beneficial effects of the present invention are as follows: the present invention supports users to efficiently complete the exploration and analysis of high-dimensional data subspaces in a visualized form of map metaphor, provides an overview of the exploration space, helps users understand the relationship between subspaces, and builds their mental maps; reveals the interaction between dimensions and data patterns, and provides guidance for dimensional decision-making; guides users to browse the exploration space, so that users can understand representative subspaces and their differences in a short time. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 A schematic diagram of a flow chart of a high-dimensional data subspace visualization exploration and analysis method provided in an embodiment of the present invention;
[0036] Figure 2 A user exploration workflow provided in an embodiment of the present invention;
[0037] Figure 3 A schematic diagram of visual encoding of a subspace map provided in an embodiment of the present invention;
[0038] Figure 4 A schematic diagram of a process for constructing a subspace map provided in an embodiment of the present invention;
[0039] Figure 5 A schematic diagram of the structure of a high-dimensional data subspace visualization exploration and analysis system provided in an embodiment of the present invention;
[0040] Figure 6 Schematic diagram of a high-dimensional data subspace visualization exploration and analysis system based on a subspace map provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0041] In order to make the technical problems solved by the present invention, the technical solutions adopted and the technical effects achieved more clearly, the technical solutions of the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.
[0042] The embodiment of the present invention provides a high-dimensional data subspace visualization exploration and analysis method, which is used to provide an overview of the exploration space, help users understand the relationship between subspaces, and build their mental maps; reveal the interaction between dimensions and data patterns, and provide guidance for dimensional decision-making; guide users to browse the exploration space, so that they can understand representative subspaces and their differences in a short time. In order to achieve the above objectives, the present invention mainly includes the following steps: 1. Process high-dimensional data, search for valuable subspaces from the data and extract their features; 2. Construct a subspace map; 3. Draw subspace maps and other views. Figure 1 As shown, including:
[0043] 1. Processing high-dimensional data
[0044] S100, search for valuable subspaces from high-dimensional data sets and extract their features, the features including subspace similarity, subspace clustering, data patterns, dimensional patterns and representative subspaces.
[0045] In S100, searching for valuable subspaces from a high-dimensional data set includes: searching for potentially valuable subspaces based on a subspace clustering algorithm SURFING (SUbspace Relevant For Clustering), where SURFING assumes that a subspace with non-uniform distance distribution of data items is more likely to have a meaningful data structure (such as clustering).
[0046] S100 includes:
[0047] (1) Extracting subspace similarity: Calculate the k-NN list of each data item in each subspace obtained by the search, take the similarity between the k-NN lists of two subspaces corresponding to a certain data item as the similarity of the two subspaces on the data item, and average the similarities of the two subspaces on all data items to obtain the similarity of the two subspaces.
[0048] (2) Extract subspace clusters: Use the DBSCAN clustering algorithm to perform hierarchical clustering on each subspace obtained by the search. In order to obtain information of different granularities, the subspaces are hierarchically clustered using DBSCAN. To simplify the map metaphor, only two layers of clustering are currently considered. The clusters in the first layer serve as the input of the second layer.
[0049] (3) Extracting data patterns: Evaluate data items that have similar k-NN lists in all subspaces of the same cluster as the data pattern of the cluster. As mentioned earlier, data items with high k-NN consistency are considered stable between two subspaces. Further evaluate which data items have similar k-NN lists in all subspaces of the same cluster. This result is called the data pattern of the cluster, which reveals the common data structure of all member subspaces.
[0050] (4) Extracting dimensional patterns: Calculate the number of times each dimension appears in all subspaces of the same cluster, and take the dimensions that appear in a certain cluster at a number that reaches the corresponding high preset number and the dimensions that appear less than the corresponding low preset number as the dimensional pattern of the cluster. The dimensional pattern is determined by the combination of dimensions shared by all member subspaces. Calculate the number of times each dimension appears in all subspaces of the same cluster, and dimensions that are often included or not included are called feature dimensions. The number of times a dimension appears in the neighboring subspaces of a certain subspace is taken as the dimensional pattern of the subspace.
[0051] (5) Extract representative subspace: Select the subspace with the highest average similarity to other member subspaces as the representative subspace. Each clustering data and dimensional pattern requires a representative subspace to display. Generally, the subspace with the highest average similarity to other member subspaces is selected as the representative subspace.
[0052] 2. Constructing subspace maps
[0053] S200, constructing a subspace map based on the extracted features, including determining anchor point locations and subspace traversal order, determining a layout by grid tiling, and adding additional map metaphors.
[0054] Subspace maps use map metaphors to present the characteristics of subspaces and guide users to explore. In subspace maps, the entire exploration space is visualized as a geographic map. It contains three elements: geographic environment, region, and transportation. In terms of geographic environment, the system constructs a map with only one regular surface by tiling hexagons. Landscape is defined as the data pattern of each subspace. It is usually the main interest of tourists (i.e. users). Dimensions are similar to natural factors that determine landscapes. Land with similar landscapes are connected together to form continents. Land that is not connected to the continent becomes an island. Oceans are used to divide continents and have no special meaning. From a regional perspective, subspaces are viewed as cities. Each continent formed by similar subspaces is conceptualized as a country, which is divided into provinces based on a finer granularity of similarity. Some cities within a country may not be so similar to other cities, and they are called municipalities. For each country and province, a representative city (national capital or provincial capital) is selected to show its main characteristics. From a transportation perspective, sightseeing routes are planned for tourists. The system sets flight routes to facilitate long-distance travel between different countries. It also sets up land and sea routes connecting neighboring cities on the same continent and across the sea, respectively, to provide routes that are more stable to pattern changes. The subspace map is constructed by finding a suitable grid to display each subspace. The construction is done in three steps. First, an anchor point is placed on the map for each country and outlier. It determines the relative positions of the first-level clusters and outliers. Then, the cluster members are placed according to a pre-set similarity-based traversal order. After the map layout is completed, other map metaphors such as capitals, natural factors, and routes are placed on the map to make it more informative.
[0055] In S200, determining the anchor point position and the subspace traversal order includes:
[0056] (1) Determine the anchor point location: The reasonable locations of different countries and islands are obtained by drawing the dimension reduction projection of all subspaces. For each country, its capital is regarded as the anchor point. For each island, it is regarded as the anchor point and its location on the hexagonal grid is determined.
[0057] (2) Determine the order of traversal of the subspace: In the first level, determine the order of capitals and islands of all countries, and in the second level, determine the order of all cities in the same country. For accuracy, an empty list is created, and the city with the smallest average dissimilarity to other cities in the list is added to the list. This allows the map to put similar cities together. This process continues until all cities in the same country are included. Different countries are traversed separately.
[0058] In S200, the layout is determined by grid tiling, including:
[0059] (1) After placing the anchor points according to the determined anchor point positions, the cities of the same country are placed according to the determined subspace traversal order. A location queue needs to be established, which is used to maintain the currently available grids. Each subspace of the incoming hexagonal grid is accommodated by the first grid in the location queue, and the queue is continuously updated. After a grid is used, adjacent grids that meet the predetermined conditions are added to the queue in clockwise order. The predetermined conditions include not being occupied by other subspaces and not being disabled grids. Disabled grids are used to avoid borders between different countries and islands.
[0060] (2) If two adjacent subspaces belong to the same cluster, the current position queue is updated; otherwise, the current position queue is cleared and a new position queue is established for the new cluster.
[0061] (3) After determining the layout of all countries and islands, make the layout more compact by narrowing the gaps between them, enlarge the remaining grids in the center to fill the entire view space, and reflect the subspace similarity through color. Since the spatial relationship between countries and islands may be affected by compression, color can be used to reflect the subspace similarity to compensate for the information loss.
[0062] In S200, additional map metaphors are added, including:
[0063] The representative subspace of each cluster is displayed as the capital or provincial capital. For each city / region, its natural factor is the dimensional pattern and its landscape is the data pattern. After the user selects the starting and ending cities, the corresponding land and sea routes are obtained by calculating the routes with the least dissimilarity through the cities. The route network is made more concise through the minimum spanning tree algorithm.
[0064] 3. Draw subspace maps and other views
[0065] S300: Draw a subspace map and other auxiliary views, and complete subspace exploration through interaction with the user.
[0066] S300 includes:
[0067] Based on the visualization toolkit D3, subspace maps and other auxiliary views are drawn, and subspace list view and map detail view are provided to assist users in completing exploration and analysis. The subspace list view is used to display the dimensional information of the subspace, and the map detail view is used to display the dimensional mode and data mode of the subspace; at the country or province level, the map detail view is used to display the data mode and dimensional mode of the clustering, and at the city level, the map detail view is used to display the data mode and dimensional mode of the selected subspace.
[0068] Specifically, subspace maps and other views can be drawn in the display. The drawing part is completed based on the visualization toolkit D3. In addition to the subspace map, subspace list view and map detail view are also provided to assist users in completing exploration and analysis. The subspace list view displays the dimensional information of the subspace. The map detail view displays the dimension and data pattern. At the country or province level, this view displays clustered data and dimensional patterns. At the city level, it only displays the information of the selected subspace. Users can interact with these views, such as changing the map level or switching to travel mode, to complete subspace exploration.
[0069] Figure 2 The workflow of user exploration is shown. After constructing the subspace map based on the input high-dimensional data, the map first shows an overview with multiple subspace clusters. Users can drill down to different levels to explore at the cluster level, subcluster level, and subspace level. By analyzing various patterns and pattern transitions, you can gain insight into dominant dimensions and stable data patterns, etc.
[0070] Figure 3 The visual encoding of the subspace map is shown. The encoding is divided into three aspects: geographical environment, region and transportation.
[0071] Figure 4 The construction process of the subspace map is shown. First, anchor points are placed on the hexagonal grid, then other subspaces are placed according to the traversal order and available grids, and finally other map metaphors are introduced.
[0072] The method of the embodiment of the present invention is used to support users to efficiently complete the exploration and analysis of high-dimensional data subspaces in the form of a visualization of map metaphors. The layout strategy adopted has high space utilization and will not be blocked. It can display hundreds or thousands of subspaces and has high scalability. Since the form of the map is familiar to users, the visualization form of the present invention can well help users understand the relationship between subspaces. The rich metaphors in the map effectively reveal the complex interactions between dimensions and data patterns. These metaphors can also guide users to track the changes in patterns between different subspaces, allowing them to understand representative subspaces and their differences in a short time.
[0073] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.
[0074] According to another aspect of the embodiments of the present invention, a high-dimensional data subspace visualization exploration and analysis system is also provided. Figure 5 As shown, including:
[0075] An extraction module 100, for searching for valuable subspaces from a high-dimensional data set and extracting their features, wherein the features include subspace similarity, subspace clustering, data patterns, dimensional patterns, and representative subspaces;
[0076] A construction module 200 is used to construct a subspace map based on the extracted features, including determining the anchor point positions and the subspace traversal order, determining the layout by grid tiling, and adding additional map metaphors;
[0077] The drawing module 300 is used to draw the subspace map and other auxiliary views, and complete the subspace exploration through interaction with the user.
[0078] Figure 6 The high-dimensional data subspace visualization exploration and analysis system based on the subspace map is shown. In the embodiment of the present invention, the user can use the traditional keyboard and mouse to input, the calculation process is processed by the Intel Xeon dual-core processor with 8GB memory, and the display process is realized by a 23-inch 1920×1080 pixel display driven by the NVIDIA Quadro2000 graphics card.
[0079] It should be noted that the above-mentioned high-dimensional data subspace visualization exploration and analysis system and the above-mentioned high-dimensional data subspace visualization exploration and analysis method belong to the same inventive concept, and the specific implementation methods will not be repeated here.
[0080] The system using the embodiment of the present invention supports users to efficiently complete the exploration and analysis of high-dimensional data subspaces in the form of a visualization of map metaphors. The layout strategy adopted has high space utilization and will not be blocked. It can display hundreds or thousands of subspaces and has high scalability. Since the form of the map is familiar to users, the visualization form of the present invention can well help users understand the relationship between subspaces. The rich metaphors in the map effectively reveal the complex interactions between dimensions and data patterns. These metaphors can also guide users to track the changes in patterns between different subspaces, allowing them to understand representative subspaces and their differences in a short period of time.
[0081] The present invention also provides a storage medium, in which a computer program is stored, wherein when the computer program is run, a high-dimensional data subspace visualization exploration and analysis method of the present invention can be executed. The storage medium includes at least one of the following: a floppy disk, an optical disk, a DVD, a hard disk, a flash memory, a U disk, a CF card, an SD card, an MMC card, an SM card, a memory stick (Memory Stick), an xD card, etc. The high-dimensional data subspace visualization exploration and analysis method of the present invention is converted into data (computer program) and burned into the above storage medium. For example, a hard disk with the computer program of the method of the present invention is put into a computer for operation, and then the high-dimensional data subspace visualization exploration and analysis method of the present invention can be realized.
[0082] The present invention also provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute a high-dimensional data subspace visualization exploration and analysis method of the present invention. The memory belongs to the aforementioned storage medium, and can store the computer program of the high-dimensional data subspace visualization exploration and analysis method of the present invention, and the processor can process the data in the memory, and the electronic device can be a computer, a mobile phone, or any other device including a memory and a processor. After the computer is started, the processor is started to run the computer program of the high-dimensional data subspace visualization exploration and analysis method of the present invention in the memory, and then the high-dimensional data subspace visualization exploration and analysis method of the present invention can be implemented.
[0083] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A high-dimensional data subspace visualization exploration and analysis method, characterized in that: include: S100, searching for valuable subspaces from a high-dimensional data set and extracting their features, wherein the features include subspace similarity, subspace clustering, data patterns, dimensional patterns, and representative subspaces; S100 includes: (1) Extracting subspace similarity: Calculate the k-NN list of each data item in each subspace obtained by the search, take the similarity between the k-NN lists of two subspaces corresponding to a certain data item as the similarity of the two subspaces on the data item, and average the similarities of the two subspaces on all data items to obtain the similarity of the two subspaces; (2) Extract subspace clusters: Use the DBSCAN clustering algorithm to perform two-layer clustering on each subspace obtained from the search, and the clusters in the first layer serve as the algorithm input for the second layer; (3) Extracting data patterns: Evaluate data items with similar k-NN lists in all subspaces of the same cluster as the data patterns of the cluster; (4) Extracting dimensional patterns: Calculate the number of times each dimension appears in all subspaces of the same cluster. Dimensions whose number of appearances in a cluster reaches the corresponding high preset number and whose number of appearances is lower than the corresponding low preset number are taken as the dimensional pattern of the cluster. The number of times a dimension appears in the neighboring subspace of a subspace is taken as the dimensional pattern of the subspace. (5) Extract representative subspace: Select the subspace with the highest average similarity to other member subspaces as the representative subspace; S200, constructing a subspace map based on the extracted features, including determining anchor point locations and subspace traversal order, determining a layout by grid tiling, and adding additional map metaphors; S300: Draw the subspace map and other auxiliary views, and complete subspace exploration through interaction with the user.
2. The method according to claim 1, characterized in that S100 includes: Based on the subspace clustering algorithm, the subspace with non-uniform distance distribution pattern of data items is searched out as a valuable subspace.
3. The method according to claim 1, characterized in that In S200, determining the anchor point position and the subspace traversal order includes: (1) Determine the anchor point location: The reasonable locations of different countries and islands are obtained by drawing the dimension reduction projection of all subspaces. For each country, its capital is regarded as the anchor point. For each island, it is regarded as the anchor point and its location on the hexagonal grid is determined. (2) Determine the subspace traversal order: Determine the order of the capitals and islands of all countries at the first level, and determine the order of all cities in the same country at the second level. That is, create an empty list and add the city with the smallest average dissimilarity to other cities in the list to the list. This process continues until all cities in the same country are included. Different countries are traversed separately.
4. The method according to claim 3, characterized in that In S200, the layout is determined by grid tiling, including: (1) After placing the anchor points according to the determined anchor point positions, the cities in the same country are placed according to the determined subspace traversal order, that is, a position queue is established, and each subspace passed into the hexagonal grid is accommodated by the first grid in the position queue, and the queue is continuously updated. After a grid is used, adjacent grids that meet predetermined conditions are added to the queue in a clockwise order, and the predetermined conditions include not being occupied by other subspaces and not being a disabled grid; (2) If two adjacent subspaces belong to the same cluster, the current position queue is updated; otherwise, the current position queue is cleared and a new position queue is established for the new cluster; (3) After determining the layout of all countries and islands, the layout is made more compact by narrowing the gaps between them and the remaining grids are enlarged and centered to fill the entire view space, with colors reflecting subspace similarity.
5. The method according to claim 4, characterized in that In S200, additional map metaphors are added, including: The representative subspace of each cluster is displayed as the capital or provincial capital. For each city / region, its natural factor is the dimensional pattern and its landscape is the data pattern. After the user selects the starting and ending cities, the corresponding land and sea routes are obtained by calculating the routes with the least dissimilarity through the cities. The route network is made more concise through the minimum spanning tree algorithm.
6. The method according to any one of claims 1 to 5, characterized in that: S300 includes: The subspace map and other auxiliary views are drawn based on the visualization toolkit D3, and a subspace list view and a map detail view are provided to assist users in completing exploration and analysis. The subspace list view is used to display the dimensional information of the subspace, and the map detail view is used to display the dimensional pattern and data pattern of the subspace; at the country or province level, the map detail view is used to display the data pattern and dimensional pattern of the cluster, and at the city level, the map detail view is used to display the data pattern and dimensional pattern of the selected subspace.
7. A high-dimensional data subspace visualization exploration and analysis system, characterized in that: include: The extraction module is used to search for valuable subspaces from high-dimensional data sets and extract their features, including subspace similarity, subspace clustering, data patterns, dimensional patterns, and representative subspaces; specifically, it includes: (1) Extracting subspace similarity: Calculate the k-NN list of each data item in each subspace obtained by the search, take the similarity between the k-NN lists of two subspaces corresponding to a certain data item as the similarity of the two subspaces on the data item, and average the similarities of the two subspaces on all data items to obtain the similarity of the two subspaces; (2) Extract subspace clusters: Use the DBSCAN clustering algorithm to perform two-layer clustering on each subspace obtained from the search, and the clusters in the first layer serve as the algorithm input for the second layer; (3) Extracting data patterns: Evaluate data items with similar k-NN lists in all subspaces of the same cluster as the data patterns of the cluster; (4) Extracting dimensional patterns: Calculate the number of times each dimension appears in all subspaces of the same cluster. Dimensions whose number of appearances in a cluster reaches the corresponding high preset number and whose number of appearances is lower than the corresponding low preset number are taken as the dimensional pattern of the cluster. The number of times a dimension appears in the neighboring subspace of a subspace is taken as the dimensional pattern of the subspace. (5) Extract representative subspace: Select the subspace with the highest average similarity to other member subspaces as the representative subspace; The construction module is used to construct the subspace map based on the extracted features, including determining the anchor point location and subspace traversal order, determining the layout by grid tiling, and adding additional map metaphors; The drawing module is used to draw the subspace map and other auxiliary views, and complete the subspace exploration by interacting with the user.
8. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute a high-dimensional data subspace visualization exploration and analysis method as described in any one of claims 1 to 6 when running.
9. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is configured to run the computer program to execute the high-dimensional data subspace visualization exploration and analysis method described in any one of claims 1 to 6.
