High-dimensional space visualization method based on data board

Abstract

The invention relates to a high-dimensional space visualization method based on a data board, and the method comprises the steps: carrying out the dimension analysis and space dimension reduction of a high-dimensional experiment space, and obtaining a low-dimensional experiment space and a two-dimensional experiment space; data processing is conducted on the high-dimensional experiment space, the high-dimensional experiment space comprises text type data and list data, and the text type data and the list data are mapped into a single numerical value type and visually displayed in a data board; the data board comprises five panels: an evolution progress visualization panel, an experiment point evolution visualization panel, a factor evolution visualization panel, a score distribution visualization panel and a score trend visualization panel. According to the method, the problem that the high-dimensional experimental space is difficult to visualize due to high dimension and dynamic evolution is solved, the high-dimensional experimental space data deduced by simulation can be simply and intuitively presented to the user, and the user can more easily understand the sampling condition and the evolution direction of the experimental space by means of the assistance of an intelligent algorithm.

Description

Technical Field

[0001] The present application belongs to the field of visualization technology, and in particular relates to a high-dimensional space visualization method based on a data dashboard. Background Art

[0002] High-dimensional experimental spaces are characterized by high data dimensions, diverse data types, and difficulty visualizing their evolutionary processes. Current large-scale visualization screens have limited page content capacity, weak data presentation capabilities, insufficient scenario-based integration, and limited interactive methods, making them unable to meet the visualization needs of high-dimensional data. Therefore, there is an urgent need to develop a high-dimensional space visualization method based on data dashboards to address the difficulty of presenting high-dimensional data and multiple business modules on the same page, while ensuring simple operation and intuitive presentation. Summary of the Invention

[0003] The main purpose of the present invention is to provide a high-dimensional space visualization method based on a data dashboard, aiming to solve the problem that high-dimensional data and multiple business modules are difficult to present on the same page.

[0004] To achieve the above objectives, the present invention provides a high-dimensional space visualization method based on a data dashboard.

[0005] The present invention provides a high-dimensional space visualization method based on a data dashboard, the method comprising the following steps: performing dimensional analysis on the high-dimensional experimental space to determine spatial dimensions and visualization data; the visualization data includes the evolution progress of running rounds, the evolution status of experimental factors, the experimental indicator score, the comprehensive evaluation score, and the evolution status of experimental points; the experimental point includes at least one of the experimental factors;

[0006] Performing spatial dimensionality reduction on the high-dimensional experimental space to obtain a low-dimensional experimental space, wherein the low-dimensional experimental space includes at least one experimental point, a set of experimental indicator scores, and at least one comprehensive evaluation score; performing dimensionality reduction on the low-dimensional experimental space to obtain a two-dimensional experimental space, wherein the two-dimensional experimental space includes at least one experimental factor, at least one experimental indicator score, and at least one comprehensive evaluation score;

[0007] The high-dimensional experimental space includes text type data and list data, and data processing is performed on the high-dimensional experimental space, including mapping the text type data and the list data into a single numerical type, and visually displaying them in the data dashboard;

[0008] The data dashboard includes five panels, namely, an evolution progress visualization panel, an experimental point evolution visualization panel, a factor evolution visualization panel, a score distribution visualization panel and a score trend visualization panel; wherein, the evolution progress visualization panel displays the progress status of the high-dimensional experimental space of the current running round of deduction based on a pie chart; the experimental point evolution visualization panel is based on a situation diagram, and the clustering movement of the experimental points is used to represent the current exploration focus and the degree of convergence of the experimental points; the factor evolution visualization panel displays the evolution status of the experimental factors based on a cloud-rain diagram; the score distribution visualization panel displays the comprehensive evaluation score distribution status of the current experimental point after the current round of deduction based on a box plot, to assist in judging the evolution trend of the current experimental point; the score trend visualization panel displays the current experimental indicator score change trend and possible future changes based on a curve chart, to assist in judging the future change trend of the current experimental point.

[0009] The data panel adopts a layout of one large and four small panels. The background of the panel is a white dotted rectangle, and the four corners of the rectangle are blue "L" shapes. The upper left corner of the panel is the panel title, and the upper right corner or directly above is the setting function, which is represented by a gear box. The bottom or side of the panel supports sliding bars for dragging or zooming.

[0010] Optionally, performing spatial dimensionality reduction on the high-dimensional experimental space includes:

[0011] The dimension of the high-dimensional experimental space Ω(N,L) is N, and the number of rounds is L. The high-dimensional experimental space also evolves with the rounds. The dimension calculation formula of the high-dimensional experimental space is as follows:

[0012] N=N1+N2+N3

[0013] N1=N 11 +N 12 +…+N 1i

[0014] N2=N 21 +N 22 +…+N 2j

[0015] N3=1

[0016] Among them, N1 is the experimental factor dimension, i is the total number of experimental factors, N2 is the experimental indicator score dimension, j is the total number of experimental indicators, and N3 is the comprehensive evaluation score dimension.

[0017] Optionally, the high-dimensional experimental space Ω(N,L) is subjected to spatial dimensionality reduction to obtain low-dimensional experimental spaces (Ω(N1,L), Ω(N2,L), Ω(N3,L)), wherein Ω(N1,L) is the experimental point, Ω(N2,L) is the experimental index score set, and Ω(N3,L) is the comprehensive evaluation score; the low-dimensional experimental space (Ω(N1,L), Ω(N2,L) is subjected to dimensionality reduction to obtain two-dimensional experimental spaces (Ω(N 11 ,L),Ω(N 12 ,L),...,Ω(N 1i ,L)) and (Ω(N 21 ,L),Ω(N 22 ,L),...,Ω(N 2j ,L)), where (Ω(N 11 ,L),Ω(N 12 ,L),...,Ω(N 1i ,L)) are the experimental factor evolution states of experimental factors 1, 2, ..., i, (Ω(N 21 ,L),Ω(N 22 ,L),...,Ω(N 2j ,L)) are the experimental index scores of experimental indicators 1, 2,..., j respectively.

[0018] Optionally, performing data processing on the high-dimensional experimental space includes:

[0019] The mapping relationship between text type data and list data is as follows:

[0020] {b|b=f(a1,a2),a1,a2∈Ω(N,L)}

[0021] Among them, a1 is the text type data point in the original experimental space, a2 is the list type data point in the original experimental space, f is the mapping relationship, and b is the numerical type data point after conversion in the experimental space.

[0022] Optionally, in the evolution progress visualization panel, the pie chart is located in the center of the evolution progress visualization panel, the entire pie chart is transparent, and the progress bar changes color gradually; the currently selected evolution iteration target is indicated directly above the pie chart, and the iteration target includes running rounds and comprehensive evaluation scores; two lines of text are displayed in the center of the pie chart, the first line is the current deduction round, expressed in a numerical value, and the second line is the current evolution progress, expressed in percentage; the remaining estimated completion time of the current round and the remaining estimated total completion time are displayed directly below the pie chart; the prediction of the remaining estimated completion time of the current round and the remaining estimated total completion time adopts a big data statistical prediction algorithm, and a statistical prediction model is determined based on the current evolution progress and completion time, and the prediction model is used to update the remaining completion time in real time.

[0023] Select the evolution iteration target through the setting function. When the evolution iteration target is the running round, the ring progress calculation formula is as follows: Among them, p i is the current deduction round, p0 is the upper limit of the round, y i The deduction progress of each sample in the current deduction round;

[0024] When the iteration target is a comprehensive evaluation score, the formula for calculating the ring progress is as follows: Among them, maxp i is the highest comprehensive evaluation score of the experimental point in the current deduction round, p0 is the expected score, y i It indicates the deduction progress of each sample in the current deduction round.

[0025] Optionally, in the factor evolution visualization panel, click the setting function to pop up a pop-up box for switching experimental factors. The pop-up box has a gray-black rounded rectangular background, and the experimental factor selection box is a radio button with a blue hollow circle. When one of the experimental factors is selected, the blue hollow circle jumps to a blue solid circle, and the experimental factor distribution status of the experimental factor in multiple rounds is displayed in the panel. The experimental factor distribution status in each round is represented by a cloud-rain graph, and the cloud-rain graph uses red, green, purple, brown, and orange as a group for cyclic display; the panel can be slid vertically; the clouds in the cloud-rain graph represent the value distribution of the current experimental factor, and the rain represents the value size and number of the current experimental point; the horizontal axis of the cloud-rain graph represents the score, and the vertical axis represents the evolution round, wherein the vertical lines are distributed at 20 equal intervals.

[0026] Optionally, in the score distribution visualization panel, the horizontal axis of the box plot represents the round, and the vertical axis represents the score value, wherein the horizontal distribution is divided into 10 equal intervals, and the panel can be slid horizontally to display the comprehensive evaluation scores of all rounds; the comprehensive evaluation score of each round is represented in the form of a box plot combined with a dot-shaped scatter plot, and is displayed in a cycle through red, green, purple, brown and orange; the scatter plot has the same color as the box plot, and the transparency of the scatter plot is lower than that of the box plot; the smaller the score distribution range and the higher the upper limit in the box plot, the better the evolution of the current experimental point, and vice versa.

[0027] Optionally, in the scoring trend visualization panel, click the setting function to pop up an indicator selection pop-up box, which includes multiple indicators. The pop-up box has a dark gray rounded rectangular background, and the indicator selection box is a check box. The check box is a rectangular box with a dark blue border. When one of the indicators is selected, the rectangular box is filled with a dark blue check mark; the panel uses multiple curves to represent the scores of multiple indicators in multiple rounds, and the curves corresponding to the indicators are the same color as the color blocks before the indicator options in the pop-up box; the horizontal axis of the curve represents the round, and the vertical axis represents the score, which is horizontally distributed at 10 equal intervals. The panel can be slid horizontally to show the indicator change trends of the multiple indicators in all rounds; the higher the score, the better the evolution direction of the current experimental point, and vice versa; when the mouse hovers over a point in the curve, a dot and a dialogue bubble appear on the curve, and the bubble displays a brief overview of the data at that point, with white text.

[0028] The scoring trend visualization panel is also provided with a prediction curve button. When the prediction curve button is in the display state, the subsequent prediction curve is displayed to indicate the change state of the experimental indicator score in future rounds. The prediction curve is presented by a dotted line and is predicted using a least squares regression prediction algorithm.

[0029] Optionally, in the experimental point evolution visualization panel, the system sequentially assigns an initial position to each experimental point as a label of the experimental point, clusters the experimental points using a clustering method, and uses the obtained cluster cluster as the current exploration focus direction; wherein the clustering method includes but is not limited to Euclidean distance.

[0030] Optionally, the experimental point evolution visualization panel is a large panel, in which the experimental points are displayed by dots, the color of the dots gradually becomes lighter from the center to the edge, and the dots are repeatedly zoomed outward from the center point to present a flashing effect; the dots include gray dots and red dots, the gray dots represent the experimental points of the previous two rounds, and the red dots represent the experimental points of the previous round; the panel supports vertical and horizontal sliding;

[0031] Click on a situation point in the situation diagram to display the values ​​of each experimental factor and the exploration proportion in the current experimental point in a list. If the evolution direction of the experimental factor research visualization, the score distribution visualization, and the score trend visualization does not meet expectations, the exploration proportion will be modified and sent to the space exploration module. In the next round of deduction, the space exploration algorithm will adjust the exploration direction according to the modification.

[0032] An embodiment of the present invention provides a high-dimensional space visualization method based on a data dashboard, which performs dimensional analysis on the high-dimensional experimental space to determine spatial dimensions and visualization data, wherein the visualization data includes the evolution progress of running rounds, the evolution status of experimental factors, experimental indicator scores, comprehensive evaluation scores, and the evolution status of experimental points; the experimental points include at least one experimental factor; spatial dimensionality reduction is performed on the high-dimensional experimental space to obtain a low-dimensional experimental space, wherein the low-dimensional experimental space includes at least one experimental point, a set of experimental indicator scores, and at least one comprehensive evaluation score; the low-dimensional experimental space is dimensionality reduced to obtain a two-dimensional experimental space, wherein the two-dimensional experimental space includes at least one experimental factor, at least one experimental indicator score, and at least one comprehensive evaluation score; the high-dimensional experimental space includes text type data and list data, and data processing of the high-dimensional experimental space includes: mapping the text type data and the list data into a single numerical type, and visually displaying them in the data dashboard; the data panel includes five panels; the five panels include an evolution progress visualization panel, an experimental point evolution visualization panel, a factor evolution visualization panel, a score distribution visualization panel, and a score trend visualization panel. The Evolution Progress visualization panel uses a donut chart to display the progress of the currently simulated experimental space; the Factor Evolution visualization panel uses a cloud-rain chart to display the evolution of experimental factors; the Score Distribution visualization panel uses a boxplot to display the distribution of the comprehensive evaluation scores of the current experimental point after the current round of simulation, helping to determine the evolutionary trend of the current experimental point; the Score Trend visualization panel uses a curve chart to display the current experimental indicator score change trend and possible future changes, helping to determine the future trend of the current experimental point; and the Experimental Point Evolution visualization panel uses a situation map to display the current exploration focus. The data dashboard uses a large and four small layout. The background of each panel is a white dotted rectangle with blue "L" corners. The panel title is located in the upper left corner, and the settings function is represented by a gear box in the upper right corner or directly above. The bottom and sides of the panel support sliders for dragging and zooming. The data dashboard visualizes high-dimensional data after dimensionality reduction, solving the problem of presenting high-dimensional data and multiple business modules on the same page.

[0033] Compared with the prior art, this application has the following beneficial effects:

[0034] This paper proposes a high-dimensional space visualization method based on data dashboards. This method has four characteristics:

[0035] (1) Simple and intuitive, it can present the high-dimensional experimental space data of simulation deduction to users in a simple and intuitive way;

[0036] (2) Easy to understand: users can more easily understand the sampling situation and evolution direction of the experimental space through the data dashboard;

[0037] (3) Intelligent assistance, which can use intelligent algorithms to predict the completion time and indicator scores of spatial evolution, reduce unnecessary waiting time, and observe and evaluate future evolution directions;

[0038] (4) Communication and sharing support friendly human-computer interaction, making it convenient for users to adjust the exploration direction of the experimental space. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the spatial dimensionality reduction principle of high-dimensional data in one embodiment

[0040] Figure 2 Schematic diagram of the overall design of a data dashboard in one embodiment

[0041] Figure 3 Schematic diagram of the design of an evolution progress visualization panel in one embodiment

[0042] Figure 4 Schematic diagram of the design of an evolution progress visualization panel in one embodiment

[0043] Figure 5 Schematic diagram of experimental point evolution visualization panel design in one embodiment

[0044] Figure 6 Schematic diagram of factor evolution visualization panel design in one embodiment

[0045] Figure 7 Schematic diagram of a score distribution visualization panel design in one embodiment

[0046] Figure 8 Schematic diagram of a scoring trend visualization panel design in one embodiment

[0047] Figure 9 Schematic diagram of Step 1 of the high-dimensional experimental space visualization process in one embodiment

[0048] Figure 10 Schematic diagram of Step 2 of the high-dimensional experimental space visualization process in one embodiment

[0049] Figure 11 Schematic diagram of Step 3 of the high-dimensional experimental space visualization process in one embodiment

[0050] Figure 12 Schematic diagram of Step 4 of the high-dimensional experimental space visualization process in one embodiment

[0051] Figure 13Schematic diagram of Step 5-a of the high-dimensional experimental space visualization process in one embodiment

[0052] Figure 14 Schematic diagram of Step 5-b of the high-dimensional experimental space visualization process in one embodiment

[0053] Figure 15 Schematic diagram of Step 6 of the high-dimensional experimental space visualization process in one embodiment DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions, and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and examples. It is apparent that the described embodiments are only a portion of the embodiments of the present invention, not all of them. It should be understood that all other embodiments obtained by persons of ordinary skill in the art without creative effort are within the scope of protection of this invention.

[0055] A Data Dashboard is an information display tool that uses a graphical interface to display key performance indicators and other important data, helping users quickly understand and analyze data so that decision makers and managers can monitor business status, project progress, or system performance in real time.

[0056] As mentioned above, high-dimensional experimental space has the characteristics of high data dimension, multiple data types, and difficult to intuitively present the evolution process.

[0057] Therefore, in order to solve the problem that high-dimensional data is difficult to present on the same page with multiple business modules, the present invention is implemented through the following technical solutions:

[0058] Perform spatial dimension analysis on the high-dimensional experimental space to determine the spatial dimension and visualization data; specifically, the visualization data includes the evolution progress of the running rounds, the evolution status of the experimental factors, the experimental indicator scores, the comprehensive evaluation scores, and the evolution status of the experimental points; the experimental points include at least one experimental factor. Perform spatial dimensionality reduction on the high-dimensional experimental space to obtain a low-dimensional experimental space, which includes at least one experimental point, a set of experimental indicator scores, and at least one comprehensive evaluation score; perform dimensionality reduction on the low-dimensional experimental space to obtain a two-dimensional experimental space, which includes at least one experimental factor, at least one experimental indicator score, and at least one comprehensive evaluation score; the high-dimensional experimental space includes text-type data and list data, and perform data processing on the high-dimensional experimental space, including mapping the text-type data and list data to a single numerical type, and visually displaying them in the data dashboard;

[0059] The data dashboard includes five panels: evolution progress visualization panel, experimental point evolution visualization panel, factor evolution visualization panel, score distribution visualization panel and score trend visualization panel; among them, the evolution progress visualization panel displays the progress status of the high-dimensional experimental space of the current running round of deduction based on a donut chart; the experimental point evolution visualization panel is based on a situation diagram, and the clustering movement of the experimental points is used to represent the current exploration focus and the degree of convergence of the experimental points; the factor evolution visualization panel displays the evolution status of the experimental factors based on a cloud-rain diagram; the score distribution visualization panel displays the distribution status of the comprehensive evaluation score of the current experimental point after the current round of deduction based on a box plot, to assist in judging the evolution trend of the current experimental point; the score trend visualization panel displays the current experimental indicator score change trend and possible future changes based on a curve chart, to assist in judging the future change trend of the current experimental point.

[0060] In one embodiment of the present invention, the visualization of the high-dimensional experimental space is a visual display of the evolution progress of the high-dimensional experimental space in terms of the distance convergence target, experimental points, experimental factors, experimental indicator scores and comprehensive evaluation scores.

[0061] In one embodiment of the present invention, the visualization dashboard includes a control setting panel, a sample list panel, and an experimental space visualization panel; the control setting panel is located on the left side of the data dashboard, the sample list panel and the experimental space visualization panel are located on the right side of the data dashboard, and the experimental space visualization panel is the data dashboard, which includes five panels;

[0062] The data dashboard is composed of five panels, with a layout of one large and four small panels. The evolution progress visualization panel and the experimental point evolution visualization panel are located in the upper part of the data dashboard, and the experimental point evolution visualization panel occupies a larger area of ​​the data dashboard. The factor evolution visualization panel, the score distribution visualization panel and the score trend visualization panel are located in the lower part of the data dashboard. Each panel uses a white dotted rectangle as the background, and the four corners of the rectangle are highlighted with a blue "L" shape. The upper left corner is the panel title, the upper right corner or the top is the setting function, and the content beyond the panel area supports dragging or zooming by dragging the bottom or side slider. The overall design of the dashboard is as follows Figure 1 shown.

[0063] In one embodiment of the present invention, the high-dimensional experimental space is firstly subjected to dimensional analysis, specifically including: assuming that the dimension of the experimental space is N and the number of rounds is L, the experimental space is not only high-dimensional but also evolves with the number of rounds, then the high-dimensional experimental space dimension calculation formula is

[0064] N=N1+N2+N3

[0065] N1=N 11 +N 12 +…+N 1i

[0066] N2=N 21 +N 22 +…+N 2j

[0067] N3=1

[0068] Among them, N1 is the experimental factor dimension, i is the total number of experimental factors, N2 is the experimental indicator score dimension, j is the total number of experimental indicators, and N3 is the comprehensive evaluation score dimension.

[0069] like Figure 2 As described, the high-dimensional experimental space is subjected to spatial dimensionality reduction. In one embodiment of the present invention, a spatial dimensionality reduction method is used to decompose the original direct visualization problem of the high-dimensional space and the round into a visualization combination presentation problem of each visualization data of the experimental space and the round, and finally a two-dimensional experimental space is obtained, and the two-dimensional experimental space is data mapped and visualized in the data dashboard. The specific process is as follows: first, the high-dimensional experimental space Ω(N,L) is subjected to spatial dimensionality reduction to obtain a low-dimensional experimental space (Ω(N1,L), Ω(N2,L), Ω(N3,L)), wherein Ω(N1,L) is the experimental point, Ω(N2,L) is the experimental indicator score set, and Ω(N3,L) is the comprehensive evaluation score; at this time, the comprehensive evaluation score has been reduced to a two-dimensional experimental space, and the experimental point and indicator dimensions are still higher than the two-dimensional experimental space, and further dimensionality reduction processing to the two-dimensional experimental space is required for visualization display; then, the low-dimensional experimental space (Ω(N1,L), Ω(N2,L) is subjected to dimensionality reduction to obtain two-dimensional experimental spaces (Ω(N1,L), Ω(N2,L) respectively. 11 ,L),Ω(N 12 ,L),...,Ω(N 1i ,L)) and (Ω(N 21 ,L),Ω(N 22 ,L),...,Ω(N 2j ,L)), where (Ω(N 11 ,L),Ω(N 12 ,L),...,Ω(N 1i ,L)) are the experimental factor evolution states of experimental factors 1, 2, ..., i, (Ω(N 21 ,L),Ω(N 22 ,L),...,Ω(N 2j ,L)) are the experimental index scores of experimental indicators 1, 2,..., j respectively.

[0070] In one embodiment of the present invention, the evolutionary progress of the running round refers to the progress status of the high-dimensional experimental space deduced by the current running round. Therefore, the evolutionary progress of the running round does not require dimensionality reduction processing.

[0071] In this embodiment, the data types of the high-dimensional experimental space mainly include numerical values ​​and texts, and the numerical types are further divided into single numerical values ​​and list data. Using a data dashboard to display the coordinated association of multiple views to realize the visualization of the experimental space faces the problem of visual display of text type data and list data. The present invention establishes a mapping relationship between text type data and numerical type data, and maps the text type data and list data in the high-dimensional experimental space to a single numerical type, so that all data can be realized based on the data dashboard. The mapping relationship is

[0072] {b|b=f(a1,a2),a1,a2∈Ω(N,L)}

[0073] Among them, a1 is the text type data point in the original experimental space, a2 is the list type data point in the original experimental space, f is the mapping relationship, and b is the numerical type data point after conversion in the experimental space.

[0074] In one embodiment of the present invention, the evolution progress visualization panel is displayed using a donut chart, which is mainly used to display the progress of the currently deduced experimental space. The donut is located in the center of the panel. The donut is transparent as a whole, and the progress bar changes color gradually, with the color value transitioning from #36BBFE to #A338FE. The outer circle of the donut is slightly shaded to the lower right, and the inner circle is slightly shaded from the upper left corner to the lower right, reflecting the overall three-dimensional sense of the donut. Two lines of text are arranged in the center of the donut. The first line is the current deduction round, expressed in numerical values, using black fonts, and the second line is the current evolution progress, expressed in percentages, using gray fonts. The text directly below the donut is displayed in a centered black font, and the effect is shown in the figure below. Figure 3 Click the drop-down menu behind the word "Evolution Progress" to switch the display of the current progress of "Run Round" or "Evaluation Score". It can be understood that the evaluation score here is actually the comprehensive evaluation score. The setting function is in the upper right corner of the evolution progress visualization panel. Through the setting, the evolution iteration target can be selected. When the iteration target is selected as round, the formula for calculating the circular progress is:

[0075]

[0076] Among them, p i is the current deduction round, p0 is the upper limit of the round, y i It indicates the deduction progress of each sample in the current deduction round.

[0077] When the iteration target is selected as the expected score, the formula for calculating the ring progress is:

[0078]

[0079] Among them, maxp i is the highest comprehensive evaluation score of the experimental points in the current deduction round, p0 is the expected score, yi It is the deduction progress of each sample in the current deduction round.

[0080] The bottom of the Evolution Progress Visualization Panel shows the remaining estimated completion time and the remaining estimated total completion time for the current round. The remaining time prediction uses a big data statistical prediction algorithm. The basic principle is to summarize the statistical prediction model from the big data of the current progress and completion time, and then use the prediction model to update the remaining completion time in real time. The design diagram of the Evolution Progress Visualization Panel is as follows: Figure 4 shown.

[0081] In one embodiment of the present invention, the evolution of experimental points is visualized using a situation diagram. The system assigns an initial position to each experimental point in sequence. The position serves as the label of the experimental point. Then, the clustering movement of the experimental points represents the current exploration focus. The clustering criterion is the Euclidean distance between the experimental factors in each experimental point. The effect is achieved by Figure 5 shown.

[0082] As mentioned earlier, the experimental point evolution visualization panel occupies a large area of ​​the data dashboard. In practice, it occupies the largest area of ​​the entire data dashboard. Experimental points are displayed as dots (the color of the dots gradually lightens from the center to the edges), and are repeatedly zoomed outward from the center point, creating a flashing effect. Gray dots represent experimental points from the previous two rounds, and red dots represent experimental points from the previous round. The entire panel supports zooming and mouse dragging, allowing users to quickly browse or select experimental points for adjustment.

[0083] Furthermore, clicking on a situation point can display the values ​​of each experimental factor and the exploration ratio in the current experimental point in a list format. When the user observes the visualization of the evolution of experimental factors, the visualization of the score distribution, and the visualization of the score trend and finds that the evolution direction does not meet expectations, the exploration ratio can be modified and sent to the spatial exploration module. In the next round of deduction, the spatial exploration algorithm will adjust the exploration direction according to the user's opinions.

[0084] In one embodiment of the present invention, the evolution of experimental factors is visualized using a cloud-rain diagram. A single cloud-rain diagram represents the distribution of experimental factors within a single round, and multiple cloud-rain diagrams represent the evolution of experimental factors over multiple rounds. The clouds in the cloud-rain diagram represent the overall value distribution of the current experimental factors, and the rain represents the size and number of the current experimental factor values. The horizontal axis of the cloud-rain diagram represents the experimental factor values, and the vertical axis represents the evolution rounds. The effect is shown in the figure below. Figure 6 The upper right corner of the Factor Evolution Visualization Panel contains the settings function, which allows you to choose to display different experimental factors.

[0085] The cloud-and-rain chart in the factor evolution visualization panel displays scores on the horizontal axis and rounds on the vertical axis. Vertical lines are spaced evenly every 20 points along the vertical axis. The panel can be scrolled vertically to display the evolution data for all rounds. The cloud-and-rain chart illustrates the evolution of each experimental factor in each round, with red, green, purple, brown, and orange colors displayed in a cycling pattern. Clicking the gear will bring up a pop-up window for selecting experimental factors. The pop-up window uses a gray-black rounded rectangular background and a standard radio button (a hollow blue circle). Selecting an option displays a solid blue circle within the circle.

[0086] In one embodiment of the present invention, the score distribution visualization is displayed using a box plot, which is mainly used to present the distribution of the comprehensive evaluation score of the current experimental point after a round of deduction, and to assist in judging whether the evolution trend of the current experimental point is evolving in a good direction. The smaller the score distribution range and the higher the upper limit, the better the evolution of the current experimental point, and vice versa. The horizontal axis of the box plot is the round, and the vertical axis is the score value. The effect is achieved as shown in the figure Figure 7 shown.

[0087] The score distribution visualization panel displays rounds on the horizontal axis and scores on the vertical axis, with horizontal lines spaced evenly across every 10 points. The panel can be scrolled horizontally to display the score data for all rounds. Each round's score is displayed as a boxplot combined with a scatterplot, with red, green, purple, brown, and orange colors displayed in a cyclical manner. Scatterplots are drawn using dots of the same color as the corresponding boxplots, but with a lower transparency to prevent them from obscuring the boxplots.

[0088] In one embodiment of the present invention, the score trend visualization is implemented using a visualization method based on a curve chart, which is mainly used to present the current experimental indicator score change trend and possible future changes, and to assist in judging whether the future change trend of the current experimental point is evolving in a good direction. The higher the experimental indicator score, the better the evolution direction of the current experimental point, and vice versa. The horizontal axis of the curve chart is the round, and the vertical axis is the experimental indicator score. The effect is achieved as shown in the figure below. Figure 8 The upper right corner of the score trend visualization panel is the settings function, which allows you to choose to display different experimental indicators.

[0089] Just below the score trend visualization panel setting function is a checkbox for whether to select the prediction curve. Checking the checkbox can use a curve to display the change status of the indicator scores in future rounds. The prediction algorithm uses the least squares regression prediction algorithm.

[0090] Scoring trends are visualized, with rounds on the horizontal axis and scores on the vertical axis. Horizontal lines are evenly spaced every 10 points, and the panel can be scrolled horizontally to display the indicator's changing trends across all rounds. Each indicator is represented as a broken line, with each line displayed in red, green, purple, brown, orange, and other colors. When the prediction curve button is in the "Show" state, the subsequent prediction curve of the broken line is displayed, and the prediction curve is presented as a dotted line. Click the settings button in the upper right corner to open the indicator selection pop-up window. The pop-up window uses a dark gray rounded rectangular background. The selection box is a common checkbox, that is, a rectangle with a dark blue border. When it is checked, the rectangle is filled with a dark blue check mark. In the pop-up window, the color block before the indicator option corresponds to the color of the rectangle in the panel. On the line chart, where the user hovers the mouse, a dot and a speech bubble appear on the broken line. The bubble uses white text to briefly summarize the data at the hovered position.

[0091] The technical solution of the embodiment of the present invention determines the spatial dimension and visualization data by performing dimensional analysis on the high-dimensional experimental space; wherein, the visualization data includes the evolution progress of the running rounds, the evolution status of the experimental factors, the experimental indicator scores, the comprehensive evaluation scores and the evolution status of the experimental points; the experimental points include at least one experimental factor; the high-dimensional experimental space is spatially reduced to obtain a low-dimensional experimental space; wherein, the low-dimensional experimental space includes at least one experimental point, a set of experimental indicator scores and at least one comprehensive evaluation score; the low-dimensional experimental space is dimensionalized to obtain a two-dimensional experimental space; wherein, the two-dimensional experimental space includes at least one experimental factor, at least one experimental indicator score and at least one comprehensive evaluation score; the high-dimensional experimental space includes text type data and list data, and data processing of the high-dimensional experimental space includes: mapping the text type data and the list data into a single numerical type, and visually displaying them in a data dashboard; the data dashboard includes five panels, namely, an evolution progress visualization panel, an experimental point evolution visualization panel, a factor evolution visualization panel, a score distribution visualization panel and a score trend visualization panel. The Evolution Progress Visualization Panel uses a donut chart to display the progress of the currently simulated experimental space; the Factor Evolution Visualization Panel uses a cloud-rain chart to display the evolution of experimental factors; the Score Distribution Visualization Panel uses a boxplot to display the distribution of the comprehensive evaluation scores of the current experimental point after the current round of simulation, helping to identify the evolutionary trend of the current experimental point; the Score Trend Visualization Panel uses a curve chart to display the current experimental indicator score change trend and possible future changes, helping to identify the future trend of the current experimental point; and the Experimental Point Evolution Visualization Panel uses a situation map to display the current exploration focus. The data dashboard uses a large and four small layout, with the background of each panel being a white dotted rectangle with blue "L" corners. The panel title is located in the upper left corner, and the settings function is represented by a gear box in the upper right corner or directly above. Sliders are available on the bottom and sides of the panel for dragging and zooming. The data dashboard visualizes high-dimensional data after dimensionality reduction, solving the problem of presenting high-dimensional data and multiple business modules on the same page, making the visualization page simple to operate and intuitive.

[0092] To further understand the present invention, the following describes the use process of the high-dimensional experimental space visualization data dashboard, which includes 6 steps:

[0093] Step 1: Fold the control settings panel on the left, click the Experimental Space Visualization tab, and enter the Experimental Space Visualization Data Panel. The operation steps are as follows: Figure 9 As shown;

[0094] Step 2: Observe the evolution progress visualization panel. Click the evolution progress drop-down menu at the top of the panel to switch the display of running rounds / evaluation scores. After the switch is completed, the corresponding prompt information of the selected content will be displayed at the bottom of the circle. The operation steps are as follows: Figure 10 As shown;

[0095] Step 3: Observe the factor evolution visualization panel. You can drag the slider vertically to view data from other rounds. You can click the settings button in the upper right corner to switch to viewing the evolution data of other experimental factors. The operation steps are shown in 11;

[0096] Step 4: Observe the score distribution visualization panel and drag the slider horizontally to view data from other rounds. Figure 12 As shown;

[0097] Step 5: Observe the scoring trend visualization panel and drag the slider horizontally to view data from other rounds.

[0098] The prediction curve button is in the "show" state by default. Click it to switch to the "hide" state. Accordingly, the prediction curve (dashed line) is hidden in the panel. Figure 13 As shown;

[0099] Click the settings button in the upper right corner to open the indicator selection pop-up window, where you can select which indicator curves to display. Multiple selections are supported. Figure 14 As shown:

[0100] Step 6: Observe the experimental point evolution visualization panel. Click on the clustered experimental point in the panel to pop up detailed information on the historical experimental point, including the value status of each experimental factor at the experimental point, the exploration weight in the sampling algorithm and causal recommendation. If the user observes that the trend of changes in the experimental indicator score and the comprehensive score does not meet expectations, or observes that the value of a certain experimental factor does not meet their own experience judgment value, the exploration weight can be edited. In the personal recommendation ratio after the experimental factor is valued, if the experimental factor needs to be explored intensively, a higher percentage value is assigned, otherwise a smaller value is assigned. Then click the "Confirm" button to send the user's recommended data to the background. The operation steps are as follows: Figure 15 shown.

[0101] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A high-dimensional space visualization method based on a data dashboard, characterized in that: The method comprises: Performing dimensional analysis on the high-dimensional experimental space to determine spatial dimensions and visualization data; the visualization data includes the evolution progress of running rounds, the evolution status of experimental factors, the experimental indicator scores, the comprehensive evaluation scores, and the evolution status of experimental points; the experimental points include at least one of the experimental factors; Performing spatial dimensionality reduction on the high-dimensional experimental space to obtain a low-dimensional experimental space, wherein the low-dimensional experimental space includes at least one experimental point, a set of experimental indicator scores, and at least one comprehensive evaluation score; performing dimensionality reduction on the low-dimensional experimental space to obtain a two-dimensional experimental space, wherein the two-dimensional experimental space includes at least one experimental factor, at least one experimental indicator score, and at least one comprehensive evaluation score; The high-dimensional experimental space includes text type data and list data, and data processing on the high-dimensional experimental space includes: mapping the text type data and the list data into a single numerical type, and visually displaying them in the data dashboard; The data dashboard includes an evolution progress visualization panel, an experimental point evolution visualization panel, a factor evolution visualization panel, a score distribution visualization panel and a score trend visualization panel; wherein, the evolution progress visualization panel displays the progress status of the high-dimensional experimental space of the current running round of deduction based on a donut chart; the experimental point evolution visualization panel is displayed based on a situation diagram, and the clustering movement of the experimental points is used to represent the current exploration focus direction and the degree of convergence of the experimental points; the factor evolution visualization panel displays the evolution status of the experimental factors based on a cloud-rain diagram; the score distribution visualization panel displays the distribution status of the comprehensive evaluation score of the current experimental point after the current round of deduction based on a box plot, and assists in judging the evolution trend of the current experimental point; the score trend visualization panel displays the current experimental indicator score change trend and possible future changes based on a curve chart, and assists in judging the future change trend of the current experimental point; The data dashboard adopts a layout of one large and four small panels. The background of each panel is a white dotted rectangle with four blue "L" corners. The upper left corner of the panel is the panel title, and the upper right corner or the top is the setting function, which is represented by a gear box. The bottom or side of the panel supports sliding bars for dragging or zooming.

2. The high-dimensional experimental space visualization method based on a data dashboard according to claim 1 is characterized in that: The performing dimensional analysis on the high-dimensional experimental space includes: The dimension of the high-dimensional experimental space Ω(N,L) is N, and the number of rounds is L. The high-dimensional experimental space also evolves with the rounds. The dimension calculation formula of the high-dimensional experimental space is as follows: N=N1+N2+N3 N1=N 11 +N 12 +…+N 1i N2=N 21 +N 22 +…+N 2j N3＝1 Among them, N1 is the experimental factor dimension, i is the total number of experimental factors, N2 is the experimental indicator score dimension, j is the total number of experimental indicators, and N3 is the comprehensive evaluation score dimension.

3. The high-dimensional experimental space visualization method based on a data dashboard according to claim 1 is characterized in that: The performing spatial dimensionality reduction on the high-dimensional experimental space includes: The high-dimensional experimental space Ω(N,L) is subjected to spatial dimensionality reduction to obtain low-dimensional experimental spaces (Ω(N1,L),Ω(N2,L),Ω(N3,L)), wherein Ω(N1,L) is the experimental point, Ω(N2,L) is the experimental index score set, and Ω(N3,L) is the comprehensive evaluation score; the low-dimensional experimental space (Ω(N1,L),Ω(N2,L) is subjected to dimensionality reduction to obtain two-dimensional experimental spaces (Ω(N 11 ,L),Ω(N 12 ,L),...,Ω(N 1i ,L)) and (Ω(N 21 ,L),Ω(N 22 ,L),...,Ω(N 2j ,L)), where (Ω(N 11 ,L),Ω(N 12 ,L),...,Ω(N 1i ,L)) are the experimental factor evolution states of experimental factors 1, 2, ..., i, (Ω(N 21 ,L),Ω(N 22 ,L),...,Ω(N 2j ,L)) are the experimental index scores of experimental indicators 1, 2,..., j respectively.

4. The high-dimensional experimental space visualization method based on a data dashboard according to claim 1 is characterized in that: The data processing of the high-dimensional experimental space includes: The mapping relationship between text type data and list data is as follows: {b|b=f(a1,a2),a1,a2∈Ω(N,L)} Among them, a1 is the text type data point in the original experimental space, a2 is the list type data point in the original experimental space, f is the mapping relationship, and b is the numerical type data point after conversion in the experimental space.

5. The high-dimensional experimental space visualization method based on a data dashboard according to any one of claim 1, characterized in that: In the evolution progress visualization panel, the donut chart is located in the center of the evolution progress visualization panel, the donut chart is transparent as a whole, and the progress bar changes color gradually; the currently selected evolution iteration target is indicated directly above the donut chart, and the iteration target includes the running round and the comprehensive evaluation score; two lines of text are displayed in the center of the donut chart, the first line is the current deduction round, expressed in a numerical value, and the second line is the current evolution progress, expressed in percentage; the remaining estimated completion time of the current round and the remaining estimated total completion time are displayed directly below the donut chart; the prediction of the remaining estimated completion time of the current round and the remaining estimated total completion time adopts a big data statistical prediction algorithm, and a statistical prediction model is determined based on the current evolution progress and completion time, and the prediction model is used to update the remaining completion time in real time; Select the evolution iteration target through the setting function. When the evolution iteration target is the running round, the ring progress calculation formula is as follows: Among them, p i is the current deduction round, p0 is the upper limit of the round, y i The deduction progress of each sample in the current deduction round; When the iteration target is a comprehensive evaluation score, the formula for calculating the ring progress is as follows: Among them, maxp i is the highest comprehensive evaluation score of the experimental points in the current deduction round, p0 is the expected score, y i It indicates the deduction progress of each sample in the current deduction round.

6. The high-dimensional experimental space visualization method based on a data dashboard according to any one of claim 1, characterized in that: In the factor evolution visualization panel, click the setting function to pop up the experimental factor switching selection dialog box, which has a gray-black rounded rectangular background and a factor selection box with a blue hollow circle as the radio button. When one of the experimental factors is selected, the blue hollow circle jumps to a blue solid circle, and the experimental factor distribution status of the experimental factor in multiple rounds is displayed in the panel. The experimental factor distribution status in each round is represented by a cloud-rain graph, which uses red, green, purple, brown, and orange as a group for cyclic display; the panel can be slid vertically; the clouds in the cloud-rain graph represent the value distribution of the current experimental factor, and the rain represents the value size and number of the current experimental point; the horizontal axis of the cloud-rain graph represents the score, and the vertical axis represents the evolution round, wherein the vertical lines are distributed at 20 equal intervals.

7. The high-dimensional experimental space visualization method based on a data dashboard according to any one of claim 1, characterized in that: In the score distribution visualization panel, the horizontal axis of the box plot represents the round, and the vertical axis represents the score value, wherein the horizontal distribution is divided into 10 equal intervals. The panel can slide horizontally to display the score data of all rounds; the score of each round is represented in the form of a box plot combined with a dot-shaped scatter plot, and is displayed in a cycle through red, green, purple, brown and orange; the scatter plot has the same color as the box plot, and the transparency of the scatter plot is lower than that of the box plot; the smaller the score distribution range in the box plot and the higher the upper limit, the better the evolution of the current experimental point, and vice versa.

8. The high-dimensional experimental space visualization method based on a data dashboard according to any one of claim 1, characterized in that: In the scoring trend visualization panel, click the setting function to pop up the indicator selection pop-up box, which includes multiple indicators. The pop-up box has a dark gray rounded rectangular background, and the indicator selection box is a check box, which is a rectangular box with a dark blue border. When one of the indicators is selected, the rectangular box is filled with a dark blue check mark; the panel uses multiple curves to represent the scores of multiple indicators in multiple rounds, and the curves corresponding to the indicators are the same color as the color blocks before the indicator options in the pop-up box; the horizontal axis of the curve represents the round, and the vertical axis represents the score, which is horizontally distributed at 10 equal intervals. The panel can be slid horizontally to show the indicator change trends of the multiple indicators in all rounds; the higher the experimental indicator score, the better the evolution direction of the current experimental point, and vice versa; when the mouse hovers over a point in the curve, a dot and a dialogue bubble appear on the curve, and a brief summary of the data of the point is displayed in the bubble, where the text is white; The scoring trend visualization panel is also provided with a prediction curve button. When the prediction curve button is in the display state, the subsequent prediction curve is displayed to indicate the change state of the experimental indicator score in future rounds. The prediction curve is presented by a dotted line and is predicted using a least squares regression prediction algorithm.

9. The high-dimensional experimental space visualization method based on a data dashboard according to any one of claim 1, characterized in that: In the experimental point evolution visualization panel, the system sequentially assigns an initial position to each experimental point as a label of the experimental point, clusters the experimental points using a clustering method, and uses the obtained cluster clusters as the current exploration focus direction; wherein the clustering method includes but is not limited to Euclidean distance.

10. The high-dimensional experimental space visualization method based on a data dashboard according to claim 9, characterized in that: The experimental point evolution visualization panel is a large panel in which the experimental points are displayed as dots. The color of the dots gradually becomes lighter from the center to the edge, and the dots are repeatedly scaled outward from the center point, presenting a flickering effect. The dots include gray dots and red dots. The gray dots represent the experimental points of the previous two rounds, and the red dots represent the experimental points of the previous round. The panel supports vertical and horizontal sliding. Click on a situation point in the situation diagram to display the values ​​of each experimental factor and the exploration proportion in the current experimental point in a list. If the evolution direction of the experimental factor evolution visualization, score distribution visualization, and score trend visualization does not meet expectations, the exploration proportion will be modified and sent to the spatial exploration module. In the next round of deduction, the spatial exploration algorithm will adjust the exploration direction according to the modification.

Images