A multi-dimensional risk information visualization method and system based on cognitive load optimization

Abstract

The application provides a multi-dimensional risk information visualization method and system based on cognitive load optimization, and relates to the technical field of data visualization. It includes: social risk multi-dimensional data attribute analysis and classification; calculating the perception efficiency coefficient of each channel; fusing scene prior, data-driven and user-defined three-source signal to calculate information importance weight, and combining dimension complexity to construct cognitive load function; under the constraint of channel mutual exclusion, the optimal mapping configuration of data attribute to visual channel is solved with the goal of minimizing total cognitive load; the current task type of the user is identified, the view importance weight is calculated, and after stability control processing, the screen space is dynamically optimized and allocated through a nonlinear allocation algorithm; finally, a multi-view linkage interface is rendered and presented. Through the perception efficiency driven visual encoding optimization and task driven layout dynamic adaptation, the cognitive load of the management personnel is significantly reduced, and the risk judgment efficiency and system usability are improved.

Description

Technical Field

[0001] This application relates to the field of data visualization technology, specifically to a method and system for visualizing multidimensional risk information based on cognitive load optimization. Background Technology

[0002] The visualization needs in the field of social risk governance have distinct industry characteristics: risk data naturally carries multi-dimensional attributes, such as time, space, semantics, and relational dimensions. To support managers in making multi-level decisions and judgments from macro-level trends to micro-level evidence, existing visualization systems typically adopt a multi-view dashboard architecture, presenting various visualization components such as geographic information system views, time-series trend charts, and relational network diagrams in parallel on the same screen.

[0003] However, existing technologies have several shortcomings. First, current system design philosophies are data integrity-oriented, striving to present as much risk dimension information as possible on a single interface, leading to severe information overload. Human working memory capacity is limited; when the density of visual elements exceeds the cognitive processing threshold, managers must expend additional cognitive resources to filter information, resulting in decision fatigue and judgment bias. Furthermore, existing systems lack a quantitative mapping model from data attributes to cognitive load, making it impossible to predict and optimize the cognitive load level of the interface during the design phase, or to dynamically adjust the granularity of information presentation during runtime.

[0004] Secondly, the visual channel is used arbitrarily, failing to follow the principles of perceptual psychology. The visual channel is a mapping channel from data attributes to graphic attributes, and different channels have significantly different transmission efficiencies for the human perceptual system. Existing systems often rely on developer experience or default configurations of chart libraries when selecting visual channels, lacking a scientific basis for selection. For example, quantitative risk scores requiring precise comparison are mapped to a color saturation channel with low discrimination, while qualitative event types requiring only rough classification are mapped to independent shape symbols that consume a large amount of cognitive resources, resulting in a serious waste of cognitive resources.

[0005] Secondly, the view layout lacks task orientation, and multi-view linkage leads to frequent reflows and visual jitter. Existing dashboard systems typically use fixed grids or manual configuration for multi-view layouts, with each view maintaining a constant screen space quota. When users switch between different analysis tasks, core views cannot obtain sufficient display space, while irrelevant views occupy valuable screen resources. Some systems attempt to dynamically adjust the layout, but lack stability control, resulting in excessively frequent layout adjustments. Each reflow triggers a visual repositioning process for the user, causing interface jitter and severely impacting the continuity and immersion of analysis tasks. Furthermore, existing technologies lack scientifically quantifiable visual coding configuration methods, and the optimization process relies on human experience, making it difficult to cope with the combinatorial explosion problem when data dimensions increase. Summary of the Invention

[0006] The purpose of this application is to provide a method, system, electronic device and storage medium for visualizing multidimensional risk information based on cognitive load optimization, which can realize the visualization of multidimensional social risk information that is cognitively friendly, task-adaptive and interactively stable.

[0007] In a first aspect, embodiments of this application provide a method for visualizing multidimensional risk information based on cognitive load optimization, including: Obtain a social risk multidimensional dataset, perform type parsing on each data attribute in the dataset, and classify it into one of the following: quantitative attribute, ordinal attribute, categorical attribute, relational attribute, hierarchical attribute, and composite multidimensional attribute; for composite multidimensional attributes, further perform sub-attribute decomposition to break them down into several basic type attributes; Based on the pre-defined visual channel perception effectiveness ranking, a decay model is used to calculate the quantified perception efficiency coefficient for each visual channel, and a mapping rule library from attribute type to candidate visual channel is established. For each data attribute, its information importance weight and dimensional complexity are calculated; the information importance weight is obtained by fusing three information sources: scenario prior, data-driven, and user-defined. Based on the information importance weight, dimensional complexity, and perception efficiency coefficient, a cognitive load function is constructed. With the goal of minimizing the total cognitive load, the optimal mapping configuration from each data attribute to the visual channel is solved under the constraint of channel mutual exclusion. Identify the user's current risk assessment task type; dynamically calculate the layout importance weight and alarm importance weight of each view based on three types of driving signals: data changes, interaction intent, and scenario priors. The layout importance weights are subjected to stability control processing, which includes at least one of interactive locking, risk hysteresis, cooling constraints and smoothing processing. Based on the processed weights, under the constraints of minimum visibility lower limit and screen area conservation, a nonlinear allocation algorithm is used to dynamically calculate the target screen space quota for each view and achieve a smooth transition of the layout. Based on the optimal mapping configuration and the target screen space quota, a multi-view interactive interface is rendered and presented; wherein, the views achieve interactive linkage, including scrolling linkage, cross-view synchronous highlighting, and shared filter conditions, through a global state manager.

[0008] Secondly, embodiments of this application provide a multi-dimensional risk information visualization system based on cognitive load optimization, including: The attribute classification module is used to obtain a social risk multidimensional dataset, perform type parsing on each data attribute in the dataset, and classify it into one of the following: quantitative attribute, ordinal attribute, categorical attribute, relational attribute, hierarchical attribute, and composite multidimensional attribute; for composite multidimensional attributes, further perform sub-attribute decomposition to break them down into several basic type attributes. The channel mapping module is used to rank the visual channels based on the preset visual channel perception effectiveness, calculate the quantified perception efficiency coefficient for each visual channel using a decay model, and establish a mapping rule library from attribute type to candidate visual channels. The load optimization module is used to calculate the information importance weight and dimensional complexity of each data attribute. The information importance weight is obtained by fusing three information sources: scene prior, data-driven, and user-defined. Based on the information importance weight, dimensional complexity, and perceptual efficiency coefficient, a cognitive load function is constructed. With the goal of minimizing the total cognitive load, the optimal mapping configuration from each data attribute to the visual channel is solved under constraints including channel mutual exclusion. The task identification module is used to identify the type of risk assessment task currently being performed by the user; based on three types of driving signals—data changes, interaction intent, and scenario priors—it dynamically calculates the layout importance weight and alarm importance weight of each view. A stability control module is used to perform stability control processing on the layout importance weights. The stability control processing includes at least one of interactive locking, risk hysteresis, cooling constraints, and smoothing processing. Based on the processed weights, under the constraints of minimum visibility lower limit and screen area conservation, a nonlinear allocation algorithm is used to dynamically calculate the target screen space quota for each view and achieve a smooth transition of the layout. The rendering module is used to render and present a multi-view interactive interface based on the optimal mapping configuration and the target screen space quota. The views interact with each other through a global state manager, including interactive filtering, cross-view synchronized highlighting, and shared filter conditions.

[0009] Thirdly, this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multidimensional risk information visualization method based on cognitive load optimization as described in the first aspect.

[0010] Fourthly, this application provides a computer-readable storage medium storing computer-readable instructions that can be executed by a processor to implement the multidimensional risk information visualization method based on cognitive load optimization as described in the first aspect.

[0011] Compared to existing technologies, the multi-dimensional risk information visualization method based on cognitive load optimization provided in this application firstly ensures that important risk information is always presented through the most efficient visualization channel by automatically allocating visual channels driven by perception efficiency, significantly reducing the cognitive load on managers. Secondly, task-driven view layout optimization ensures that the core views required for the current assessment task receive sufficient screen space, improving the efficiency and accuracy of risk assessment tasks. Thirdly, the layout stability control mechanism effectively suppresses interface jitter caused by alarm triggering and interactive operations, enhancing the stability and immersiveness of multi-view linkage interaction. Finally, the cognitive load optimization solution algorithm transforms the visual encoding configuration process, which originally relied on manual experience, into an automatically solvable mathematical optimization problem, achieving automation and scientification of the configuration. As can be seen, this invention combines the visual perception efficiency theory in cognitive psychology with the multidimensional data characteristics of social risks, and establishes a quantitative mathematical model between data attributes, visual channels, and cognitive load. Under the mutual exclusion constraint of channels, it solves the optimal visual encoding configuration with the minimum total cognitive load. At the same time, it introduces a task-driven view importance assessment mechanism, dynamically allocates the screen space quota of each view according to the user's current assessment task, and suppresses layout jitter through stability control measures, thus realizing a cognitively friendly, task-adaptive, and interactively stable visualization of multidimensional social risk information. Attached Figure Description

[0012] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 The flowchart illustrates a method for visualizing multidimensional risk information based on cognitive load optimization, as provided in an embodiment of this application. Figure 2 This illustration shows a specific visualization process for multidimensional risk information based on cognitive load optimization provided in an embodiment of this application; Figure 3 This illustration shows a schematic diagram of the structure of a multidimensional risk information visualization system based on cognitive load optimization provided in an embodiment of this application; Figure 4 A schematic diagram of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0013] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0014] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains.

[0015] Furthermore, the terms "first" and "second," etc., are used to distinguish different objects, not to describe a specific order. Additionally, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to those processes, methods, products, or apparatuses.

[0016] The CogniRisk-Vis visualization system of this invention is implemented using a B / S architecture, with front-end and back-end separated.

[0017] Please refer to Figure 1 , Figure 1 This is a flowchart of a multidimensional risk information visualization method based on cognitive load optimization provided in an embodiment of this application, including the following steps: S101. Data Attribute Classification: Obtain the social risk multidimensional dataset, perform type parsing on each data attribute in the dataset, and classify it into one of the following: quantitative attribute, ordinal attribute, categorical attribute, relational attribute, hierarchical attribute, and composite multidimensional attribute; for composite multidimensional attributes, further perform sub-attribute decomposition to break them down into several basic type attributes.

[0018] The quantitative attributes include risk deviation score, estimated scale of dissemination impact, or attribution contribution value; the ordinal attributes include warning level or risk evolution trend level; the categorical attributes include event type or information source type; the relational attributes include dissemination link or subject interaction relationship; and the hierarchical attributes include administrative division level or event classification level.

[0019] Social risk multidimensional datasets typically originate from diverse and heterogeneous sources, including but not limited to: alarm record databases (timestamps, geographic locations, event descriptions, involved parties), social media information collection systems (posted content, propagation chains, interaction relationships), geographic information systems (administrative division levels, regional population density, distribution of sensitive locations), risk scoring engines (deep support vector data describing DSVDD deviation scores, context-corrected scores, SHAP attribution contribution values), and propagation extrapolation model outputs (risk level evolution predictions, social impact scope estimations). For example, a social risk multidimensional dataset arising from a public health event.

[0020] The purpose of this step is to classify the attributes of each dimension in the social risk multidimensional dataset into six basic types according to their semantic features and mathematical properties, so as to establish a classification basis for subsequent visual channel mapping.

[0021] For ease of understanding, the present invention also provides, as follows: Figure 2 The diagram illustrates a specific process for visualizing multidimensional risk information based on cognitive load optimization.

[0022] refer to Figure 2 This step involves examining each data attribute in the aforementioned multidimensional dataset. a i The execution type resolution classifies it into one of the following six attribute types: 1. Quantitative Attributes (Q): These are continuous or discrete numerical attributes with explicit mathematical meaning and an absolute zero point. In social risk scenarios, typical quantitative attributes include: risk deviation score s∈[0, 100], estimated propagation impact size n, and SHAP attribution contribution value c. Quantitative attributes support arithmetic operations such as summation, difference, mean, and proportion.

[0023] 2. Ordinal Attributes (O): Discrete hierarchical attributes with a clear order but varying intervals. In social risk scenarios, typical ordinal attributes include: warning level (Level 1 / Level 2 / Level 3 / Level 4), risk evolution trend level (rising / stable / falling), and response priority (urgent / important / general). Ordinal attributes support comparison and sorting, but do not support precise arithmetic operations.

[0024] 3. Categorical Attributes (C): Discrete label attributes without inherent order. In social risk scenarios, typical categorical attributes include: event type (security incident / public safety incident / mass incident / cybersecurity incident), type of involved party (individual / organization / institution / anonymous), and type of information source (emergency call / social media / citizen complaint / manual report). Categorical attributes only support equivalence judgment and frequency statistics.

[0025] 4. Relational Attributes (R): Graph structure attributes that describe the relationships between entities. In social risk scenarios, typical relational attributes include: propagation links (carrier A is forwarded and cited by carrier B), subject interaction relationships (subject X and subject Y have opposing viewpoints), and event association relationships (event M and event N are spatially and temporally proximate or causally related). Relational attributes are essentially sets of entity pairs and need to be encoded through spatial relationship channels such as connection, proximity, and containment.

[0026] 5. Hierarchical Attributes (H): Tree-structured attributes describing nested or subordinate relationships. In social risk scenarios, typical hierarchical attributes include: administrative division hierarchy (country → province → city → district / county → street), event classification hierarchy (major category → intermediate category → minor category → subcategory), and organizational structure hierarchy (general team → branch team → brigade → squadron). Hierarchical attributes need to be encoded through channels such as inclusion, nesting, and tree layout.

[0027] 6. Composite Multidimensional Attributes (X): These are hybrid attributes that simultaneously carry multiple semantic characteristics mentioned above. They need to be decomposed and encoded separately or a composite encoding strategy should be adopted. For example, the "spatiotemporal context" attribute of a risk event contains three sub-dimensions: timestamp (quantitative / ordinal), geographic location (hierarchical / quantitative), and regional sensitivity (ordinal), which is a typical composite multidimensional attribute.

[0028] Classification rules: For each data attribute a i The following determination process will be executed: if a i If the range of values ​​is a numerical interval and there exists an absolute zero, then the type is Q; if a i If the range of values ​​is a finite ordered set, then the type is O; if a i If the range of values ​​is a finite unordered set, then the type is C; if a i If the range of values ​​is a finite ordered set, then the type is O; if a i The type is R if it describes a binary association between entities; if a i The type is H to describe nested membership between entities; if a iIf it contains multiple semantics as described above, then the type is X.

[0029] For complex multidimensional attributes, this step further performs sub-attribute decomposition, breaking them down into several basic type attributes. a i ={ a i,1 , a i,2 , …, a i,k Each sub-attribute independently participates in subsequent encoding allocation.

[0030] S102, Visual Channel Efficiency Initialization: Based on the preset visual channel perception effectiveness ranking, a decay model is used to calculate the quantified perception efficiency coefficient for each visual channel, and a mapping rule library from attribute type to candidate visual channel is established.

[0031] The purpose of this step is to establish a mathematical measurement system for channel efficiency by assigning quantitative perception efficiency coefficients to various visual channels based on research findings on visual perception effectiveness in cognitive psychology.

[0032] According to existing research, visual channels are ranked from high to low in terms of perceptual effectiveness, as shown in Table 1 below.

[0033] Table 1: Visual Channel Ordering

[0034] This invention uses an exponential decay model to calculate the perceptual efficiency coefficient of each visual channel: e channel = τ (rank-1) ； in, e channel The perceived efficiency coefficient. rank This represents the ranking of the visual channel in the above perceptual effectiveness ranking. t It is a decay factor with a value range of (0,1); in the ranking of perception effectiveness, the position channel ranks highest and the shape channel ranks lowest.

[0035] use t When = 0.8, the efficiency coefficients of each visual channel are calculated as shown in Table 2 below.

[0036] Table 2: Efficiency coefficients of the visual channel

[0037] Not all visual channels are suitable for all data attribute types. This invention establishes a mapping rule base from attribute type to candidate visual channels to ensure the consistency of encoding semantics, as shown in Table 3 below.

[0038] Table 3: Mapping rule base from attribute type to candidate visual channel

[0039] S103. Calculation of Information Importance and Dimensional Complexity: For each data attribute, calculate its information importance weight and dimensional complexity; the information importance weight is obtained by fusing three information sources: scenario prior, data-driven, and user-defined. The purpose of this step is to calculate the information importance weight and dimensional complexity for each data attribute, which will then serve as input parameters for the subsequent cognitive load function.

[0040] The information importance weight reflects the decision-making value of the data attribute in the current analysis scenario. The fusion calculation formula for the information importance weight is as follows: w imp (a i )=α · w scene (a i ) + β · w data (a i ) + γ · w user (a i )； in, w imp (a i ) For data attributes a i Information importance weight; w scene (a i ) The prior importance of a scene is determined by the type of task being assessed. For example, in situational awareness tasks, the prior importance of spatial location attributes is significantly higher than that of temporal attributes. w data (a i ) Data-driven importance is determined by the statistical characteristics of the attribute. For example, attributes with higher variance in the current subset of data have higher discriminative power and are therefore more important. w user (a i) Allows users to customize importance, enabling administrators to manually adjust the priority of certain attributes based on business experience; α , β , c The preset fusion weight coefficients, and satisfy the following conditions: a + b + c = 1. For example, α= 0.5 β= 0.3 c= 0.2.

[0041] The specific value of the prior importance of the scenario is preset by the association matrix of task type and attribute type in Table 4 below.

[0042] Table 4: Association Matrix between Task Type and Attribute Type

[0043] Data-driven importance is calculated based on data distribution characteristics: w data (a i )= ; Wherein, σ( a i σ represents the standard deviation of this attribute on the current subset of data (reflecting its discriminative power). max H represents the maximum standard deviation of all data attributes within the current dataset; a i H represents the information entropy (reflecting the amount of information) of this attribute. max This represents the maximum value of the information entropy of all data attributes within the current dataset; the clip function truncates the result to the [0, 1] interval.

[0044] Dimensional complexity reflects the cognitive processing difficulty required for data attributes to be visualized. Complexity is related to factors such as attribute type, cardinality of the value range, and value range structure. The calculation formula is as follows: d complexity (a i ) = base t × log 2 (|D(a i )| + 1 ) × ( 1 + the · struct i )； in, d complexity (ai ) For data attributes a i dimensional complexity; base t For attribute type t The corresponding baseline complexity values ​​are: Quantitative = 1.0, Ordinal = 1.2, Categorical = 1.5, Relational = 2.0, Hierarchical = 1.8, Composite = 2.5; |D(a i )| The cardinality of the attribute's value range (the number of different values); structure i To determine the structural complexity index, we set it to 0 for ordered attributes, 1 for unordered attributes, and 2 for graph / tree structure attributes. or The preferred structural complexity weighting coefficients are those used for optimization. or = 0.3.

[0045] S104. Visual encoding optimization solution: Based on the information importance weight, dimensional complexity and perceptual efficiency coefficient, construct a cognitive load function. With the goal of minimizing the total cognitive load, under the constraint of channel mutual exclusion, solve the optimal mapping configuration from each data attribute to the visual channel. One of the core innovations of this invention lies in establishing a mathematical mapping from "data attribute → visual channel → cognitive load". For each data attribute and each candidate visual channel, the cognitive load function is defined as follows: CL(a i , v j ) = (w imp (a i ) ×d complexity (a i ) / e channel (v j )； in, CL(a i , v j ) Indicates data attributes a i Mapping to the visual channel v j Cognitive load during the process; echannel (v j ) For visual channel v j The perceived efficiency coefficient; w imp (a i ) For data attributes a i Information importance weight; d complexity (a i ) For data attributes a i dimensional complexity; Function design principles: The numerator of the cognitive load function w imp (a i ) ×d complexity (a i ) This represents the total amount of information that the cognitive system needs to process for this data attribute. The higher the importance of the information, the greater the necessity for managers to pay attention to and understand this attribute; the higher the dimensional complexity, the greater the cognitive processing difficulty of this attribute itself. The product of the two represents "the original cognitive cost that managers need to pay if visual encoding optimization is not performed".

[0046] The denominator of the cognitive load function e channel (v j ) This indicates the efficiency of information transmission through the visual channel. The higher the efficiency coefficient (e.g., 1.0 for the position channel), the less cognitive resources are required to transmit the same amount of information through this channel; the lower the efficiency coefficient (e.g., 0.069 for the shape channel), the more cognitive resources are required to transmit the same amount of information through this channel.

[0047] Therefore, cognitive load function CL(a i , v j ) The physical meaning is: to transfer data attributes a i Mapping to the visual channel v j This refers to the cognitive load that managers incur to understand the code. The lower the cognitive load, the more "cognitively friendly" the code configuration is.

[0048] The total cognitive load CL total (M) represents the sum of the cognitive load of all data attribute and visual channel mapping pairs in the visualization configuration scheme M.

[0049] Total cognitive load calculation: For a visualization configuration scheme M containing n data attributes and m visual channels, where M( a i ) = v j Represents attributes a i Assigned to a channel v j The total cognitive load is: ; The optimization objective of this invention is to find a solution that makes CL... total (M) Minimal encoding configuration scheme M * .

[0050] In some embodiments, the constraints for visual channel allocation include at least one of the following: Constraint 1 (Single-channel exclusive constraint): The same visual channel can be assigned to at most one data attribute at the same view level (except for sub-attributes of composite multidimensional attributes, which are usually assigned to different views).

[0051] Constraint 2 (Attribute Type and Channel Type Matching Constraint): Data attributes can only be assigned to the candidate channel set corresponding to their attribute type. For example, quantitative attributes cannot be mapped to the hue channel, and categorical attributes cannot be mapped to the length channel.

[0052] Constraint 3 (Channel Combination Mutual Exclusion Constraint): Certain visual channel combinations can cause perceptual interference when used simultaneously, and should be avoided when allocating them at the same time. The main mutually exclusive combinations include: "Area + Length" (highly correlated and difficult to decode independently), "Hue + Saturation" (hue changes interfere with the accurate perception of saturation), "Lightness + Hue" (lightness changes are mistaken for hue changes or vice versa), and "Angle + Slope" (perceptually equivalent).

[0053] Constraint 4 (Channel Capacity Limit): The perceptual discrimination capacity of some visual channels is limited; exceeding the limit leads to a sharp decrease in discrimination accuracy. For example, the reliable upper limit for human discrimination of discrete hues is approximately 8-12, for discrete shapes approximately 5-7, and for discrete textures approximately 4-6. When the cardinality of a data attribute's value range exceeds the channel capacity limit, that attribute should not be assigned to that channel.

[0054] Constraint 5 (View Hierarchy Constraint): In a dashboard system containing multiple views, different views have independent visual channel namespaces. The same data attribute can be mapped to different visual channels in different views (for example, a risk score is mapped to color brightness in a GIS view and to the vertical axis position in a time-series view), but all of the above constraints apply within the same view.

[0055] Under the constraints established above, we seek the optimal mapping configuration from data attributes to the visual channel that minimizes the total cognitive load.

[0056] Problem Modeling: The visual channel allocation problem is modeled as a constrained combinatorial optimization problem. Decision variable x is defined. i,j ∈{0, 1}, where x i,j = 1 indicates data attribute a i Assigned to the visual channel v j x i,j = 0 indicates that it has not been allocated.

[0057] The objective function is to minimize the total cognitive load: .

[0058] This invention employs a hybrid strategy combining branch and bound with greedy initialization to solve the aforementioned integer programming problem: Phase 1: Greedy Initialization For each data attribute a i According to cognitive load CL(a i , v j ) Traverse candidate channels in ascending order v j The first channel that satisfies all constraints is selected as the initial allocation. The time complexity of the greedy initialization algorithm is O(n^2). m), which can generate feasible solutions in milliseconds, providing an initial upper bound for subsequent optimization.

[0059] Phase Two: Branch and Bound Optimization Using the total cognitive load of the greedy solution as the initial upper bound, a better solution is systematically searched using a branch and bound algorithm: 1. Branching strategy: Select the attribute with the highest dimensionality complexity among the currently unassigned attributes and branch accordingly, prioritizing the assignment of high-complexity attributes to reduce the search space; 2. Bounding strategy: Calculate the lower bound of the current partial allocation scheme (the sum of the minimum possible cognitive loads of the remaining unallocated attributes). If the lower bound is greater than the current optimal upper bound, then prune. 3. Backtracking strategy: When the current branch cannot find a better solution, backtrack to the previous step and try other channel allocations.

[0060] For typical social risk visualization scenarios (number of data attributes n≤12, number of channels m≤11), the branch and bound algorithm can converge to the global optimum in sub-seconds.

[0061] When the system contains multiple views (GIS view, time series view, relational view), each view has an independent channel namespace. This step performs the above optimization solution for each view separately, and the views are weakly coupled through the coding consistency constraint of shared data attributes (for example, the same risk score attribute should maintain the consistency of coding direction in different views, such as high score corresponding to high brightness or high position).

[0062] After optimization, the output encoding configuration scheme M is completed. * It includes the optimal mapping relationship from each data attribute to the visual channel, the corresponding cognitive load value, and the total cognitive load statistics for each view.

[0063] S105, View Importance Weight Calculation: Identify the user's current risk assessment task type; dynamically calculate the layout importance weight and alarm importance weight of each view based on three types of driving signals: data changes, interaction intent, and scenario priors. One of the purposes of this step is to identify the user's current risk assessment task type, providing task context for subsequent view importance weight calculation and layout optimization.

[0064] The assessment tasks in social risk early warning scenarios can be summarized into three basic types: 1. Situation Awareness Task: Users need to quickly understand the spatial distribution of risk events within their jurisdiction, the location of hotspot areas, and the overall risk level. This type of task heavily relies on GIS spatial views and requires identifying risk clustering patterns and spatial anomalies against a large-scale geographic background.

[0065] 2. Trend Analysis Task: Users need to understand the temporal evolution of risk events, recent fluctuation trends, and future trajectory predictions. This type of task heavily relies on time-series change views and requires identifying periodicity, abrupt change points, and confidence distributions for projected prediction intervals on the timeline.

[0066] 3. Association Mining: Users need to uncover the propagation relationships, interactions between stakeholders, and causal chains between risk events. This type of task heavily relies on network relationship views and requires identifying key propagation nodes, core links, and adversarial structures within the graph structure.

[0067] Task identification method: This invention employs a multi-source signal fusion method for task identification, combining the following three types of signals to determine the user's current task type: Signal 1: Explicit task switching signal The system provides a task switching control (three main function buttons: Status, Trend, and Association). When the user actively clicks to switch, the system directly records the current task type. T current Explicit switching has the highest confidence level and the highest priority.

[0068] Signal 2: View Interaction Behavior Signal The system monitors the frequency and type of user interactions across various views, inferring the current task tendency through implicit behavior. If the frequency of zooming, panning, and clicking operations on the GIS view is significantly higher than that on other views, and the interaction depth (drill-down level) is relatively deep, then it is inferred to be a situational awareness task. If users frequently perform operations such as selecting time ranges, hovering over curves, and viewing prediction intervals in the time series view, it is inferred that it is a trend analysis task. If users frequently perform node dragging, path tracing, and community expansion operations in the relationship view, it is inferred that this is an association mining task.

[0069] The task propensity probability of the behavioral signal is calculated as follows: ; Among them, V k For task type T k The core view collection, w v f is the behavior weight coefficient for view v. v The frequency of interaction on view v.

[0070] Signal 3: Data-driven scenario signals The system analyzes the characteristic patterns of the current data subset to determine the most likely areas for analysis. If the spatial distribution entropy of the current data subset is low (there is obvious hot spot clustering), it tends to be a situational awareness task; If the time series variance in the current data subset is high (with obvious fluctuations or abrupt changes), then trend analysis tasks are preferred. If the relationship graph density in the current data subset is high (meaning there are complex propagation paths), then the association mining task is more likely to be performed.

[0071] Task fusion decision: By combining the three types of signals, a weighted voting mechanism is used to determine the final task type: ; Wherein, λ1, λ2, and λ3 are signal fusion weights, with λ1= 0.6, λ2= 0.3, and λ3= 0.1 being preferred to satisfy the dominance of explicit signals.

[0072] The second objective of this step is to dynamically calculate the importance weight of each view in the current judgment context based on three driving signals: data changes, interaction intentions, and scenario priors.

[0073] Specifically, this invention divides view importance into two categories: layout importance (determining the size quota of the view in screen space) and alarm importance (determining whether to trigger alarm prompts and non-intrusive notifications).

[0074] Definition of three-source drive signals: Data change signals S data ( v ) : This reflects the degree of drastic change in the data presented in view v within the most recent time window. For GIS views, the data change signal is the rate of change in the spatial distribution of the risk heatmap; for time-series views, the data change signal is the year-on-year / month-on-month change in the risk score for the latest period; for relationship views, the data change signal is the number of added / deleted edges in the graph structure and the rate of change in the community structure.

[0075] ; Where ΔD(v) is the amount of data change in view v within the most recent time window Δt, D base (v) represents the baseline change.

[0076] Interaction Intent Signals S intent ( v ) : This reflects the user's recent engagement with view v. The interaction intent signal is calculated using three sub-indicators: interaction frequency, interaction depth, and interaction duration. ; Among them, f v d represents the interaction frequency (number of operations per unit time). v t represents the interaction depth (number of drill-down levels). vThe interaction duration (view dwell time) is represented by 'max', which represents the maximum value of the metric. The metrics are multiplied together after being normalized to the [0, 1] interval.

[0077] Scene prior signals S scene ( v , T current ) : View v reflects the current task type being analyzed. T current The prior importance of the signal is determined by the predefined task type and view association matrix shown in Table 5 below.

[0078] Table 5: Relationship Matrix between Task Type and View

[0079] Importance of layout I layout ( v ) Determines the priority of view v in screen space allocation: I layout (v) = w 1 · S scene (v, T current ) + w 2 · S intent (v) + w 3 · S data (v) ； in, S scene (v, T current ) represents the prior signal of the scene. S intent (v) represents the interaction intent signal. S data (v) represents the data change signal. w 1. w 2. w 3 is the preset fusion weight and w 1 +w 2 +w The ratio 3=1 is used to ensure the dominant role of the scenario prior, while also taking into account the dynamic adjustment role of user interaction intentions and data change signals.

[0080] Importance of Alarms I alert (v) Determine whether view v triggers an alert: Ialert (v) = w 1 S data (v) + w 2 S scene (v, T current ); Among them, the importance of alarms is mainly based on data change signals. w 1 = 0.6), with scene priors as an auxiliary factor ( w 2 = 0.4). When I alert (v)> i alert (Alarm threshold, preferred) i alert When = 0.7), trigger a non-intrusive alert for view v (such as a change in the view title bar color or a gentle blinking indicator, rather than a pop-up interruption).

[0081] S106. Layout stability control and space allocation: Perform stability control processing on the layout importance weights. The stability control processing includes at least one of interactive locking, risk hysteresis, cooling constraint and smoothing processing. Based on the processed weights, under the constraints of minimum visibility lower limit and screen area conservation, a nonlinear allocation algorithm is used to dynamically calculate the target screen space quota of each view and realize the smooth transition of the layout. The purpose of this step is to control the triggering conditions and update rhythm of layout adjustments through a state machine mechanism, thereby suppressing interface jitter and maintaining the user's visual spatial memory.

[0082] This invention designs a four-fold stability control mechanism to form a complete state transition system: 1. Interaction Lock Mechanism When a user performs a high-intensity interactive operation in any view (the frequency of continuous operations exceeds the threshold f) lock = 2 times / second, or in drill-down / drag mode), the system enters "interaction lock" state. In lock state, any global layout reordering triggered by data changes or alarm signals is prohibited, and only the local view directly associated with the user's current interaction is updated.

[0083] Duration T of interactive lock lock The timestamp t of the most recent interaction last and the lock-in cooling period Δt lock (Optimized 3 seconds) Decision: Locked(t) = 1 (t - t) last <Δt lock); 2. Risk Hysteresis To avoid frequent layout adjustments caused by minor data fluctuations, this invention introduces a hysteresis band. Layout updates are only triggered when the change in view importance exceeds a hysteresis threshold. ; That is, by subtracting the importance used in the previous round of stable layout from the latest calculated importance, and taking the absolute value, we obtain the change in the importance of this round of layout.

[0084] ; Among them, δ hys The preferred δ value is the hysteresis ratio threshold. hys = 0.15 (meaning an update is triggered only if the importance change exceeds 15%). In other words, the hysteresis threshold is set based on the old, stable importance value.

[0085] 3. Cooldown Constraint Set the global minimum layout update interval ΔT min (Preferred 5 seconds) Regardless of changes in importance signals, the time interval between two consecutive global layout rearrangements must not be less than this threshold: When the system is in a cooldown period, all layout update requests are temporarily stored in a buffer queue and then merged and executed all at once after the cooldown period ends.

[0086] 4. Exponential Moving Average Smoothing (EMA Smoothing) Even after passing the triple stability checks described above, abrupt changes in view importance weights can still lead to step changes in the layout. This invention applies an exponential moving average (EMA) smoothing method to the target layout size: ; in, α ema ∈(0, 1) is the EMA smoothing coefficient, preferably α ema = 0.3, which makes historical weight memory dominate (70%), while the current instantaneous signal only serves as an asymptotic adjustment factor (30%).

[0087] The smoothed importance weights are used for spatial allocation calculations, rather than the original instantaneous weights. This process ensures that even if there are sudden and drastic changes in the data, the layout adjustments are gradual and predictable.

[0088] The following section describes the non-linear allocation and smooth transition of screen space.

[0089] Specifically, based on the smoothed view importance weights, the target screen space quota for each view is calculated using a constrained nonlinear allocation algorithm, and the continuous visual evolution of the layout is achieved through temporal interpolation.

[0090] Specific execution process: 1. The following constraints need to be considered when allocating screen space: Minimum visibility lower bound constraint: Each view must maintain a minimum visible area A min (Preferably 8% of the total screen area) Even if it is of extremely low importance in the current task, it cannot be completely hidden or compressed to the point of being unrecognizable. This constraint ensures the overall accessibility of multi-view information and avoids the extreme case of a single view occupying the entire screen.

[0091] Risk safety margin constraint: If the alarm importance of view v I alert (v)> i alert Then the view will receive an additional risk protection allowance A. bonus (Preferably 5% of the total screen area) to ensure that the alarm view has enough space to display alarm details.

[0092] Total screen area conservation constraint: ; Where A(v) represents the screen display area allocated to the visualization view v; V represents the entire set of all current visualization views; A screen This indicates the total usable display area of ​​the screen.

[0093] 2. Nonlinear Allocation Algorithm This invention employs a nonlinear allocation strategy that enhances head concentration, enabling high-importance views to obtain a disproportionate amount of screen space while ensuring that low-importance views still meet the minimum visibility constraint.

[0094] First, a power-law transformation is performed on the importance of the smoothed layout: ; Here, γ>1 is the head concentration enhancement coefficient, with γ=1.5 being the preferred value. When γ>1, high importance values ​​are further amplified, while low importance values ​​are compressed, realizing the philosophy of making the important more important.

[0095] Then, the target area is solved under the constraints: ; Where |V| is the total number of views, |V alert| The number of views that triggered the alarm. If view v triggers an alarm, add an additional A to its target area. bonus .

[0096] 3. Smooth Transition Operator To avoid abrupt changes during layout adjustments, this invention introduces a time interpolation transition between adjacent frames. Let the target area of ​​the current frame be A. * (v), the current actual area is A current (v), the transient process is described by the following differential equation: ; in, k For the transition rate coefficient, preferably k = 2.0 / second, which allows the layout to complete 86% of the target distance in about 0.5 seconds (the 2x time constant rule for first-order systems).

[0097] At discrete time step t, the transition operator is implemented as follows: ; Where Δt is the frame time interval (usually taken as the display refresh cycle, approximately 16.7ms@60Hz).

[0098] During the transition, the view's borders and position move synchronously via interpolation, creating a smooth visual evolution. Throughout the transition, the visual encoding within the view remains unchanged (only the container size changes), ensuring that content rendering is unaffected by layout adjustments.

[0099] S107. Multi-view linkage rendering: Render and present a multi-view linkage interface according to the optimal mapping configuration and the target screen space quota; wherein, the views achieve linkage interaction through the global state manager, including brush linkage, cross-view synchronous highlighting and shared filter conditions.

[0100] The purpose of this step is to enable data linkage, selection synchronization, and sharing of filtering conditions among multiple views, based on optimized visual coding configuration and layout allocation.

[0101] Specific execution process: 1. Types of Interactive Linkages This invention supports the following three types of core interactive linkages: Brushing & Linking: When a user defines a selection area in a view using the selection box, lasso, or polygon tools, the selected collection of data objects S... selected Real-time synchronization to all other views. Other views are synchronized according to S. selectedAdjust the visual encoding, such as keeping the original encoding of selected objects or enhancing their highlight (e.g., increasing brightness or adding strokes), and degrading unselected objects to a weakened background state (e.g., reducing opacity to 30%).

[0102] The selected state is maintained through the Global State Manager, and cross-view synchronization is achieved using a publish-subscribe (Pub-Sub) model, with latency controlled within 50ms to ensure smooth interaction.

[0103] Cross-view Highlighting: When a user hovers their mouse over a data object, the corresponding representation of that object is simultaneously highlighted across all views. Because the same data object may be encoded using different visual channels in different views (for example, a risk event is a heatmap in a GIS view, a node on a polyline in a time-series view, and a graph node in a relationship view), the system uses a unified data object identifier (UUID) for cross-view object matching.

[0104] Highlighting styles follow these rules: Hovering object: Add a glowing halo effect with a radius of 6px and the color of the view's main color. Related object: Add a weak glow effect, radius 3px, color is neutral gray; Irrelevant objects: Reduce transparency to 40% and remove interactive responses.

[0105] Shared Filters: The system maintains the global filter condition status F, including the time range [t]. start , t end Geographic region R geo Risk level range [s] min , s max Event type set C types All views uniformly apply F as a pre-filter condition for data queries when rendering data.

[0106] When a user modifies filter criteria in any view (e.g., dragging the time selection range in the time sequence view), the modified criteria are automatically synchronized to the global state, and all other views apply the updated filter criteria in the next render frame. Changes to filter criteria do not directly trigger layout adjustments (to avoid frequent jitter during the filter exploration process); a one-time layout evaluation is only triggered when the filter criteria have been stable for more than 2 seconds.

[0107] 2. Visual coding consistency constraints In multi-view cascading, the visual encoding of the same data attribute in different views should maintain consistency in direction to avoid user confusion. High risk scores should be mapped to prominent visual features (high brightness, large area, high position, thick lines) in all views. A high level of warning should be mapped to warning visual characteristics (warm hue, high saturation, strong luminosity) in all views. The most recent time dimension should be arranged in a consistent direction in all views (usually from left to right or from bottom to top).

[0108] Encoding consistency constraints are implemented in the multi-view expansion phase through cross-view weak coupling constraints.

[0109] 3. Performance Optimization Strategies To ensure smooth interactive performance under large-scale data (tens of thousands of data points per screen), this invention adopts the following performance optimization strategies: Layered rendering: The view is divided into a background layer (static geographic base map / coordinate axis), a data layer (dynamic risk objects), and an interaction layer (selection / highlighting overlay). The background layer is only rendered once during initialization and cached as a bitmap. Incremental update: In linked state, only data objects whose visual attributes have changed are updated to avoid full redraw; Web Worker offloading: CPU-intensive operations such as data filtering and aggregation calculations are executed asynchronously in Web Worker threads, without blocking the main thread's rendering and interactive responses; LOD Adaptive: Automatically selects the rendering granularity (level of detail) based on the current view size and data density. High-density areas switch to aggregated thermal rendering, while low-density areas switch to individual point rendering.

[0110] In some embodiments, the above-described method of the present invention further includes the steps of tracing and presenting the risk evidence chain.

[0111] The purpose of this step is to support managers in continuously tracing back from macroscopic visualization results to microscopic original evidence, establishing a complete decision support chain from macroscopic situation to mesoscopic events to microscopic evidence.

[0112] Specific execution process: 1. Evidence Chain Data Model This invention defines the risk evidence chain as a quintuple E = (e1, e2, e3, e4, e5): e1 Macro-situation Layer: Risk clustering areas or abnormal points on the GIS heat map, presented by a spatial view optimized by visual coding; e2 Meso-level Event Layer: Risk event summary cards that are expanded after drilling down from macro-level points, including structured fields such as event number, occurrence time, geographical location, warning level, risk score, and event type; e3 Causal Explanation Layer: Based on the SHAP attribution method, the risk causal contribution decomposition diagram presents the core driving factors that make the event rated as high risk (three typical patterns: content intensity-driven, spatiotemporal context-sensitive, and historical frequency accumulation). e4 Propagation Inference Layer: Based on the risk level evolution prediction curve and social impact range estimate of the HGT-T time series heterogeneous graph Transformer, it presents the future trend prediction of the event; e5 Micro-evidence Layer: Original evidence materials related to the incident, including police report texts, social media screenshots, complaint texts, on-site photos / videos, and other multimodal materials.

[0113] 2. Trace the interaction path The system supports two directions of evidence chain navigation: Forward expansion: Starting with risk hotspots in the GIS macro view, users click or double-click on the target area → a pop-up event list sidebar → select a specific event → expand the cause explanation panel → view propagation projection predictions → browse related original evidence. Each step of the expansion is accompanied by smooth view transition animations and visual focus guidance, ensuring that the user's attention always follows the current exploration path.

[0114] Backtracking: When browsing micro-level evidence (such as a post from social media), users can use the "Tracing Source" button to instantly return to the macro-level context to which the evidence belongs. The system automatically restores the spatial location, timestamp, and risk level of the evidence to be highlighted in the global view, enabling rapid location from evidence to situation.

[0115] 3. Visual encoding of evidence presentation The multimodal material at the micro-evidence layer employs simplified coding that minimizes cognitive load: Textual evidence: Key sentences are automatically highlighted (based on a hybrid score of TF-IDF and sentiment intensity), avoiding the burden of reading the entire text; Image evidence: thumbnail grid + click to enlarge; thumbnails are marked with risk-colored borders (red / orange / yellow) to indicate the urgency of the evidence. Video evidence: keyframe preview + timeline risk intensity curve, supporting direct jump to the peak moment of risk intensity; Audio evidence: Speech-to-text is presented first, while the original audio is retained as a secondary option.

[0116] This invention combines the visual perception efficiency theory in cognitive psychology with the multidimensional data characteristics of social risks, and establishes a quantitative mathematical model between data attributes, visual channels, and cognitive load. Under the mutual exclusion constraint of channels, it solves the optimal visual encoding configuration with the minimum total cognitive load. At the same time, it introduces a task-driven view importance assessment mechanism, which dynamically allocates the screen space quota of each view according to the user's current assessment task, and suppresses layout jitter through stability control measures, so as to realize the visualization of multidimensional social risk information that is cognitively friendly, task-adaptive, and interactively stable.

[0117] To verify the beneficial effects of the present invention, a user comparison experiment was designed and implemented, comparing the traditional multi-view dashboard (control group) with the CogniRisk-Vis system of the present invention (experimental group).

[0118] Experimental Design: Forty-eight grassroots managers and command and dispatch personnel with experience in using social governance information systems were recruited and randomly divided into a control group (24 participants) and an experimental group (24 participants). There were no statistically significant differences between the two groups in age distribution (28-45 years), educational background (87.5% with a bachelor's degree or above), and system usage experience (average 3.2 years) (p>0.05). Three typical assessment tasks (situation recognition, trend analysis, and correlation mining) were designed, each containing three sub-scenarios, for a total of nine experimental tasks.

[0119] Measurement metrics: task completion time, task accuracy, NASA-TLX cognitive load scale, SUS system availability scale.

[0120] Experimental results: (1) Task performance results: The overall average completion time of the experimental group (200.1 seconds) was significantly shorter than that of the control group (264.2 seconds) by 24.3% (p<0.001); the overall average accuracy rate (85.5%) was 17.0 percentage points higher than that of the control group (68.5%) (p<0.001). The completion time of the experimental group in all three types of tasks was significantly lower than that of the control group, and the accuracy rate was significantly higher than that of the control group.

[0121] (2) NASA-TLX cognitive load results: The overall cognitive load weighted score of the experimental group (49.4) was significantly lower than that of the control group (66.8) by 26.0% (p<0.001). The scores decreased by 28.5% in the mental demand dimension, 29.7% in the time pressure dimension, and 37.1% in the frustration dimension.

[0122] (3) SUS usability score results: The experimental group's SUS score (81.6) reached the "excellent usability" level (>80), while the control group's score (62.3) was in the "near usable" range (<68). The difference between the two groups was highly statistically significant (p<0.001).

[0123] (4) Qualitative results of semi-structured interviews: 10 participants in the experimental group mentioned that "the interface does not suddenly jump around", reflecting the comprehensive effect of the layout stability control mechanism; 8 participants mentioned that "the meaning of color coding is consistent in different views, and there is no need to readjust", reflecting the effectiveness of coding consistency constraints.

[0124] In the above embodiments, a method for visualizing multidimensional risk information based on cognitive load optimization is provided. Correspondingly, this application also provides a system for visualizing multidimensional risk information based on cognitive load optimization. The system for visualizing multidimensional risk information based on cognitive load optimization provided in this application can implement the above-described method for visualizing multidimensional risk information based on cognitive load optimization. This system can be implemented through software, hardware, or a combination of both. For example, the system may include integrated or separate functional modules or units to perform the corresponding steps in the above methods.

[0125] Please refer to Figure 3 As shown, this application provides a multidimensional risk information visualization system 10 based on cognitive load optimization, comprising: The attribute classification module 101 is used to obtain a social risk multidimensional dataset, perform type parsing on each data attribute in the dataset, and classify it into one of the following: quantitative attribute, ordinal attribute, categorical attribute, relational attribute, hierarchical attribute, and composite multidimensional attribute; for composite multidimensional attributes, further perform sub-attribute decomposition to break them down into several basic type attributes. The channel mapping module 102 is used to rank the visual channel perception effectiveness based on a preset method, calculate the quantified perception efficiency coefficient for each visual channel using a decay model, and establish a mapping rule library from attribute type to candidate visual channel. The load optimization module 103 is used to calculate the information importance weight and dimensional complexity of each data attribute. The information importance weight is obtained by fusing three information sources: scene prior, data-driven, and user-defined. Based on the information importance weight, dimensional complexity, and perceptual efficiency coefficient, a cognitive load function is constructed. With the goal of minimizing the total cognitive load, the optimal mapping configuration from each data attribute to the visual channel is solved under the constraint of channel mutual exclusion. The task identification module 104 is used to identify the type of risk assessment task currently being performed by the user; based on three types of driving signals—data changes, interaction intent, and scenario prior—it dynamically calculates the layout importance weight and alarm importance weight of each view. The stability control module 105 is used to perform stability control processing on the layout importance weights. The stability control processing includes at least one of interactive locking, risk hysteresis, cooling constraint and smoothing processing. Based on the processed weights, under the constraints of minimum visibility lower limit and screen area conservation, a nonlinear allocation algorithm is used to dynamically calculate the target screen space quota of each view and realize the smooth transition of the layout. The rendering module 106 is used to render and present a multi-view interactive interface according to the optimal mapping configuration and the target screen space quota; wherein, the views are interconnected through a global state manager to achieve interactive linkage including brushing linkage, cross-view synchronous highlighting and shared filter conditions.

[0126] Specifically, the quantitative attributes include risk deviation score, estimated scale of propagation impact, or attribution contribution value; The ordinal attribute includes the warning level or the risk evolution trend level; The category attributes include event type or information source type; The relationship attributes include propagation links or subject interaction relationships; The hierarchical attributes include administrative division level or event classification level.

[0127] The multidimensional risk information visualization system based on cognitive load optimization provided in this application embodiment and the multidimensional risk information visualization method based on cognitive load optimization provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0128] This application also provides an electronic device corresponding to the method provided in the foregoing embodiments. The electronic device may be an electronic device for a client, such as a mobile phone, laptop, tablet computer, desktop computer, etc., to execute the above-described multidimensional risk information visualization method based on cognitive load optimization.

[0129] Please refer to Figure 4 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 4 As shown, the electronic device 20 includes: a processor 200, a memory 201, a bus 202, and a communication interface 203. The processor 200, the communication interface 203, and the memory 201 are connected via the bus 202. The memory 201 stores a computer program that can run on the processor 200. When the processor 200 runs the computer program, it executes the multidimensional risk information visualization method based on cognitive load optimization provided in any of the foregoing embodiments of this application.

[0130] The memory 201 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 203 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0131] Bus 202 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory 201 is used to store programs. After receiving an execution instruction, the processor 200 executes the program. The multi-dimensional risk information visualization method based on cognitive load optimization disclosed in any of the foregoing embodiments of this application can be applied to the processor 200, or implemented by the processor 200.

[0132] The processor 200 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 200 or by instructions in software form. The processor 200 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 201. The processor 200 reads the information in memory 201 and, in conjunction with its hardware, completes the steps of the above method.

[0133] The electronic device provided in this application embodiment and the multidimensional risk information visualization method based on cognitive load optimization provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0134] This application also provides a computer-readable storage medium corresponding to the cognitive load-optimized multidimensional risk information visualization method provided in the foregoing embodiments, wherein a computer program (i.e., a program product) is stored thereon. When the computer program is run by a processor, it executes the cognitive load-optimized multidimensional risk information visualization method provided in any of the foregoing embodiments.

[0135] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0136] The computer-readable storage medium provided in the above embodiments of this application and the multidimensional risk information visualization method based on cognitive load optimization provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and they should all be covered within the scope of the claims and specification of this application.

Claims

1. A method for visualizing multidimensional risk information based on cognitive load optimization, characterized in that, include: Obtain a social risk multidimensional dataset, perform type parsing on each data attribute in the dataset, and classify it into one of the following: quantitative attribute, ordinal attribute, categorical attribute, relational attribute, hierarchical attribute, and composite multidimensional attribute; for composite multidimensional attributes, further perform sub-attribute decomposition to break them down into several basic type attributes; Based on the pre-defined visual channel perception effectiveness ranking, a decay model is used to calculate the quantified perception efficiency coefficient for each visual channel, and a mapping rule library from attribute type to candidate visual channel is established. For each data attribute, its information importance weight and dimensional complexity are calculated; the information importance weight is obtained by fusing three information sources: scenario prior, data-driven, and user-defined. Based on the information importance weight, dimensional complexity, and perception efficiency coefficient, a cognitive load function is constructed. With the goal of minimizing the total cognitive load, the optimal mapping configuration from each data attribute to the visual channel is solved under the constraint of channel mutual exclusion. Identify the user's current risk assessment task type; Based on three types of driving signals—data changes, interaction intent, and scenario priors—the layout importance weight and alarm importance weight of each view are dynamically calculated. The layout importance weights are subjected to stability control processing, which includes at least one of interactive locking, risk hysteresis, cooling constraints and smoothing processing. Based on the processed weights, under the constraints of minimum visibility lower limit and screen area conservation, a nonlinear allocation algorithm is used to dynamically calculate the target screen space quota for each view and achieve a smooth transition of the layout. Based on the optimal mapping configuration and the target screen space quota, a multi-view interactive interface is rendered and presented; wherein, the views achieve interactive linkage, including scrolling linkage, cross-view synchronous highlighting, and shared filter conditions, through a global state manager.

2. The method according to claim 1, characterized in that, The quantitative attributes include risk deviation score, estimated scale of propagation impact, or attribution contribution value; The ordinal attribute includes the warning level or the risk evolution trend level; The category attributes include event type or information source type; The relationship attributes include propagation links or subject interaction relationships; The hierarchical attributes include administrative division level or event classification level.

3. The method according to claim 1, characterized in that, The formula for calculating the sensing efficiency coefficient is as follows: e channel = τ (rank-1) ； in, e channel The perceived efficiency coefficient. rank The ranking of the visual channel in the perceptual effectiveness ranking. τ It is a decay factor with a value range of (0,1); in the ranking of perception effectiveness, the position channel ranks highest and the shape channel ranks lowest.

4. The method according to claim 1, characterized in that, The formula for calculating the fusion of information importance weights is as follows: w imp (a i )=α · w scene (a i ) + β · w data (a i ) + γ · w user (a i )； in, w imp (a i ) For data attributes a i Information importance weight, w scene (a i ) , w data (a i ) and w user (a i ) These are, respectively, the importance of prior context, the importance of data-driven factors, and the importance of user-defined factors. α , β , γ The preset fusion weight coefficients, and satisfy the following conditions: α + β + γ = 1.

5. The method according to claim 1, characterized in that, The formula for calculating the dimensionality complexity is: d complexity (a i ) = base t × log 2 (|D(a i )| + 1 ) × ( 1 + η · struct i )； in, d complexity (a i ) For data attributes a i dimensional complexity, base t For attribute type t The corresponding baseline complexity, |D(a i )| The cardinality of the attribute's value range. struct i To determine the structural complexity index of the value range, η These are preset structural complexity weighting coefficients.

6. The method according to claim 1, characterized in that, The cognitive load function is defined as follows: CL(a i , v j ) = (w imp (a i ) ×d complexity (a i ) / e channel (v j )； in, CL(a i , v j ) Indicates data attributes a i Mapping to the visual channel v j Cognitive load during the process; e channel (v j ) For visual channel v j The perceived efficiency coefficient; w imp (a i ) For data attributes a i Information importance weight; d complexity (a i ) For data attributes a i dimensional complexity; The total cognitive load CL total (M) For visual configuration scheme M The sum of cognitive loads of all data attributes and visual channel mapping pairs.

7. The method according to claim 1, characterized in that, The channel mutual exclusion constraint includes at least one of the following: Single-channel exclusivity constraint means that the same visual channel can be assigned to at most one data attribute at the same view level; Attribute type and channel type matching constraint, that is, data attributes can only be assigned to the candidate channel set of their corresponding attribute type; Channel combination mutual exclusion constraint, that is, prohibits the simultaneous assignment of a preset mutually exclusive channel combination to different data attributes; And channel capacity limit constraint, that is, when the cardinality of the value range of data attributes exceeds the perceptual discrimination capacity limit of the visual channel, allocation is prohibited.

8. The method according to claim 1, characterized in that, The process of finding the optimal mapping configuration from each data attribute to the visual channel includes: A hybrid strategy combining branch and bound with greedy initialization is employed to solve for the optimal mapping configuration from each data attribute to the visual channel, where: In the greedy initialization phase, for each data attribute, the first candidate channel that satisfies all constraints is selected in ascending order of cognitive load, generating an initial feasible solution and serving as the upper bound; In the branch and bound optimization phase, the total cognitive load of the greedy solution is used as the upper bound, and the global optimal solution is searched through branching, bounding, and backtracking operations.

9. The method according to claim 1, characterized in that, The identification of the user's current risk assessment task type is achieved through a weighted fusion of the following three types of signals: Explicit task switching signals, triggered by user interaction with the task switching control, have the highest priority. View interaction behavior signals are implicitly inferred by monitoring the frequency, depth, and duration of user interactions on various views; And data-driven scene signals, which are judged by analyzing the spatial distribution entropy, time series variance or relation graph density characteristics of the current data subset.

10. The method according to claim 1, characterized in that, The layout importance weight I layout (v) is calculated using the following formula: I layout (v) = w1 · S scene (v, T current ) + w2 · S intent (v) + w3 · S data (v); Among them, S scene (v, T current S represents the prior signal of the scene. intent (v) represents the interaction intent signal, S data (v) represents the data change signal, and w1, w2, and w3 are preset fusion weights with w1+w2+w3=1 to ensure the dominant role of the scenario prior.

11. The method according to claim 1, characterized in that, The stability control process specifically includes: An interaction lock mechanism is implemented. When the frequency of user interaction operations exceeds a preset threshold, the system enters a locked state, preventing the triggering of a global layout reshuffle. The risk delay mechanism only triggers an update when the change in the layout importance calculated in this instance exceeds the preset delay ratio threshold of the previous value. The cooling constraint mechanism stipulates that the time interval between two adjacent global layout reorderings must not be less than the preset minimum update interval; And the exponential moving average smoothing process, the calculation formula is as follows: ； in, α ema This is the smoothing coefficient.

12. The method according to claim 1, characterized in that, The nonlinear allocation algorithm employs a power-law transformation, and its calculation formula is as follows: ； Among them, A * (v) represents the target screen space quota, A min For the minimum visible area, A screen V represents the total screen area, and |V| represents the total number of views. alert | The number of views that triggered the alarm, A bonus As a safety net for risk, layout (v) represents the layout importance after smoothing, and γ>1 is the head concentration enhancement coefficient.

13. The method according to claim 1, characterized in that, The shared filtering conditions are uniformly maintained by the global status manager, including time range, geographical region, risk level range, and event type set; When any view modifies the filter criteria, the updated criteria are automatically synchronized to all other views and take effect in the next rendering frame.

14. A multidimensional risk information visualization system based on cognitive load optimization, characterized in that, include: The attribute classification module is used to obtain a social risk multidimensional dataset, perform type parsing on each data attribute in the dataset, and classify it into one of the following: quantitative attribute, ordinal attribute, categorical attribute, relational attribute, hierarchical attribute, and composite multidimensional attribute; for composite multidimensional attributes, further perform sub-attribute decomposition to break them down into several basic type attributes. The channel mapping module is used to rank the visual channels based on the preset visual channel perception effectiveness, calculate the quantified perception efficiency coefficient for each visual channel using a decay model, and establish a mapping rule library from attribute type to candidate visual channels. The load optimization module is used to calculate the information importance weight and dimensional complexity of each data attribute. The information importance weight is obtained by fusing three information sources: scene prior, data-driven, and user-defined. Based on the information importance weight, dimensional complexity, and perceptual efficiency coefficient, a cognitive load function is constructed. With the goal of minimizing the total cognitive load, the optimal mapping configuration from each data attribute to the visual channel is solved under constraints including channel mutual exclusion. The task identification module is used to identify the type of risk assessment task currently being performed by the user. Based on three types of driving signals—data changes, interaction intent, and scenario priors—the layout importance weight and alarm importance weight of each view are dynamically calculated. A stability control module is used to perform stability control processing on the layout importance weights. The stability control processing includes at least one of interactive locking, risk hysteresis, cooling constraints, and smoothing processing. Based on the processed weights, under the constraints of minimum visibility lower limit and screen area conservation, a nonlinear allocation algorithm is used to dynamically calculate the target screen space quota for each view and achieve a smooth transition of the layout. The rendering module is used to render and present a multi-view interactive interface based on the optimal mapping configuration and the target screen space quota. The views interact with each other through a global state manager, including interactive filtering, cross-view synchronized highlighting, and shared filter conditions.

15. An electronic device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method as claimed in any one of claims 1 to 13.

16. A computer-readable storage medium, characterized in that, It stores computer-readable instructions that can be executed by a processor to implement the method as described in any one of claims 1 to 13.

Images