Big data-based information technology consulting management system and method

Abstract

The application relates to the technical field of computer information and machine learning, and particularly discloses an information technology consultation management system and method based on big data, which comprises a dynamic constraint perception module, a scheme knowledge graph module, a multi-target optimization engine module and a scheme self-adaptive evolution module. Through real-time perception of multi-dimensional dynamic constraints, semantic support is provided by using the knowledge graph, and a multi-target optimization algorithm is used to balance and optimize among targets such as cost, benefit and technical adaptability, so that the consultation scheme can intelligently select and evolve strategies according to constraint conflicts, the continuous self-adaptive evolution and optimization of the scheme in the whole life cycle are realized, and the agility and accuracy of the consultation service are improved.

Description

Technical Field

[0001] This invention belongs to the field of computer information and machine learning technology, specifically relating to an information technology consulting management system and method based on big data. Background Technology

[0002] Information technology consulting is a key service area that guides enterprises in digital transformation, optimizing business processes, and enhancing technological capabilities. With the popularization of big data technology, the consulting process is shifting from relying on expert subjective experience to data-driven decision-making, which involves the comprehensive analysis of massive amounts of industry data, technical solutions, and client constraints.

[0003] Big data-driven consulting management systems aim to generate customized technical solutions for clients through data modeling and algorithmic analysis. The fundamental goal of these systems is to integrate multi-source data and output consulting recommendations based on preset rules or models, thereby improving the scientific rigor and efficiency of the solutions.

[0004] Consulting solutions generated using static rule bases or historical case matching lack the ability to adaptively adjust and optimize based on dynamic constraints that arise during implementation. For example, when client budgets change, existing technical debt creates new limitations, or the team's talent structure shifts, the established static solutions struggle to evolve in real time to maintain their effectiveness and feasibility. This results in insufficient flexibility in the actual implementation of consulting solutions, leading to a disconnect from the dynamically changing business environment and technical conditions, ultimately impacting the successful implementation of consulting projects and the achievement of expected benefits. Summary of the Invention

[0005] The purpose of this invention is to provide an information technology consulting management system and method based on big data, so as to solve the problem that the existing technology lacks the ability to adaptively adjust and optimize according to dynamic constraints after the consulting solution is generated.

[0006] This invention provides an information technology consulting and management system based on big data. The system includes a dynamic constraint perception module, a solution knowledge graph module, a multi-objective optimization engine module, and a solution adaptive evolution module.

[0007] The dynamic constraint awareness module is used to collect and structure multi-dimensional dynamic constraint data from the customer, technology, and market sides in real time. This module specifically includes three sub-modules: a customer constraint awareness sub-module, a technology constraint awareness sub-module, and a market constraint awareness sub-module. The customer constraint awareness sub-module interfaces with enterprise resource planning (ERP) systems, project management tools, and financial systems to acquire and analyze customer budget changes, project progress data, team skill matrix changes, and business priority adjustment data in real time. The technology constraint awareness sub-module scans code repositories, infrastructure monitoring platforms, and technology debt management tools to acquire and analyze system architecture complexity data, legacy system interface data, technology stack compatibility data, and security compliance requirement changes in real time. The market constraint awareness sub-module crawls industry technical reports, open-source community updates, and supplier product update announcements to acquire and analyze new technology maturity data, mainstream technology trend data, third-party service cost changes, and supply chain risk data in real time. The dynamic constraint awareness module normalizes and timestamps all the above data to form a standardized dynamic constraint vector, which is then pushed to the solution adaptive evolution module in real time.

[0008] The solution knowledge graph module is used to construct and maintain a structured knowledge network representing entities and relationships in the field of information technology consulting. The nodes of this knowledge graph include technical component entities, business scenario entities, solution pattern entities, constraint type entities, and implementation effect indicator entities. Technical component entities include specific technical elements and their attributes, such as microservice frameworks, database systems, and message middleware. Business scenario entities include typical business requirement scenarios such as high-concurrency transactions, real-time data analysis, and hybrid cloud deployment. Solution pattern entities are composite nodes connecting technical components and business scenarios, representing verified and reusable technical architecture patterns. Constraint type entities correspond to the constraint vector categories output by the dynamic constraint awareness module. Implementation effect indicator entities include quantitative indicators such as performance improvement percentage, cost savings, implementation cycle, and system availability. Edges between nodes define various relationships between entities, including compatibility relationships between technical components, the fulfillment of business scenarios by solution patterns, the restriction relationships of constraint types on solution patterns, and the causal relationships between solution patterns and implementation effect indicators. This knowledge graph is incrementally trained and updated by continuously importing historical consulting case data, industry best practice documents, and technical white papers.

[0009] The multi-objective optimization engine module is used to perform optimization calculations on candidate consulting solutions under given dynamic constraints to generate a Pareto optimal solution set. This module receives the current solution state code and the latest dynamic constraint vector from the solution adaptive evolution module. Internally, it constructs a mathematical model containing at least three optimization objectives: total implementation cost, expected comprehensive benefit index, and coupling degree between the solution and technical debt. The total implementation cost is calculated by summing the procurement costs, deployment costs, and labor costs of all technical components in the solution. The expected comprehensive benefit index is calculated by weighted summation of the implementation effect indicators associated with the solution knowledge graph module, with weights determined by the client's business priorities. The coupling degree between the solution and technical debt is calculated by analyzing the interface complexity and dependency strength between the technical components required by the solution and legacy components in the client's existing technology stack.

[0010] The multi-objective optimization engine employs a decomposition-based multi-objective evolutionary algorithm for solving the problem. This algorithm decomposes the multi-objective problem into a series of scalar sub-problems and performs co-evolution. Specifically, the process involves: first, retrieving and generating an initial population of candidate solutions from the solution knowledge graph based on the current solution state and constraint vectors; then, evaluating each individual solution in the population against the three objectives mentioned above; next, generating a population of offspring solutions through simulated binary crossover and polynomial mutation operations; finally, using a Chebyshev aggregation function to assign each sub-problem to individuals in the population for optimization; and after a preset number of iterations, the algorithm outputs a non-dominated solution set, i.e., the Pareto front, where each solution represents a feasible optimization solution that achieves different trade-offs among multiple objectives.

[0011] The adaptive evolution module, as the core control unit of the system, is responsible for driving the continuous intelligent evolution of the consulting solution throughout its lifecycle. This module includes a solution state encoder, a constraint conflict detector, an evolution strategy controller, and a solution version manager. The solution state encoder maps the current consulting solution into a high-dimensional feature vector, which contains the set of technical components used in the solution, the architectural pattern, resource configuration parameters, and historical implementation stage markers. The constraint conflict detector continuously compares the solution state code with the latest dynamic constraint vector pushed by the dynamic constraint perception module. When any constraint condition is detected to be violated or approaching a critical threshold, the evolution process is immediately triggered.

[0012] The evolution strategy controller selects one of three preset evolution strategies based on the type and severity of constraint conflicts. The first strategy is a local parameter tuning strategy, suitable for resource configuration parameter-related constraint conflicts. The controller's multi-objective optimization engine optimizes only a small range of resource configuration parameters while maintaining the main structure of the solution. The second strategy is a component replacement strategy, suitable for technical compatibility or cost-related constraint conflicts. Guided by the solution knowledge graph, the controller's multi-objective optimization engine searches for and evaluates compatible replacement technical components. The third strategy is an architectural pattern refactoring strategy, suitable for fundamental changes in business requirements or significant technical debt constraints. Based on the new core constraints, the controller's multi-objective optimization engine retrieves matching new architectural patterns from the solution knowledge graph and initiates a new round of multi-objective optimization to generate a completely new solution. The solution version manager records the reason for each evolution trigger, the strategy adopted, the input constraint vector, the output Pareto solution set, and the final selected evolution scheme, forming a complete solution evolution chain. This chain data is simultaneously fed back to the solution knowledge graph module for knowledge updates.

[0013] Furthermore, the calculation process of the expected comprehensive benefit index of the solution is as follows: First, extract all implementation effect indicator nodes associated with the current solution from the solution knowledge graph; then, based on the business priority data provided by the customer constraint perception submodule, assign weight coefficients to each implementation effect indicator, with the sum of all weight coefficients being 1; next, based on the regression model trained on historical case data, predict the specific improvement value of the current solution for each implementation effect indicator; finally, multiply the predicted improvement value of each indicator by its corresponding weight coefficient and sum them to obtain the expected comprehensive benefit index of the solution.

[0014] Furthermore, the coupling degree between the proposed solution and technical debt is calculated using a dependency graph analysis method. This method first constructs a component dependency graph of the client's existing technology stack; then, new technology components introduced by the candidate consulting solution are added to this graph as new nodes; next, the sum of the shortest path length between the new node and existing legacy technology component nodes in the graph, and the weights of the dependency edges on the path, is calculated; finally, the coupling degree is defined as the weighted average length of all such paths, with the weights determined by the modification costs of the components involved in the path.

[0015] Furthermore, the evolutionary strategy controller selects a strategy based on a preset decision rule table. This decision rule table is organized by constraint conflict type (rows) and conflict severity level (columns), with each cell defining the preferred evolutionary strategy number. The conflict severity level is divided into three levels based on the percentage deviation of the corresponding dimension's value in the constraint vector from its initial baseline value.

[0016] Furthermore, when recording the evolutionary path, the solution version manager records the coordinates of the final solution selected by the client or system on the Pareto front for the Pareto solution set output by the multi-objective optimization engine. This coordinate information is used for subsequent analysis of the client's preference patterns and optimization of the initialization settings of the weight vector in the multi-objective optimization algorithm.

[0017] Furthermore, the system operates within a closed-loop feedback architecture. After the final evolution plan output by the adaptive evolution module is confirmed by the customer and enters the implementation phase, the actual cost data, achieved results data, and new problem data generated during the implementation process will be fed back to the dynamic constraint perception module and the solution knowledge graph module through the data acquisition channel, serving as new dynamic constraints and knowledge evidence to drive the next round of solution optimization and knowledge base enhancement.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention fundamentally changes the static and rigid generation and maintenance mode of information technology consulting solutions by constructing a collaborative system of dynamic constraint perception, solution knowledge graph, multi-objective optimization, and adaptive evolution modules. The dynamic constraint perception module realizes real-time capture and quantification of diverse and time-varying constraints such as customer budgets, technical debt, and market trends, providing accurate input signals for solution evolution.

[0019] The solution knowledge graph module structures and links scattered domain knowledge, providing a rich and reasonable semantic foundation for solution generation and optimization.

[0020] The multi-objective optimization engine module can automatically weigh and optimize multiple conflicting objectives such as cost, benefit, and technology adaptability, and output a series of Pareto optimal candidate solutions, rather than a single solution.

[0021] As an intelligent hub, the adaptive evolution module can automatically trigger and guide the most suitable evolution strategy based on constraint conflicts, enabling the consulting solution to continuously adapt to environmental changes like a living organism.

[0022] The entire system forms a complete closed loop from perception, decision-making, execution to feedback, ensuring that the consulting solution remains dynamically optimal and practically feasible throughout the entire project lifecycle, significantly improving the agility, accuracy, and long-term value of consulting services. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention; Figure 2 This is a schematic diagram of the core principle framework of the adaptive evolution module in this invention; Figure 3This is a schematic diagram of the multi-level interaction relationship and data flow of the dynamic constraint perception module in this invention; Figure 4 This is a logical flow diagram of the multi-objective optimization engine module in this invention; Figure 5 This is a schematic diagram of the structured knowledge network framework of the knowledge graph module in this invention. Detailed Implementation

[0024] Example 1: The overall architecture of the information technology consulting and management system based on big data proposed in this invention is shown in the attached figure. Figure 1 As shown, the system consists of four core functional units: a dynamic constraint perception module, a solution knowledge graph module, a multi-objective optimization engine module, and a solution adaptive evolution module. These modules are tightly coupled through standardized data interfaces and event-driven mechanisms, forming a closed-loop feedback and continuously evolving intelligent decision-making system. The system runs on a distributed computing platform, supporting high-concurrency data access, real-time inference, and large-scale optimization computation. The following will combine the attached diagram... Figure 1 To be continued Figure 5 The specific implementation methods of each module are explained in detail.

[0025] First, the dynamic constraint perception module, as the external environment perception front-end of the system, has the following structure and data flow relationship: Figure 3 As shown, this module is divided into three logically independent but data-coordinated sub-modules: the customer constraint awareness sub-module, the technology constraint awareness sub-module, and the market constraint awareness sub-module. The customer constraint awareness sub-module establishes bidirectional communication connections with the customer's enterprise resource planning system, project management tools, and financial system through a secure, authenticated application programming interface (API). Within each polling cycle (default 5 minutes, configurable from 1 to 60 minutes), this sub-module actively pulls or passively receives structured data streams from the aforementioned systems.

[0026] Specifically, the Enterprise Resource Planning (ERP) system provides records of budget balance changes and cost center allocation changes; project management tools output current task completion rates, critical path delay warnings, and human resource utilization status; and the financial system pushes adjustments to capital expenditure limits and revised cash flow forecasts. Furthermore, customer business managers can manually input business priority weight adjustment commands through a dedicated interface. These commands are encapsulated in a structured message format and then enter the data pipeline. All raw data undergoes missing value filling, outlier removal, and unit standardization processing by the data cleaning unit built into the customer constraint awareness submodule, ultimately mapping to a customer-dimensional constraint feature vector. This vector has a fixed 8-dimensionality, including budget flexibility coefficients, manpower availability index, project schedule deviation rate, skill matching degree, business priority vector (4 dimensions), and cost sensitivity threshold.

[0027] The Technology Constraint Awareness submodule performs deep scanning and monitoring of the customer's technology infrastructure layer. This submodule deploys a lightweight agent that periodically (every 10 minutes by default) accesses the customer's code repository, infrastructure monitoring platform, and technology debt management tool. From the code repository, the submodule parses source code dependencies, branching strategies, and commit frequencies, extracting system architecture complexity metrics (using the McCabe cyclomatic complexity weighted average) and interface change frequency. From the infrastructure monitoring platform, it obtains runtime performance data such as CPU utilization, memory leak trends, and network latency distribution, and uses this to infer the stability boundaries of the existing technology stack. From the technology debt management tool, it extracts the number of identified technology debts, severity level distribution, and estimated remediation costs. Simultaneously, this submodule maintains a customer-specific technology stack list, recording the version number, license type, and lifecycle status of all components in use. When a component is detected to be nearing discontinuation or has a high-risk vulnerability, a technology compliance risk signal is immediately generated.

[0028] After normalization, all the above information forms a constraint feature vector of the technical dimension, which has 7 dimensions, including the architecture complexity index, legacy system interface compatibility score, technology stack health, security compliance risk level, number of days of lag in dependent component updates, technical debt density, and estimated cost of migrating between old and new technology stacks.

[0029] The Market Constraint Awareness submodule is responsible for capturing dynamic changes in the macro-technology ecosystem from the open network environment. This submodule integrates a distributed web crawler cluster and continuously crawls industry technology reports (sources include Gartner, IDC, and Forrester), open source community dynamics, and product update announcements from mainstream vendors according to a preset keyword strategy (such as cloud-native, AI inference framework, and database as a service).

[0030] The original webpage content undergoes a natural language processing pipeline: First, key technical terms and vendor names are extracted using named entity recognition; second, a sentiment analysis model is used to determine the technology maturity trend (e.g., early adoption, mainstream adoption, decline); then, a price extractor identifies pricing changes in third-party services (e.g., API call unit price, storage fees, and hourly rates for computational instances); finally, a supply chain risk assessor calculates the probability of continuity in the supply of key technical services based on geopolitical news and logistical disruption events. All processing results are aggregated into a market-dimensional constraint feature vector, with six dimensions, including a new technology maturity index (range 0 to 1), a technology trend consistency score, third-party service cost volatility, open-source activity index, vendor lock-in risk coefficient, and supply chain disruption probability.

[0031] The central coordinator of the dynamic constraint perception module receives the feature vectors output by the three sub-modules mentioned above, performs timestamp alignment (using a system clock synchronization protocol with nanosecond-level precision), zero-padding for missing dimensions (filling with historical averages), and Z-score standardization, finally concatenating them into a 21-dimensional standardized dynamic constraint vector. This vector is pushed to the scheme adaptive evolution module in real time via a message queue, with a push frequency of no less than once every 15 minutes, and is triggered immediately when any sub-module detects a major constraint mutation (such as a budget reduction of more than 20% or the disclosure of vulnerabilities in core components).

[0032] Secondly, the structured knowledge network framework of the solution's knowledge graph module is attached. Figure 5 As shown, the core of this module is a heterogeneous knowledge graph built on a graph database. The node types in the graph are strictly limited to five categories: technical component entities, business scenario entities, solution pattern entities, constraint type entities, and implementation effect indicator entities. Technical component entity nodes include attribute fields: component name, technology domain (e.g., message middleware, relational database), version range, open-source / commercial attribute, typical deployment scale, list of compatible operating systems, and average learning curve days. For example, Apache Kafka, as a technical component entity, has the following attributes: technology domain is message middleware, version range is 2.8 to 3.5, open-source attribute is true, typical deployment scale is a cluster of 10 or more nodes, compatible operating systems are Linux and macOS, and the average learning curve is 14 days.

[0033] Business scenario entity nodes describe typical customer business requirement contexts, and their attributes include: scenario name, concurrency level (low / medium / high / extremely high), data timeliness requirements (batch processing / near real-time / real-time), data scale, compliance requirements, and disaster recovery level.

[0034] Solution pattern entities are composite hub nodes in the graph, representing proven, reusable technical architecture paradigms. Each such node is associated with multiple technical component entities (connected via relational edges) and at least one business scenario entity (connected via adaptation relational edges). Its attributes include: pattern name, applicable industry (e.g., finance, e-commerce, manufacturing), number of reference cases, average implementation time (person-months), and technical complexity rating.

[0035] Each constraint type entity node corresponds one-to-one with each dimension of the 21-dimensional vector output by the dynamic constraint perception module. For example, the budget flexibility coefficient constraint type node defines its legal value range (0.0 to 1.0), violation threshold (<0.3 is considered a serious conflict), and set of mitigable solution patterns (via reverse links constrained by relational edges). The implementation effect indicator entity node quantifies the value of the solution, and its attributes include: indicator name, unit (e.g., percentage, ten thousand yuan, day), baseline value (industry average level), improvement limit (theoretical maximum value), and measurement method.

[0036] The edges between nodes define seven semantic relationships: compatibility between technical components (with compatibility scores from 0 to 1), satisfaction of business scenarios by solution patterns (with satisfaction weights), restriction of solution patterns by constraint types (with restriction strength), improvement of implementation effect indicators by solution patterns (with improvement coefficients), influence of technical components on implementation effect indicators, sensitivity of business scenarios to constraint types, and derivative relationships between solution patterns (representing pattern evolution paths). The graph is updated daily by incrementally importing historical consulting cases (structured into CSV files, containing fields such as solution ID, components used, achieved results, and encountered problems), industry best practice documents (converted into triples after information extraction), and technical white papers (extracting technical propositions and applicable conditions through semantic parsing). The graph database's built-in graph neural network embedding model performs embedding vectorization on the entire graph weekly, generating a 128-dimensional feature vector for each node, used for subsequent similarity retrieval and inference.

[0037] Third, the logical flow of the multi-objective optimization engine module is shown in the appendix. Figure 4 As shown. The core task of this module is to solve an optimization problem with three conflicting objectives, given a dynamic constraint vector and the current scheme state encoding. The three objective functions are defined as follows: The first objective is the total implementation cost of the solution. The calculation formula is as follows:

[0038] in, This represents the total number of technical components included in the solution. This represents the purchase or licensing cost of the i-th component (this is 0 for open-source components). The deployment and configuration cost of this component is estimated based on the component's complexity and the difficulty of adapting it to the customer's environment. The required labor cost is equal to the estimated number of man-days multiplied by the unit labor cost rate. All cost items are in RMB ten thousand and are dynamically adjusted based on the latest price index provided by the market constraint perception submodule.

[0039] The second objective is the expected comprehensive benefit index of the plan. The calculation process is as follows: First, all implementation effect indicator nodes activated by the current solution are retrieved from the solution knowledge graph; then, based on the business priority vector provided by the customer constraint perception submodule... (in ), for each indicator Assign weights Next, a pre-trained gradient boosting regression model (based on the XGBoost algorithm, trained using 1000 historical cases) is invoked, and the current solution's technical component combination and customer environment characteristics are input to predict the solution's performance on each metric. Specific improvement value Ultimately, the benefit index is a weighted sum:

[0040] The third objective is the coupling between the solution and technical debt. This metric is calculated using dependency graph analysis. First, the component dependency graph of the customer's existing technology stack is obtained from the technology constraint awareness submodule. , where the node set For legacy components, edge sets For calls or data dependencies between components, each edge With weight This indicates a modification to the estimated cost of the dependency. Then, the set of new technology components from the candidate solutions will be considered. As a new node, it is added to the knowledge graph, and potential dependency edges between the new and old components are added based on the compatibility relationships in the knowledge graph. For each new node... Calculate its path to all legacy nodes. Shortest path length (Path length is defined as the sum of edge weights). Coupling is defined as all... The weighted average path length of the pair, with weights of . It itself, that is:

[0041] The smaller the value, the smoother the integration of the new solution with the old system.

[0042] The multi-objective optimization engine employs a decomposition-based multi-objective evolutionary algorithm (MOEA / D) for solving the problem. During algorithm initialization, a restricted random walk is performed in the solution knowledge graph based on the current solution state encoding to generate an initial population (size 100). Each individual solution is encoded as a variable-length integer sequence representing the selected technical component ID and its configuration parameters. In each iteration, the three objective values ​​are evaluated in parallel for each individual in the population. Subsequently, simulated binary crossover (SBX) operator (crossover distribution exponent of 20) and multinomial mutation (mutation distribution exponent of 20, mutation probability of 0.1) are used to generate offspring. The algorithm maintains a uniformly distributed set of 100 weight vectors, each corresponding to a scalar quantum problem. The multi-objective values ​​are aggregated into a single fitness value using a Chebyshev aggregation function. After 200 iterations, the algorithm converges and outputs a non-dominated solution set, i.e., the Pareto front. Each solution on this front represents a feasible solution that achieves different balances among cost, benefit, and coupling.

[0043] Fourth, the adaptive evolution module, as the core control center of the system, has the following principle framework: Figure 2 As shown, this module comprises four key sub-units: a scheme state encoder, a constraint conflict detector, an evolution strategy controller, and a scheme version manager.

[0044] The solution state encoder is responsible for compressing the complete state of the current consulting solution into a 512-dimensional floating-point feature vector. The encoding process consists of four steps: First, the list of technical component IDs used in the solution is mapped into a vector sequence through a pre-trained component embedding matrix (128 dimensions); second, the architecture pattern IDs are expanded into 32-dimensional vectors through one-hot encoding; third, resource configuration parameters (such as the number of servers, memory size, and number of replicas) are normalized and concatenated into a 64-dimensional vector; finally, historical implementation stage markers (such as proof of concept, detailed design, and go-live preparation) are encoded into 8-dimensional classification vectors. All the above vectors are fused through a multilayer perceptron (MLP) to output the final 512-dimensional state code. This code is persistently stored and used as the initial search point for the multi-objective optimization engine.

[0045] The constraint conflict detector continuously monitors the constraint vector stream from the dynamic constraint awareness module. It maintains a conflict rule base, where each rule defines a constraint dimension, its legal range, and a critical threshold. For example, the legal range for the budget flexibility coefficient is [0.3, 1.0], and the critical threshold is 0.25. When any dimension value in a newly arrived constraint vector exceeds the legal range or falls below the critical threshold, the detector immediately generates a conflict event object, including the conflict type, current value, baseline value, percentage deviation, and severity level (1 to 3). The severity level classification rules are: deviation <15% is Level 1 (minor), 15% ≤ deviation <30% is Level 2 (moderate), and deviation ≥30% is Level 3 (severe).

[0046] Upon receiving a conflict event, the evolutionary strategy controller queries a pre-defined decision rule table. This table is a 3x3 matrix, with row indices representing conflict type (customer, technology, market) and column indices representing severity level (1, 2, 3). Each cell specifies a priority strategy number (1, 2, or 3). Strategy 1 involves local parameter tuning, adjusting only resource configuration parameters (e.g., increasing the number of servers to address performance degradation); Strategy 2 involves component replacement, searching the knowledge graph for functionally equivalent but lower-cost or more compatible alternative components; Strategy 3 involves architectural pattern restructuring, abandoning the current pattern and reselecting an infrastructure paradigm based on new core constraints. Based on the selected strategy, the controller sends corresponding optimization instructions to the multi-objective optimization engine, including specifying which variables to fix, which component spaces to search, and the weight bias of the objective function.

[0047] The solution version manager is responsible for full lifecycle tracking. Each time an evolution is triggered, it creates a new solution version record, containing: the parent version ID, trigger timestamp, conflict event details, the evolutionary strategy adopted, the input constraint vector, the complete Pareto solution set output by the multi-objective optimization engine (including three objective values ​​for each solution), and the finally selected sub-solution ID. Specifically, for the Pareto solution set, the manager records the coordinates of the client-selected solution in the objective space. This coordinate is used for cluster analysis of customer preferences. All version records form an immutable evolutionary chain, stored in a blockchain-style log. This chain data is exported in batches daily to feed back into the solution knowledge graph module—new solution-effect relationships are extracted into new causal edges, and failed evolutionary attempts are marked as negative samples for optimizing the prediction model.

[0048] The entire system operates within a closed-loop feedback architecture. Once the final evolved solution output by the adaptive evolution module receives electronic signature confirmation from the client, the implementation phase begins. During implementation, lightweight probes deployed in the client's environment collect real-time feedback data, including actual cost consumption, performance achievements, and user satisfaction scores. This data, after anonymization and aggregation, flows back to the dynamic constraint perception module (as new evidence of client constraints) and the solution knowledge graph module (as new instances of implementation effectiveness metrics), driving the next round of more precise solution optimization. This closed loop ensures the continuous evolution of system knowledge and the increasing accuracy of solution recommendations.

[0049] Example 2: Based on Example 1 above, this example further refines the interactive filtering mechanism of Pareto solution set in the multi-objective optimization engine module and enhances the support capability of the scheme adaptive evolution module for multi-customer collaborative scenarios.

[0050] After outputting the Pareto front, the multi-objective optimization engine does not immediately deliver all solutions, but instead initiates an interactive, visual filtering process. The system projects the Pareto solution set onto a 3D objective space and renders it as an interactive scatter cloud using WebGL. Client decision-makers can dynamically adjust the relative importance weights of the three objectives by dragging sliders, and the system highlights the 1 to 3 candidate solutions that are optimal under the current weights in real time. Simultaneously, each solution node displays thumbnails of its key technical components, a bar chart of the expected implementation cycle, and a radar chart showing compatibility with the client's existing technology stack. Clients can click on any solution to expand its complete contextual view in the solution knowledge graph, including the applicable business scenarios, related historical success stories, and potential technical risk warnings. All clickstream and dwell time data generated during this interactive process are recorded and used to learn the client's implicit preference model online. This model is used to initialize the weight vector distribution of MOEA / D in the next optimization, making it more focused on areas that the client may be interested in.

[0051] In multi-customer collaborative scenarios, the solution version manager of the solution adaptive evolution module has been extended to support cross-customer knowledge migration. When multiple customers face similar constraint conflicts (such as simultaneously encountering a significant price increase from a cloud service provider), the system automatically identifies these customer groups and injects collective intelligence prompts into their respective evolution paths. For example, if customer A successfully adopts a component replacement strategy and chooses an open-source alternative when facing a surge in third-party service costs, and the implementation is effective, this experience will be anonymized and pushed as a recommendation strategy to customers B and C who are experiencing the same conflict. The knowledge migration is triggered based on the cosine similarity of the constraint vectors (threshold set at 0.85) and the Jaccard similarity of the business scenario (threshold set at 0.7). The migration content includes not only the final selected solution but also the complete evolution path, encountered pitfalls, and avoidance measures. This mechanism significantly accelerates the solution optimization process for new customers and improves the robustness of the overall solution.

[0052] Furthermore, the market constraint perception submodule of the dynamic constraint perception module introduces an adversarial verification mechanism. To prevent web crawlers from obtaining false or misleading market information, this submodule establishes a credibility score for each data source. The score is based on historical accuracy (the degree to which the source's past predictions match actual market trends), content consistency (the degree of overlap with other highly credible sources), and publisher authority (whether it is a well-known research institution or a leading manufacturer). When the credibility score of a data source is below 0.6, the information it provides is subject to a decay factor (0.3 times the weight) before normalization. This mechanism effectively filters out market noise and ensures the authenticity of constraint inputs.

[0053] Through the above enhancements, while maintaining the core architecture, the system has significantly improved human-machine collaboration efficiency and cross-customer knowledge reuse capabilities, further strengthening its adaptability and value creation capabilities in complex and ever-changing information technology consulting environments.

Claims

1. An information technology consulting management system based on big data, characterized in that, include: The dynamic constraint perception module is used to collect and structure multi-dimensional dynamic constraint data from the customer side, the technology side, and the market side in real time to form a standardized dynamic constraint vector. The solution knowledge graph module is used to construct and maintain a structured knowledge network representing entities and relationships in the field of information technology consulting. The nodes of the knowledge network include technical component entities, business scenario entities, solution model entities, constraint type entities, and implementation effect indicator entities. The multi-objective optimization engine module is used to receive the dynamic constraint vector and the current scheme state code, and perform optimization calculations on the candidate consultation schemes based on the mathematical model containing the optimization objectives to generate a Pareto optimal solution set. The optimization objectives include the total implementation cost of the scheme, the expected comprehensive benefit index of the scheme, and the coupling degree between the scheme and technical debt. The adaptive evolution module is used to drive the continuous intelligent evolution of the consulting solution throughout its entire life cycle. The adaptive evolution module includes a solution state encoder, a constraint conflict detector, an evolution strategy controller, and a solution version manager. The scheme state encoder is used to map the current consultation scheme into a high-dimensional feature vector; The constraint conflict detector is used to continuously compare the high-dimensional feature vector output by the scheme state encoder with the latest dynamic constraint vector pushed by the dynamic constraint perception module, and triggers the evolution process when the constraint condition is detected to be violated or approaching the critical threshold. The evolution strategy controller is used to select one of the preset evolution strategies to execute based on the conflict type and severity level detected by the constraint conflict detector, so as to instruct the multi-objective optimization engine module to perform the corresponding type of optimization calculation. The scheme version manager is used to record the complete link data triggered by each evolution, and feeds the link data back to the scheme knowledge graph module for knowledge updates.

2. The information technology consulting management system based on big data according to claim 1, characterized in that, The dynamic constraint perception module specifically includes a customer constraint perception submodule, a technology constraint perception submodule, and a market constraint perception submodule. The customer constraint perception submodule is used to obtain and parse customer budget change data, project progress data, team skill matrix change data, and business priority adjustment data in real time by connecting with the application programming interface of the enterprise resource planning system, project management tools, and financial system. The technology constraint awareness submodule is used to acquire and parse system architecture complexity data, legacy system interface data, technology stack compatibility data, and security compliance requirement change data in real time by scanning code repositories, infrastructure monitoring platforms, and technology debt management tools. The market constraint perception submodule is used to crawl industry technical reports, open source community dynamics and supplier product update announcements to obtain and parse new technology maturity data, mainstream technology trend data, third-party service cost change data and supply chain risk data in real time.

3. The information technology consulting management system based on big data according to claim 1, characterized in that, The calculation process for the expected comprehensive benefit index of the proposed scheme is as follows: First, extract all implementation effect indicator nodes associated with the current solution from the solution knowledge graph module; Then, a weight coefficient is assigned to each implementation effect indicator based on the customer's business priority data, and the sum of all weight coefficients is 1; Next, the regression model trained based on historical case data predicts the specific improvement of the current solution on each implementation effect indicator; Finally, the predicted improvement value of each indicator is multiplied by its corresponding weight coefficient and then summed to obtain the expected comprehensive benefit index of the scheme.

4. The information technology consulting management system based on big data according to claim 1, characterized in that, The coupling degree between the proposed solution and technical debt is calculated using dependency graph analysis. The calculation process of this method is as follows: First, construct a component dependency graph of the client's existing technology stack; then, add new technology components introduced by the candidate consulting solutions as new nodes to this graph. Next, calculate the sum of the shortest path length between the newly added node and the existing legacy technology component nodes in the graph, and the weights of the dependent edges on the path; Ultimately, coupling is defined as the weighted average length of all such paths, with the weights determined by the modification costs of the components involved in the path.

5. The information technology consulting management system based on big data according to claim 1, characterized in that, The multi-objective optimization engine module uses a decomposition-based multi-objective evolutionary algorithm for solving the problem. The execution process of this algorithm is as follows: First, based on the current scheme state and constraint vector, a set of initial candidate scheme populations is retrieved and generated in the scheme knowledge graph module. Then, the at least three optimization objectives are evaluated for each individual scheme in the population; next, a population of offspring schemes is generated by simulating binary crossover and polynomial mutation operations. Then, the Chebyshev aggregation function is used to assign each subproblem to individuals in the population for optimization; After a preset number of iterations, the algorithm outputs the non-dominated solution set as the Pareto optimal solution set.

6. The information technology consulting management system based on big data according to claim 1, characterized in that, The evolution strategy controller selects an evolution strategy based on a preset decision rule table. The decision rule table is arranged with constraint conflict type as the row and conflict severity level as the column. Each cell defines the evolution strategy number that should be given priority. The severity level of the conflict is determined based on the percentage by which the value of the corresponding dimension in the constraint vector deviates from its initial baseline value.

7. The information technology consulting management system based on big data according to claim 6, characterized in that, The preset evolution strategies include local parameter tuning strategies, component replacement strategies, and architecture pattern reconstruction strategies. The local parameter tuning strategy instructs the multi-objective optimization engine module to perform optimization on a small scale only for resource configuration parameters while keeping the main structure of the scheme unchanged. The component replacement strategy instructs the multi-objective optimization engine module, under the guidance of the scheme knowledge graph module, to find replaceable compatible technology components and evaluate and optimize them. The architecture pattern refactoring strategy instructs the multi-objective optimization engine module to retrieve a matching new architecture pattern in the solution knowledge graph module based on the new core constraints, and to initiate a new round of multi-objective optimization to generate a completely new solution entity.

8. The information technology consulting management system based on big data according to claim 1, characterized in that, When recording the evolution path, the solution version manager records the position coordinates of the solution finally selected by the customer or system on the Pareto front for the Pareto solution set output by the multi-objective optimization engine module. The position coordinates are used for subsequent analysis of the customer's preference patterns and optimization of the initialization settings of the weight vector in the multi-objective optimization engine module.

9. The information technology consulting management system based on big data according to claim 1, characterized in that, The system operates in a closed-loop feedback architecture. After the final evolution scheme output by the scheme adaptive evolution module enters the implementation stage, the real cost data, achievement data, and new problem data generated during the implementation process will be fed back to the dynamic constraint perception module and the scheme knowledge graph module through the data acquisition channel as new dynamic constraints and knowledge evidence.

10. A big data-based information technology consulting management method, characterized in that, The information technology consulting management system based on big data, as described in any one of claims 1-9, is used to manage information technology consulting.

Images