Nuclear energy customer priority evaluation system
By utilizing a nuclear energy customer priority assessment system, multi-objective Pareto front analysis and optimized particle swarm optimization algorithm are employed to integrate multi-dimensional data, resolving the issue of conflicting multiple indicators in market assessment methods. This achieves optimal allocation of customer resources in the nuclear energy industry and improves the accuracy and efficiency of the assessment system.
Patent Information
- Application Number
- CN202510879137.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-27
- Publication Date
- 2025-10-31
AI Technical Summary
Existing market assessment methods cannot effectively capture the conflicts and dynamic changes in weights among multiple indicators. Traditional customer screening models have the drawback of unresolved target conflicts and are difficult to adapt to the diverse needs of customers. The nuclear energy industry lacks an assessment system for optimal resource allocation based on the heterogeneity of customer groups.
A nuclear energy customer priority assessment system is adopted, including a data layer module, a model layer module, and an application layer module. It utilizes multi-objective Pareto front analysis and optimized particle swarm optimization algorithm to integrate multi-dimensional data. Priority is ranked based on the probability of future visits from countries, national purchasing power, and national security index. Combined with machine learning prediction models and optimized particle swarm optimization algorithm, the weights are dynamically adjusted to achieve multi-dimensional equilibrium.
It significantly improves the algorithm convergence and global search capability of the evaluation system, enabling accurate ranking of target customer countries worldwide and providing a scientific basis for enterprises' international expansion and resource allocation.
Smart Images
Figure CN120875602A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of nuclear energy technology strategy, and in particular to a nuclear energy customer priority assessment system. Background Technology
[0002] In traditional international and even market promotion, accurately assessing potential markets and rationally allocating resources across numerous countries is crucial for corporate strategic decision-making. However, existing market assessment methods often rely on static weight models or simple weighted summation methods, failing to effectively capture conflicts between multiple indicators and dynamic changes in weights. Traditional customer screening models (such as the linear weighted average power method, AHP) suffer from the core flaw of unsolvable objective conflicts. Linear models cannot effectively handle such nonlinear contradictions. This contradiction stems from the inherent characteristics of multi-objective optimization—an irreconcilable competitive relationship exists between objectives, thus requiring more complex mathematical tools to achieve multidimensional equilibrium. Furthermore, the traditional "one-size-fits-all" promotion model is ill-suited to the diverse needs of customers. Currently, the nuclear energy industry lacks an assessment system that leverages a tiered strategy to achieve optimal resource allocation based on the heterogeneity of customer groups. Summary of the Invention
[0003] This invention provides a nuclear energy customer priority assessment system to solve the problem in the existing technology that the nuclear energy industry lacks a system for prioritizing customers based on the heterogeneity of customer groups.
[0004] The technical solution of the present invention is as follows:
[0005] This invention proposes a nuclear energy customer priority assessment system, which includes a data layer module, a model layer module, and an application layer module. The data layer module integrates multi-dimensional data of target customer countries. The model layer module reads the data from the data layer module, combines multi-objective Pareto front analysis modeling and optimized particle swarm optimization algorithm to prioritize target customer countries based on three aspects: country's future visit probability, country's purchasing power, and national security index. The application layer module reads the target customer country priority ranking from the model layer module and outputs a nuclear energy customer country priority ranking table based on the ranking score.
[0006] In some embodiments, the multidimensional data integrated by the data layer module includes nominal GDP, regional development coordination index, participation in international security cooperation, cross-regional infrastructure cooperation, whether it is an emerging economy, Asia-Pacific economic cooperation, regional security and economic sharing status, technological cooperation, risk index, nuclear energy demand, visiting organizations, job level, and number of visits.
[0007] In some embodiments, the model layer module establishes a set V for the target customer countries through multi-objective Pareto front analysis modeling, as shown in formula (1);
[0008] V = {v1, v2, ..., v n} (1)
[0009] Among them, v i For the i-th target customer country, where each v i Define its characteristic index vector f(vi) as shown in formula (2);
[0010] f(vi) = (pi, b) i , si) (2)
[0011] Where pi represents the probability of future visits from the target customer's country; b i represents the purchasing power score of the target customer's country; si represents the national security index of the target customer; the model layer module takes the joint maximization of the feature vector of the target customer's country across all dimensions as its objective, and uses the optimized particle swarm optimization algorithm to perform calculations to obtain the priority ranking of the target customer's countries.
[0012] In some embodiments, the model layer module uses machine learning to learn a mapping function of the future probability of visiting the target customer's country, constructs a prediction model, and calculates the future probability of visiting the target customer's country.
[0013] In some embodiments, the prediction model of the model layer module is as shown in formula (3):
[0014] pi = f_visit(zi) (3)
[0015] Where pi represents the probability of future visits from the target customer's country, f_visit is the mapping function, and zi is the feature vector of the target customer's country; the mapping function f_visit is specifically shown in formula (4):
[0016] f_visit: Z → [0,1] (4)
[0017] Where Z is the set of feature vectors zi of all target customer countries; zi is the feature vector of the target customer country as shown in formula (5):
[0018] z i =(z i1 ,z i2 ,...,z im (5)
[0019] Where the eigenvector z i The elements (z) in i1 ,z i2 ,...,z imThis includes total number of visits, number of days since the last visit, visit frequency, seasonal preference characteristics, and unit type. When the prediction model calculates the probability of future visits to the target customer's country through machine learning, it quantifies the contribution of each element to the prediction of the probability of visits through feature importance analysis.
[0020] In some embodiments, the target customer's country purchasing power score b i The specific calculation method is shown in formula (6):
[0021] b i =α1(GDP) i -min(GDP)) / (max(GDP)-min(GDP))+α2(GDP per capita i -min (per capita)
[0022] GDP)) / (max(GDP per capita)-min(GDP per capita))(6)
[0023] Among them, GDP i Let $\mathbf{i}$ be the GDP of target customer country $i$ in the current year, $\min(GDP)$ be the lowest GDP of target customer country $i$ over the past 5 years, $\max(GDP)$ be the highest GDP of target customer country $i$ over the past 5 years, and $\mathbf{i}$ be the GDP per capita. i α1 is the GDP per capita of target customer country i in the current year, min(GDP per capita) is the lowest GDP per capita of target customer country i in the past 5 years, max(GDP per capita) is the highest GDP per capita of target customer country i in the past 5 years, and α1 and α2 are user-defined parameters, where α1+α2=1.
[0024] In some embodiments, the target customer's national security index si is calculated using the specific method described in formula (7):
[0025] si = (1 - Risk Score / Maximum Risk Score) (7)
[0026] Among them, the risk score and the maximum risk score are obtained from the data layer module and are the absolute values of the risk index of the data layer module.
[0027] In some embodiments, the model layer module aims to jointly maximize the feature vectors of the target customer country across all dimensions, as shown in formula (8):
[0028] max f(vi) = (pi, b) i , si), vi ∈ V (8)
[0029] Where pi represents the probability of future visits from the target customer's country; b irepresents the purchasing power score of the target customer's country; si represents the national security index of the target customer; f(vi) is the feature vector of the target customer's country; solve the Pareto optimal solution set P according to formula (8). The Pareto optimal solution set P is defined as the absence of another solution vi such that vi is better than vi on all objectives.
[0030] In some embodiments, the model layer module employs an optimized particle swarm optimization algorithm for computation. Each particle in the particle swarm is represented as a weight vector (w1, w2, w3), where w1 controls the relative importance of the target customer country's future visit probability pi, and w2 controls the target customer country's purchasing power score b. i The relative importance of w3 is used to determine the relative importance of the national security index si of the target customer. The optimized particle swarm algorithm uses Latin hypercube sampling to intelligently initialize the particle position. When the particle approaches the Pareto front boundary, a bounce mechanism is triggered. The particle velocity update formula of the optimized particle swarm algorithm is as shown in formula (9):
[0031] v i k+1 = w×v i k + c1 ×r1 × (pbest i - x i ) + c2 × r2 × (gbest - x i (9)
[0032] Where w is the inertia weight, which decreases linearly from 0.9 to 0.4; c1 and c2 are learning factors, which are dynamically adjusted using the Spearman rank correlation coefficient; r1 and r2 are random numbers, and v i k Let v be the velocity vector of particle i during the k-th iteration. i k+1 pbest is the velocity vector of particle i at the (k+1)th iteration. i Let be the best position found by particle i in the historical iterations, and gbest be the best position found by the entire particle swarm in the historical iterations; the particle position update formula for optimizing the particle swarm algorithm is as shown in formula (10):
[0033] x i k+1 = x i k + v i k+1 (10)
[0034] Where, x i k+1 Let x be the position vector of particle i at the (k+1)th iteration. ik Let be the position vector of particle i at the k-th iteration; after each particle update, optimize the particle swarm algorithm to perform weight normalization to ensure that the sum of all weights is 1. The normalization formula is as shown in formula (11):
[0035] w normalized = w / sum(w) (11)
[0036] Among them, w normalized is the weight vector after weight normalization, and w is the weight vector before normalization.
[0037] In some embodiments, when the particle swarm swarm stagnates, the optimized particle swarm algorithm of the model layer module reinitializes the 20% of particles with the worst individual fitness through a restart strategy. If the change in the global optimal fitness of the optimized particle swarm algorithm is less than the set early stopping threshold in 30 consecutive iterations, the model layer module will terminate the algorithm early.
[0038] The implementation of this invention has the following beneficial effects:
[0039] This invention proposes a nuclear energy customer priority assessment system. It utilizes three indicators—national future visit probability, national security index, and national purchasing power score—to prioritize target customer countries globally through weight optimization. The invention employs several strategies, including Latin hypercube sampling initialization, dynamic inertia weight and learning factor adjustment, boundary bounce strategy, and early stopping mechanism, significantly improving the convergence and global search capability of the assessment system algorithm. This method can accurately rank target customer countries in the global market, providing a scientific basis for enterprises' international expansion and resource allocation. Attached Figure Description
[0040] Figure 1 This is a three-dimensional schematic diagram of the Pareto optimal solution of a nuclear energy customer priority assessment system proposed in an embodiment of the present invention;
[0041] Figure 2 This is a schematic diagram of the convergence curve of the optimized particle swarm optimization algorithm for a nuclear energy customer priority assessment system proposed in an embodiment of the present invention.
[0042] Figure 3 This is a feature vector correlation heatmap of a nuclear energy customer priority assessment system proposed in an embodiment of the present invention. Detailed Implementation
[0043] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0044] like Figures 1 to 3 As shown, this invention proposes a nuclear energy customer priority assessment system, which includes a data layer module, a model layer module, and an application layer module.
[0045] The data layer module integrates multidimensional data on the target customer's country, including 13 indicators, as detailed in Table 1 below. Table 1. Pareto Dimension Definition Indicators NO. Evaluation indicators Notes 1 Nominal GDP Using the 2023 nominal GDP (in US dollars) published by the International Monetary Fund (IMF) as an indicator, it reflects a country's current economic strength and affects its ability to make independent investments or import technology. 2 Regional Development Coordination Index Used to quantitatively assess the balance and sustainability of a specific region across various dimensions of economic and social development, reflecting the overall stability of the macro-environment. 3 international security cooperation participation To measure the participation of research subjects in multilateral security governance mechanisms. 4 Cross-regional infrastructure cooperation Whether the research subjects participate in transnational cooperation initiatives with infrastructure connectivity as the core objective reflects the breadth of their external economic links. 5 Is it an emerging economy? Indicate whether the research subject is an emerging market country to reflect its role in the cooperation. 6 Asia-Pacific Economic Cooperation The assessment evaluates the extent to which the research subjects are integrated into the Asia-Pacific regional economic integration process, covering institutional cooperation in areas such as trade and investment. 7 Regional security and economic sharing status The study identifies whether the research subjects have joined cross-regional multilateral cooperation mechanisms aimed at promoting integrated security and economic development. 8 Technical cooperation Does the existence of nuclear power or nuclear technology cooperation provide historical evidence for technology transfer and trust building? 9 Risk Index The overall risk level of project implementation is reflected by factors such as comprehensive credit rating, sovereign debt risk, and frequency of natural disasters. 10 Nuclear energy demand Human assessment of the country's current energy structure to determine its actual market demand for SMR technology. 11 Visiting Unit The type of institution or organization to which the survey subjects belong is used to measure organizational-level decision-making influence and resource mobilization capabilities. 12 Job level The survey respondents' job level within their workplace reflects their influence and power in the decision-making chain. 13 Number of visits The cumulative number of visits to China by survey respondents or their affiliated organizations for exchanges or negotiations in the field of nuclear energy is used to measure the frequency of bilateral interactions and the sustainability of the willingness to cooperate. The model layer module reads data from the data layer module and combines multi-objective Pareto modeling frontier analysis with particle swarm optimization (PSO) algorithm. The PSO algorithm uses Latin hypercube sampling (LHS) to initialize the particle swarm, dynamic inertia weights and boundary bounce strategy to prioritize target customer countries based on three aspects: probability of future visits, purchasing power and national security index. The model layer module establishes a set V for target customer countries through multi-objective Pareto modeling, as shown in formula (1). V = {v1, v2, ..., v n} (1) Among them, v i For the i-th target customer country, where each v i Define its feature index vector f(vᵢ) as shown in formula (2); f(vᵢ) = (pᵢ, bᵢ, sᵢ) (2) Where pᵢ represents the probability of a future visit from the target customer's country; bᵢ represents the purchasing power score of the target customer's country; and sᵢ represents the national security index of the target customer.
[0054] For pi, representing the probability of a target customer country's future visit refers to the likelihood of a representative from that country visiting again within the next six months, which is a typical time series evolution prediction problem. Considering that the visiting behavior is driven by multiple factors such as historical frequency, time distribution, and unit nature, this paper introduces machine learning methods to establish a more generalizable prediction model, as shown in formula (3):
[0055] pi = f_visit(zi) (3)
[0056] Where pi represents the probability of future visits from the target customer country, f_visit is the mapping function, and zi is the feature vector of the target customer country. The task of predicting the probability of visits can be formalized as a regression problem, with the goal of learning a mapping function f_visit, as shown in formula (4).
[0057] f_visit: Z → [0,1] (4)
[0058] Where Z is the set of feature vectors zi of all target customer countries; let the historical behavior records of a certain country i be the feature vector zi, as shown in formula (5).
[0059] z i =(z i1 ,z i2 ,...,z im (5)
[0060] eigenvector z i The elements (z) in i1 ,z i2 ,...,z im This includes, but is not limited to: total number of visits (Count), number of days since the last visit (Recency), visit frequency (Frequency), seasonality, and business type. The feature vectors of all target customer countries collectively constitute the input space Z. The prediction model ultimately outputs the probability of a visit to each country within the next six months, pi∈[0,1]. Furthermore, feature importance analysis quantifies the contribution of each feature to the predicted probability of a visit, providing a basis for subsequent decision-making.
[0061] Target customer country purchasing power score b i The specific calculation method is shown in formula (6):
[0062] b i =α1(GDP) i -min(GDP)) / (max(GDP)-min(GDP))+α2(GDP per capita i -min (per capita)
[0063] GDP)) / (max(GDP per capita)-min(GDP per capita)) (6)
[0064] Among them, GDP iLet $\mathbf{i}$ be the GDP of target customer country $i$ in the current year, $\min(GDP)$ be the lowest GDP of target customer country $i$ over the past 5 years, $\max(GDP)$ be the highest GDP of target customer country $i$ over the past 5 years, and $\mathbf{i}$ be the GDP per capita. i α1 represents the GDP per capita of target customer country i in the current year; α2 represents the lowest GDP per capita of target customer country i over the past 5 years; α3 represents the highest GDP per capita of target customer country i over the past 5 years; α4 and α5 are user-defined parameters, where α1 + α2 = 1. α6 represents the national security index of the target customer country. i The specific calculation method is shown in formula (7):
[0065] s i = (1 - Risk Score / Maximum Risk Score) (7)
[0066] The risk score and maximum risk score are obtained from the data layer module and are the absolute values of the risk index of the data layer module. The risk score is normalized by dividing it by the maximum risk score, and then reverse-processed to obtain the target customer's national security index s. i .
[0067] In a multi-indicator evaluation system, there are often irreconcilable conflicts between different objectives. The optimal solution for a single objective cannot take all indicators into account. Therefore, the Multi-Objective Optimization (MOO) method is introduced. The objective of this problem is to jointly maximize the feature vector of the target customer country across all dimensions, as shown in formula (8):
[0068] max f(v i )=(p i ,b i ,s i ), v i ∈V (8)
[0069] The Pareto optimal solution set P is defined as: there is no other solution v. i Make v i It is superior to vi on all objectives, i.e., the solution is: P = {v} i ∈V|v does not exist j ∈V such that v j Dominate v i}
[0070] The model layer module performs calculations using an optimized particle swarm optimization algorithm based on the Pareto objective. A bounce mechanism is triggered when particles approach the Pareto front boundary. This calculation generates a priority ranking table of target customer countries and a 3D Pareto front visualization map. Figure 1 As shown.
[0071] In the optimized particle swarm optimization algorithm, each particle is represented as a weight vector (w1, w2, w3), where w1 controls the probability p of future visits from the target customer's country. i The relative importance of w2 is used to control the purchasing power score of the target customer's country. i The relative importance of W3 for target customer national security index i The relative importance of particles is emphasized. The model layer module employs Latin hypercube sampling (LHS) for intelligent initialization of particle positions, ensuring that particles are uniformly distributed in the search space.
[0072] The velocity and position of particles in the optimized particle swarm optimization algorithm are updated using the following formulas (9) and (10):
[0073] v i k+1 =w×v i k +c1×r1×(pbest i -x i )+c2×r2×(gbest-x i (9)
[0074] x i k+1 =x i k +v i k+1 (10)
[0075] Where w is the inertia weight, c1 and c2 are learning factors, r1 and r2 are random numbers, and v i k Let v be the velocity vector at the k-th iteration. i k+1 This represents the velocity vector at the (k+1)th iteration. The inertia weight w decreases linearly from 0.9 to 0.4, and the learning factors c1 and c2 are adaptively adjusted. The learning factors c1 and c2 are dynamically corrected using the Spearman rank correlation coefficient. (pbest) i Let x be the optimal position found by particle i in the historical iterations, and gbest be the optimal position found by the entire particle swarm in the historical iterations. i k+1 Let x be the position vector of particle i at the (k+1)th iteration. i k Let be the position vector of particle i during the k-th iteration.
[0076] After each particle update, the particle swarm optimization algorithm performs weight normalization to ensure that the sum of all weights is 1. This avoids solutions that do not meet the constraints. The normalization formula is shown in formula (11):
[0077] w normalized =w / sum(w) (11)
[0078] Among them, w normalized Here, w is the weight vector after weight normalization, and w is the weight vector before normalization. The weight vector is (w1, w2, w3), which ensures that w1 + w2 + w3 = 1 after each side is updated.
[0079] Meanwhile, the optimized particle swarm optimization algorithm introduces a restart strategy. When the iteration count meets the requirement of every 50 iterations, the 20% of particles with the worst fitness are reinitialized using the restart strategy. In addition, the optimized particle swarm optimization algorithm also has an early stopping mechanism. The early stopping mechanism will terminate the algorithm early if there is no significant progress in 30 consecutive iterations (i.e., the change in the global optimal fitness is less than the set early stopping threshold 1e-8), thus avoiding invalid computation.
[0080] The application layer module reads and sorts the target customer countries by priority from the model layer module, and outputs a nuclear energy customer country priority sorting table.
[0081] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.
Claims
1. A nuclear energy customer priority assessment system, characterized in that, The system comprises a data layer module, a model layer module, and an application layer module. The data layer module integrates multidimensional data on target customer countries. The model layer module reads the data from the data layer module and, combining multi-objective Pareto front analysis modeling and optimized particle swarm optimization algorithm, prioritizes target customer countries based on three aspects: probability of future visits, purchasing power, and national security index. The application layer module reads the target customer country priority ranking from the model layer module and outputs a nuclear energy customer country priority ranking table based on the ranking score.
2. The nuclear energy customer priority assessment system according to claim 1, characterized in that, The data layer module integrates multidimensional data including nominal GDP, regional development coordination index, participation in international security cooperation, cross-regional infrastructure cooperation, whether it is an emerging economy, Asia-Pacific economic cooperation, regional security and economic sharing status, technological cooperation, risk index, nuclear energy demand, visiting organizations, job level, and number of visits.
3. A nuclear energy customer priority assessment system according to claim 2, characterized in that, The model layer module establishes a set V for the target customer countries through multi-objective Pareto front analysis modeling, as shown in formula (1); V={v1,v2,...,v n } (1) Among them, v i For the i-th target customer country, where each v i Define its feature index vector f(v) i ), specifically as shown in formula (2); f(v i )=(p i ,b i ,s i ) (2) Where, p i Indicates the probability of future visits from the target customer's country; b i This indicates the purchasing power score of the target customer's country; s i The target customer's national security index is represented by the model layer module, which uses the joint maximization of the feature vectors of the target customer's country across all dimensions as its objective and employs an optimized particle swarm optimization algorithm to perform calculations and obtain the priority ranking of the target customer's country.
4. A nuclear energy customer priority assessment system according to claim 3, characterized in that, The model layer module uses machine learning to learn the mapping function of the future visit probability of the target customer's country, constructs a prediction model, and calculates the future visit probability of the target customer's country.
5. A nuclear energy customer priority assessment system according to claim 4, characterized in that, The prediction model of the model layer module is as shown in formula (3): p i =f_visit(z i ) (3) Where, p i The z represents the probability of future visits from the target customer's country, f_visit is the mapping function, and z represents the probability of future visits from the target customer's country. i The feature vector of the target customer's country; the mapping function f_visit is as shown in formula (4): f_visit:Z→[0,1] (4) Where Z is the feature vector z of all target customer countries. i A collection; z i The feature vector for the target customer's country is specifically shown in formula (5): With i =(of i1 ,With i2 ,...,With im ) (5) Where the eigenvector z i The elements (z) in i1 ,z i2 ,...,z im The data includes total number of visits, number of days since the last visit, visit frequency, seasonal preference characteristics, and unit type. When the prediction model calculates the probability of future visits to the target customer's country through machine learning, it quantifies the contribution of each element to the prediction of the probability of visits through feature importance analysis.
6. A nuclear energy customer priority assessment system according to claim 3, characterized in that, The target customer's country purchasing power score b i The specific calculation method is shown in formula (6): b i =α1(GDP) i -min(GDP)) / (max(GDP)-min(GDP))+α2(GDP per capita i -min(GDP per capita)) / (max(GDP per capita)-min(GDP per capita)) (6) Among them, GDP i Let $\mathbf{i}$ be the GDP of target customer country $i$ in the current year, $\min(GDP)$ be the lowest GDP of target customer country $i$ over the past 5 years, $\max(GDP)$ be the highest GDP of target customer country $i$ over the past 5 years, and $\mathbf{i}$ be the GDP per capita. i α1 is the GDP per capita of target customer country i in the current year, min(GDP per capita) is the lowest GDP per capita of target customer country i in the past 5 years, max(GDP per capita) is the highest GDP per capita of target customer country i in the past 5 years, and α1 and α2 are user-defined parameters, where α1+α2=1.
7. A nuclear energy customer priority assessment system according to claim 3, characterized in that, The target customer's national security index i The specific calculation method is shown in formula (7): s i = (1 - Risk Score / Maximum Risk Score) (7) Among them, the risk score and the maximum risk score are obtained from the data layer module and are the absolute values of the risk index of the data layer module.
8. A nuclear energy customer priority assessment system according to claim 3, characterized in that, The model layer module takes the joint maximization of the feature vectors of the target customer's country across all dimensions as its objective, as shown in formula (8): max f(v i )=(p i ,b i ,s i ),v i ∈V (8) Where, p i Indicates the probability of future visits from the target customer's country; b i This indicates the purchasing power score of the target customer's country; s i f(v) represents the national security index of the target customer. i ) represents the feature vector of the target customer's country; the Pareto optimal solution set P is obtained according to formula (8), wherein the Pareto optimal solution set P is defined as having no other solution v. i Make v i It outperforms vi in all objectives.
9. A nuclear energy customer priority assessment system according to claim 8, characterized in that, The model layer module uses an optimized particle swarm optimization algorithm for computation. Each particle in the swarm is represented as a weight vector (w1, w2, w3), where w1 controls the probability p of future visits from the target customer's country. i The relative importance of w2 is used to control the purchasing power score of the target customer's country. i The relative importance of W3 for target customer national security index i To assess the relative importance of particles, the optimized particle swarm optimization algorithm employs Latin hypercube sampling for intelligent initialization of particle positions, triggering a bounce mechanism when a particle approaches the Pareto front boundary. The velocity update formula for the particles in the optimized particle swarm optimization algorithm is shown in formula (9): v i k+1 =w×v i k +c1×r1×(pbest i -x i )+c2×r2×(gbest-x i ) (9) Where w is the inertia weight, which decreases linearly from 0.9 to 0.4; c1 and c2 are learning factors, which are dynamically adjusted using the Spearman rank correlation coefficient; r1 and r2 are random numbers, and v i k Let v be the velocity vector of particle i during the k-th iteration. i k+1 pbest is the velocity vector of particle i at the (k+1)th iteration. i Let be the optimal position found by particle i in the historical iterations, and gbest be the optimal position found by the entire particle swarm in the historical iterations; the particle position update formula of the optimized particle swarm algorithm is as shown in formula (10): x i k+1 =x i k +v i k+1 (10) Where, x i k+1 Let x be the position vector of particle i at the (k+1)th iteration. i k Let be the position vector of particle i at the k-th iteration; after each particle update, optimize the particle swarm algorithm to perform weight normalization to ensure that the sum of all weights is 1. The normalization formula is as shown in formula (11): w normalized =w / sum(w) (11) Among them, w normalized is the weight vector after weight normalization, and w is the weight vector before normalization.
10. A nuclear energy customer priority assessment system according to claim 8, characterized in that, When the particle swarm optimization algorithm of the model layer module stagnates, it reinitializes the 20% of particles with the worst individual fitness through a restart strategy. If the change in the global optimal fitness is less than the set early stopping threshold in 30 consecutive iterations of the optimized particle swarm optimization algorithm, the model layer module will terminate the algorithm early.