Demand analysis and purchase method and system for new product project management

Abstract

The invention belongs to the technical field of demand analysis, and discloses a demand analysis and purchasing method and system for new product project management. The method comprises the following steps: collecting M groups of project demand data; performing clustering analysis on the project demand data to obtain n demand clusters; constructing a demand structure relation model for each demand cluster, and obtaining an effective quality set based on the demand structure relation model; optimizing the demand structure relation model to obtain an optimal demand distribution pattern; according to the optimal demand distribution pattern, supplier optimization selection is carried out, and an optimal supplier combination is obtained; the optimal supplier combination is used as a purchasing scheme, the purchasing risk is evaluated according to the effective quality set, risk coping measures are formulated, high-risk demands and suppliers are accurately recognized, differentiated risk coping strategies are formulated, the purchasing risk is prevented and reduced from the source, and it is guaranteed that the project is smoothly propelled.

Description

Technical Field

[0001] The present invention relates to the technical field of demand analysis, and more specifically, to a demand analysis and procurement method and system for new product project management. Background Art

[0002] As market competition intensifies, companies are increasingly in need of launching new products and services; the successful implementation of new product projects is crucial for companies to maintain their competitiveness and expand into new markets. However, new product projects usually involve complex demand analysis and procurement processes, which require weighing multiple factors, such as demand priority, technical complexity, supplier capabilities, cost control, etc., which brings many challenges to the efficient advancement of projects.

[0003] In the implementation of new product projects, traditional demand analysis and procurement methods have many shortcomings, which have brought many obstacles to the smooth progress of the project; first, the existing methods are difficult to fully grasp the inherent correlation between demands, and often consider each demand as an independent individual, ignoring the mutual influence and constraint relationship between demands. This isolated analysis method cannot accurately assess the actual importance and complexity of each demand, resulting in unreasonable resource allocation and inefficient project execution; secondly, in the supplier selection link, the existing methods usually only consider a single factor, such as cost or quality, and lack a comprehensive evaluation of multi-dimensional indicators. This simplistic approach makes it difficult to fully grasp the actual capabilities of suppliers, which may lead to improper selection, affect the quality of new product delivery, and increase procurement risks; thirdly, the traditional methods lack scientificity in demand distribution and resource allocation, and are prone to concentrated or over-dispersed demand, resulting in waste of resources or excessive pressure on the supply chain, which seriously affects the efficient advancement of the project; in addition, the existing methods have obvious shortcomings in procurement risk management and control, lack of differentiated identification and response to different demands and supplier risks, and cannot timely discover and resolve potential risks, which may lead to project delays or quality problems, causing significant losses to the company.

[0004] In view of this, the present invention proposes a demand analysis and procurement method and system for new product project management to solve the above problems. Summary of the invention

[0005] In order to overcome the above defects of the prior art and to achieve the above objectives, the present invention provides the following technical solutions: A demand analysis and procurement method for new product project management, comprising: step 1, collecting M groups of project demand data; Step 2: Perform cluster analysis on the project demand data to obtain n demand clusters; Step 3: construct a demand structure relationship model for each demand cluster, and obtain a valid quality set based on the demand structure relationship model; Step 4: Optimize the demand structure relationship model to obtain the optimal demand distribution pattern; Step 5: According to the optimal demand distribution pattern, optimize the supplier selection and obtain the optimal supplier combination; Step 6: Use the optimal supplier combination as the procurement plan, evaluate procurement risks based on the effective quality set, and develop risk response measures.

[0006] Furthermore, the project demand data includes basic project information, demand information and cost information; Basic project information includes project name, project type, project budget and project period; Demand information includes demand source, demand priority and demand category; cost information includes the estimated procurement cost and implementation cost for each demand.

[0007] Furthermore, the cluster analysis method includes: Standardize or normalize the project demand data, define clustering features, which include demand priority, demand category, procurement cost, and implementation cost; and represent the clustering features as a multidimensional vector; Define that the project demand data is a mixture of K probability distributions, and then obtain the mixture model corresponding to the project demand data; K is an integer greater than 1; each probability distribution corresponds to a potential cluster; Then the probability density function of the mixture model is ;in, is the covariate vector, It is The mixed weights of probability distributions satisfy that the sum of the mixed weights of all probability distributions is 1; It is The density function of a probability distribution; For the The parameters of the density function of a probability distribution, The original project requirement data; Randomly initialize the parameters of the density function of each probability distribution, and for each data point y, calculate its initial posterior probability under each probability distribution ,in, represents the hidden cluster label; ;in, It is a data point Likelihood under the kth probability distribution; based on the calculated posterior probability, re-estimate the comprehensive parameters of each probability distribution, which include the parameters of the mixing weight and density function; the re-estimation formula is: ;in, is the index of the data point; is the total number of data points; is the mixing weight of the kth probability distribution after re-estimation; ; is the parameter of the density function of the kth probability distribution after re-estimation; repeat until the preset maximum number of iterations is reached; Set the number of clusters n. For different numbers of clusters n, use the number of clusters as the number of probability distributions and iteratively update the corresponding mixed models. For each mixed model, calculate its log-likelihood value. and the number of parameters p; based on the log-likelihood and the number of parameters p to calculate the fit indices ;in, is the adjustment factor; select The number of clusters with the smallest value is taken as n; Take n at this time as the number of probability distributions, calculate the posterior probability of each data point belonging to each probability distribution, and assign each data point to the probability distribution with the largest posterior probability. The probability distribution at this time is the demand cluster.

[0008] Furthermore, the covariate vector is obtained by: Select project type, project budget and project cycle as covariate features; map the selected covariate features to a multidimensional feature space; at this time, the project demand data is used as data points in the multidimensional feature space; calculate the distance between any two data points in the multidimensional feature space, and construct a distance matrix D; set a seed point set S, randomly select a data point as the first seed point s1, add s1 to the seed point set S, iteratively select seed points, for the current seed point set S, calculate the closest distance between each data point that is not a seed point and the seed point in S, select the data point with the largest closest distance as the new seed point, add the new seed point s_new to the seed point set S, set a distance threshold, and repeat until the closest distance of all data points that are not seed points to the seed points in S is less than the distance threshold, stop the iterative selection; starting from each seed point in the seed point set S, extend several scan lines in the multidimensional feature space, and define the direction of the scan line; Along each scan line, the data points encountered are marked as the neighborhood points of the corresponding seed point. For each seed point, all the data points in the neighborhood formed by its neighborhood points are regarded as a cluster. The statistics of the covariate features of the data points in each cluster are calculated, and these statistics are constructed into a covariate vector.

[0009] Furthermore, the method of defining the direction of the scan line includes: For each data point , calculate its density value in the multidimensional feature space and the density gradient vector ;in, is the preset step size; for each seed point s, obtain its density value and density gradient vector; Initially, the direction of the scan line is defined as the direction of the density gradient vector of the seed point s; recorded as the initial direction; the scan line is extended from the seed point s along the initial direction, and a step length is taken on the scan line , calculate the new data point at this step size The density gradient vector ; Adjust the direction of the scan line to The scan line direction is adjusted repeatedly to extend in the direction where the density increases fastest. The maximum length of the scan line is set to L_max. When the length of the scan line exceeds L_max or encounters an area where the density gradient vector is 0, the scan line is terminated. The next scan line is restarted from the seed point, and the direction is determined by the current density gradient vector.

[0010] Furthermore, the method of constructing the demand structure relationship model includes: The data points in the demand cluster are regarded as the demand; for each demand cluster, a two-dimensional grid structure is constructed, and the two-dimensional grid structure contains a number of regular grids; For each demand in the demand cluster, a mapping function is used to map the demand to a grid point on the grid of the two-dimensional grid structure; a grid field is obtained; the scope of interaction is defined, and the interaction between the demands is calculated; the calculation formula for the interaction is: ;in, For demand and demand the interactions between them; For demand The priority of For demand The priority of For demand The cost, For demand Costs; and is the weight coefficient; Defining the Energy Field of the Lattice Field ; ;in Grid Field operators on ; is the quality parameter; is the coupling constant; is a differential operator, which represents the difference between adjacent grid points; for each grid point , calculate the local effective mass ; ;in, is the traversal index of the grid point, traversal and grid point Interacting grid points; At this time, the two-dimensional lattice structure integrating the local effective mass and energy field is used to construct the demand structure relationship model; The method for obtaining the effective quality set includes: Initialize the local effective mass of all grid points, use the mean field approximation or other numerical methods to solve the energy field, obtain the field operator, and repeatedly calculate the local effective mass of each grid point based on the field operator until the preset number of iterations is reached; fix the local effective mass of each grid point at this time to form an effective mass set.

[0011] Furthermore, the method for obtaining the optimal demand distribution pattern includes: Obtain the initial demand distribution pattern from the demand structure relationship model, that is, the initial mapping position of each demand on the grid; set the initial temperature T0, temperature drop rate a1, and end temperature Tend; define the objective function as the energy field and the weighted sum of the uniformity of demand distribution; The formula for demand distribution uniformity is: ;in, Grid The demand density on To meet the demand for uniform distribution, is the average value of the demand density of all grid points; is the number of rows in the two-dimensional lattice structure, is the number of columns of the two-dimensional grid structure; the demand density is the weighted sum of the number and priority of the demand at the corresponding grid point; Calculate the value of the objective function corresponding to the initial demand distribution pattern ; Set the current temperature , set an iteration number for each temperature; at the current temperature, iterate: For the demand structure relationship model, its neighborhood structure is defined as 8 neighborhoods; a fixed radius is preset, and for each grid point, the demand density within the fixed radius around it is calculated; the maximum neighborhood radius Rmax, the minimum neighborhood radius Rmin and the density threshold are set; for grid points with demand density less than or equal to the density threshold, its neighborhood radius is set to Rmin; for grid points with demand density greater than the density threshold, its neighborhood radius is set to Rmax; for each demand, the corresponding adaptive neighborhood is constructed according to the neighborhood radius of the grid point where it is located; Randomly select a demand, and randomly select a new location based on the adaptive neighborhood of the grid point where the demand is currently located; obtain a new distribution pattern; calculate the value of the objective function of the new distribution pattern , calculate the energy difference ; like If it is less than or equal to 0, the new distribution pattern is accepted and updated. ;like If greater than 0, then the probability Accept new distribution patterns; Repeat until the number of iterations at that temperature is reached; lower the temperature and repeat the iteration, lowering the temperature to the temperature drop rate a1 multiplied by the current temperature as the temperature for the next iteration; until the current temperature is less than Tend; at this time, the new distribution pattern finally output is the optimized demand distribution pattern.

[0012] Furthermore, the method for obtaining the optimal supplier combination includes: Collect relevant information about potential suppliers, including supplier name, supplier type, supplier capabilities, supplier rating, and supplier cost; Determine the constraints for supplier selection based on demand and budget; formalize the constraints as hard constraints; build a model for evaluating supplier quality based on relevant information of potential suppliers; the output of the model is the supplier's quality score; Treat each demand in the optimal demand distribution pattern as a molecule; assign an initial quality value to each molecule, determined based on the priority and complexity index of the demand; construct an optimization function; the optimization function is the weighted sum of the supplier's quality score and the quality value of the molecule; Combine suppliers, assign a supplier to each molecule, obtain the supplier combination, and calculate the value of the optimization function under the current supplier combination; use the optimization algorithm to iterate the optimization of the supplier combination, stop the optimization iteration when the value of the optimization function converges, and output the supplier combination at this time, which is the optimal supplier combination.

[0013] A demand analysis and procurement system for new product project management, which is used to implement a demand analysis and procurement method for new product project management, comprising: a data collection module, which is used to collect M groups of project demand data; The preliminary analysis module is used to cluster the project demand data and obtain n demand clusters; A model building module is used to build a demand structure relationship model for each demand cluster and obtain a valid quality set based on the demand structure relationship model; Optimization module, used to optimize the demand structure relationship model and obtain the optimal demand distribution pattern; The supplier matching module is used to optimize supplier selection according to the optimal demand distribution pattern to obtain the optimal supplier combination; The risk prediction module is used to use the optimal supplier combination as a procurement plan, evaluate procurement risks based on the effective quality set, and formulate risk response measures; each module is connected through wired and / or wireless means to achieve data transmission between modules.

[0014] Technical effects and advantages of a demand analysis and procurement method and system for new product project management of the present invention: The present invention can deeply explore the intrinsic connection and influence between demands, breaking through the limitation of traditional methods that only consider a single demand. By establishing a demand structure relationship model, the actual importance and complexity of each demand can be accurately evaluated, laying the foundation for subsequent resource allocation and risk management; secondly, it realizes the balanced distribution of demands in different regions, avoids the inefficiency of resource concentration or excessive dispersion, and improves the overall resource utilization efficiency; thirdly, combining the supplier quality model with the optimization algorithm can comprehensively consider multiple key factors, select the optimal supplier combination, ensure the high-quality delivery of new product projects, and reduce the total procurement cost; in addition, the introduction of effective quality collection and supplier risk assessment mechanism can accurately identify high-risk demands and suppliers, formulate differentiated risk response strategies, prevent and reduce procurement risks from the source, and ensure the smooth progress of the project. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a demand analysis and procurement method for new product project management according to the present invention; Figure 2 A schematic diagram of a demand analysis and procurement system for new product project management according to the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Example 1 See also Figure 1 As shown, this embodiment provides a demand analysis and procurement method for new product project management, including: Step 1: Collect project demand data of group M; Step 2: Perform cluster analysis on the project demand data to obtain n demand clusters; M and n are both integers greater than 1; Step 3: construct a demand structure relationship model for each demand cluster, and obtain a valid quality set based on the demand structure relationship model; Step 4: Optimize the demand structure relationship model to obtain the optimal demand distribution pattern; Step 5: According to the optimal demand distribution pattern, optimize the supplier selection and obtain the optimal supplier combination; Step 6: Use the optimal supplier combination as the procurement plan, evaluate procurement risks based on the effective quality set, and develop risk response measures.

[0018] Project demand data includes basic project information, demand information and cost information; basic project information includes project name, project type, project budget and project cycle; demand information includes demand source, demand priority and demand category; cost information is the estimated procurement cost and implementation cost for each demand; The project name is in the form of a string, which is used to uniquely identify the project. The project type, demand source, demand priority and demand category are represented by digital codes. For the source of demand, 1 represents from customers, 2 represents from internal sources, 3 represents regulatory requirements, etc.; for the requirement category, 1 represents functional requirements, 2 represents non-functional requirements, 3 represents constraints, etc.

[0019] Methods for cluster analysis include: Standardize or normalize the project demand data so that different features are at the same level, define clustering features, which include demand priority, demand category, procurement cost, and implementation cost; and represent the clustering features as a multidimensional vector.

[0020] Define that the project demand data is a mixture of K probability distributions, and then obtain the mixture model corresponding to the project demand data; K is an integer greater than 1; each probability distribution corresponds to a potential cluster; Then the probability density function of the mixture model is ;in, is the covariate vector, It is The mixed weights of probability distributions satisfy that the sum of the mixed weights of all probability distributions is 1; It is Density function of a probability distribution, such as Gaussian distribution, Student's t distribution, etc. For the The parameters of the density function of a probability distribution, The original project requirement data.

[0021] The project type, project budget and project cycle are selected as covariate features; the selected covariate features are mapped to a multidimensional feature space; at this time, the project demand data is used as a data point in the multidimensional feature space; the distance between any two data points in the multidimensional feature space is calculated, and a distance matrix D is constructed; the distance is calculated as the Euclidean distance or cosine similarity between the vector components of the covariate features.

[0022] Set a seed point set S, randomly select a data point as the first seed point s1, add s1 to the seed point set S, iteratively select seed points, for the current seed point set S, calculate the closest distance between each data point that is not a seed point and the seed point in S, select the data point with the largest closest distance as the new seed point, add the new seed point s_new to the seed point set S, set a distance threshold, and repeat until the closest distance from all data points that are not seed points to the seed points in S is less than the distance threshold, stop the iterative selection.

[0023] Starting from each seed point in the seed point set S, several scan lines are extended in the multidimensional feature space; and the direction of the scan line is defined; for each data point , calculate its density value in the multidimensional feature space and the density gradient vector ;in, is the preset step length; For each seed point s, obtain its density value and density gradient vector; initially, define the direction of the scan line as the direction of the density gradient vector of the seed point s; record it as the initial direction; this ensures that the scan line extends in the direction of increasing density; extend the scan line from the seed point s along the initial direction, and take a step length on the scan line , calculate the new data point at this step size The density gradient vector ; Adjust the direction of the scan line to The scan line direction is adjusted repeatedly to extend in the direction where the density increases fastest. The maximum length of the scan line is set to L_max. When the length of the scan line exceeds L_max or encounters an area where the density gradient vector is 0, the scan line is terminated. The next scan line is restarted from the seed point, and the direction is determined by the current density gradient vector.

[0024] Along each scan line, the data points encountered are marked as neighborhood points of the corresponding seed point. The same data point may be covered by scan lines of multiple seed points. In this case, conflicts need to be resolved (e.g., the seed point with the closest distance wins).

[0025] For each seed point, all data points in the neighborhood formed by its neighboring points are regarded as a cluster, and the statistics of the covariate characteristics of the data points in each cluster (such as mean, variance, etc.) are calculated, and these statistics are used to form a covariate vector , which is input into the mixed model as the covariate information of the corresponding cluster.

[0026] Randomly initialize the parameters of the density function of each probability distribution, and for each data point y, calculate its initial posterior probability under each probability distribution ,in, Represents a hidden cluster label; it actually reflects the potential clustering affiliation of the data point and is a potential random variable that needs to be inferred and estimated based on the observed data.

[0027] ;in, It is a data point The likelihood under the kth probability distribution is expressed as From the The probability of a probability distribution is calculated; according to the calculated posterior probability, the comprehensive parameters of each probability distribution are re-estimated, and the comprehensive parameters include the parameters of the mixing weight and the density function.

[0028] The re-estimated formula is: ;in, is the index of the data point; is the total number of data points (the number of groups of project demand data); is the mixing weight of the re-estimated kth probability distribution.

[0029] ; is the parameter of the density function of the re-estimated kth probability distribution; repeat until the preset maximum number of iterations is reached.

[0030] Set the number of clusters n. For different numbers of clusters n, use the number of clusters as the number of probability distributions and iteratively update the corresponding mixed models. For each mixed model, calculate its log-likelihood value. and the number of parameters p; based on the log-likelihood and the number of parameters p to calculate the fit indices ;in, is the adjustment factor, usually a natural number between 2 and 5; The number of clusters with the smallest value is taken as n; n at this time is taken as the number of probability distributions, the posterior probability of each data point belonging to each probability distribution is calculated, and each data point is assigned to the probability distribution with the largest posterior probability. The probability distribution at this time is the demand cluster.

[0031] Ways to build a demand structure relationship model include: The data points in the demand cluster are regarded as demands; for each demand cluster, a two-dimensional lattice structure is constructed, and the two-dimensional lattice structure contains a number of regular grids; for each demand in the demand cluster, a mapping function (such as linear mapping or nonlinear mapping) is used to map the demand to a grid point on the grid of the two-dimensional lattice structure; and a grid field is obtained.

[0032] Define the scope of interaction (such as nearest neighbor or next nearest neighbor) and calculate the interaction between requirements; the calculation formula for interaction is: ;in, For demand and demand the interactions between them; For demand Priority (demand priority), For demand Priority (demand priority), For demand Cost (purchase cost and implementation cost), For demand Cost (purchase cost and implementation cost); and is the weight coefficient.

[0033] Defining the Energy Field of the Lattice Field ; ;in Grid The field operator on , which can be understood as describing the intensity or existence of the demand at that grid point; is a quality parameter, understood as the inherent importance of the requirement or the underlying cost; is the coupling constant, which describes the strength of the interaction between the demands; is a differential operator, which represents the difference between adjacent grid points; it needs to be explained that as well as Written in this form, it follows from the Pauli Exclusion Principle, which means that two identical field operators cannot occupy the same quantum state; this is different from ordinary number multiplication.

[0034] For each grid point , calculate the local effective mass ; ;in, is the traversal index of the grid point, traversal and grid point Interacting lattice points; At this time, the two-dimensional lattice structure integrating the local effective mass and energy field is the construction requirement structure relationship model.

[0035] Initialize the local effective mass of all lattice points, solve the energy field using the mean field approximation or other numerical methods to obtain the field operator, and repeatedly calculate the local effective mass of each lattice point based on the field operator until the preset number of iterations is reached; Fix the local effective mass of each lattice point at this time to form an effective mass set. Correspondingly, a local effective mass distribution map can be drawn to identify high / low effective mass regions, corresponding to high / low priority or complexity requirement concentration areas; Through the above steps, the effective mass of each lattice point considering the interaction between requirements is calculated, and then the relationship between requirements and the overall structure can be better understood.

[0036] The ways to obtain the optimized requirement distribution pattern include: Obtain the initial requirement distribution pattern from the requirement structure relationship model, that is, the initial mapping position of each requirement on the lattice; Set the initial temperature T0, the temperature decrease rate a1 (0 < a1 < 1), and the termination temperature Tend.

[0037] Define the objective function as the weighted sum of the energy field and the requirement distribution uniformity, which is used to evaluate the quality of the requirement distribution pattern; The formula for the requirement distribution uniformity is: ; where is the requirement density on lattice point , is the requirement distribution uniformity, is the average value of the requirement densities of all lattice points; is the number of rows of the two-dimensional lattice structure, is the number of columns of the two-dimensional lattice structure; The requirement density is the weighted sum of the number and priority of requirements on the corresponding lattice point; Calculate the value of the objective function corresponding to the initial requirement distribution pattern ; Set the current temperature , and set an iteration number for each temperature; At the current temperature, perform iterations: For the requirement structure relationship model, define its neighborhood structure as the set of adjacent positions where requirements can move on the lattice; Specifically, it is the 8-neighborhood (up, down, left, right, and the four diagonals); Preset a fixed radius, and for each lattice point, calculate the requirement density within the fixed radius around it; Set the maximum neighborhood radius Rmax, the minimum neighborhood radius Rmin, and the density threshold; For lattice points with a requirement density less than or equal to the density threshold, set their neighborhood radius to Rmin; For lattice points with a requirement density greater than the density threshold, set their neighborhood radius to Rmax; Or use a non-linear method to map the requirement density to the interval [Rmin, Rmax]. For each demand, a corresponding adaptive neighborhood is constructed according to the neighborhood radius of the grid point where it is located; the adaptive neighborhood includes the grid point and all grid points within its radius; a demand is randomly selected, and a new location is randomly selected according to the adaptive neighborhood of the grid point where the demand is currently located; a new distribution pattern is obtained; the value of the objective function of the new distribution pattern is calculated , calculate the energy difference ; like If it is less than or equal to 0, the new distribution pattern is accepted and updated. ;like If greater than 0, then the probability Accept new distribution patterns; Repeat until the number of iterations at that temperature is reached; lower the temperature and repeat the iteration, lowering the temperature to the temperature drop rate a1 multiplied by the current temperature as the temperature for the next iteration; until the current temperature is less than Tend; at this time, the new distribution pattern finally output is the optimized demand distribution pattern.

[0038] Ways to obtain the best supplier combination include: Collect relevant information about potential suppliers, including supplier name, supplier type, supplier capability, supplier rating, and supplier cost; pre-process relevant information, such as filling missing values ​​and standardization, to make the data computable.

[0039] Based on demand and budget, determine the constraints for supplier selection, such as cost cap, delivery cycle requirements, quality requirements, etc.; formalize the constraints as hard constraints.

[0040] Based on the relevant information of potential suppliers, a model for evaluating supplier quality is constructed, and machine learning and other methods can be used; the output of the model is the supplier's quality score, which is used for subsequent optimization; specifically, non-numeric data is encoded, such as converting the supplier type into a numerical label; features that help evaluate supplier quality, such as on-time delivery rate, quality pass rate, customer satisfaction, etc., are extracted from the original data to construct a feature matrix, with each row corresponding to a supplier and each column corresponding to a feature; a quality score or grade is labeled for each supplier as the target value for supervised learning; the quality score can be obtained by manual evaluation or calculated based on historical performance; a machine learning algorithm is selected as the basic framework of the model, such as decision trees, random forests, support vector machines, etc.

[0041] Use the feature matrix to train the model. According to the performance of the model, adjust the model hyperparameters, such as the maximum depth of the decision tree and the number of trees in the random forest. Use cross-validation, grid search and other techniques to tune the model to improve its generalization ability on unseen data. Make the tuned model persistent for subsequent use, and regularly collect new supplier data to update the model.

[0042] Treat each demand in the optimal demand distribution pattern as a molecule; assign an initial quality value to each molecule, which is determined based on the priority and complexity index of the demand; the complexity index reflects the difficulty of realizing the demand, which is quantified by the technical complexity of the demand and the resource requirements; Construct an optimization function; the optimization function is the weighted sum of the supplier's quality score and the molecule's quality value, and the weight can be adjusted according to actual conditions.

[0043] Combine suppliers and assign a supplier to each molecule. It should be noted that suppliers can be repeated. Get supplier combinations and calculate the value of the optimization function under the current supplier combination. Use optimization algorithms (such as simulated annealing, genetic algorithms, etc.) to optimize the supplier combination iteratively. When the value of the optimization function converges, stop the optimization iteration and output the supplier combination at this time, which is the optimal supplier combination. Through the above steps, combined with the demand distribution pattern and supplier data, the optimal supplier combination that meets each demand cluster can be obtained.

[0044] The optimal supplier combination obtained through the optimization algorithm is used as the procurement plan for the project; for each demand cluster, the corresponding supplier is listed, as well as the specific demand list that needs to be purchased from the supplier; the requirements that each supplier needs to provide are summarized to form a purchase order for each supplier; the purchase order, supplier information, delivery time, etc. are organized into a formal procurement plan document.

[0045] The effective quality set reflects the actual importance or complexity of each requirement after considering the interactions between requirements. For requirements with higher effective quality, there is a greater risk in their implementation, and the supplier's delivery quality will be more critical.

[0046] According to the value of local effective quality, a risk score is assigned to each demand, and a nonlinear mapping function is used for mapping. For each supplier, the sum of the risk scores of the demands it is responsible for is summarized as the risk value of the supplier. The supplier's risk value is combined with its quality score (weighted sum) to obtain a comprehensive score. Based on the comprehensive score, the supplier is comprehensively rated to obtain a supplier rating. Specifically, the threshold is set in segments and compared with the comprehensive score to obtain the corresponding supplier rating. High-risk and high-quality suppliers need to be paid special attention. Based on the supplier rating, risk response measures are formulated, such as increasing inspections, increasing liquidated damages, etc., to reduce procurement risks.

[0047] During the procurement execution process, continuously monitor the delivery quality and progress of suppliers; for suppliers with major risks, adjust the response measures in a timely manner according to the evaluation results; if a supplier is unable to deliver on time or the quality is seriously unqualified, the supplier portfolio can be re-optimized based on the remaining demand; the new supplier portfolio needs to re-evaluate the risks, and adjust the procurement plan and risk response measures accordingly.

[0048] This embodiment can deeply explore the internal connection and influence between demands, breaking through the limitation of traditional methods that only consider a single demand. By establishing a demand structure relationship model, the actual importance and complexity of each demand can be accurately evaluated, laying the foundation for subsequent resource allocation and risk management; secondly, it realizes the balanced distribution of demands in different regions, avoids the inefficiency of resource concentration or excessive dispersion, and improves the overall resource utilization efficiency; thirdly, combining the supplier quality model with the optimization algorithm can comprehensively consider multiple key factors, select the optimal supplier combination, ensure the high-quality delivery of new product projects, and reduce the total procurement cost; in addition, the introduction of effective quality collection and supplier risk assessment mechanism can accurately identify high-risk demands and suppliers, formulate differentiated risk response strategies, prevent and reduce procurement risks from the source, and ensure the smooth progress of the project.

[0049] Example 2 See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides a demand analysis and procurement system for new product project management, including: Data collection module, used to collect M group project demand data; The preliminary analysis module is used to perform cluster analysis on the project demand data to obtain n demand clusters; M and n are both integers greater than 1; A model building module is used to build a demand structure relationship model for each demand cluster and obtain a valid quality set based on the demand structure relationship model; Optimization module, used to optimize the demand structure relationship model and obtain the optimal demand distribution pattern; The supplier matching module is used to optimize the supplier selection according to the optimal demand distribution pattern to obtain the optimal supplier combination; The risk prediction module is used to use the optimal supplier combination as a procurement plan, evaluate procurement risks based on the effective quality set, and formulate risk response measures; each module is connected through wired and / or wireless means to achieve data transmission between modules.

[0050] Example 3 This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the demand analysis and procurement method for new product project management provided above is implemented.

[0051] Since the electronic device introduced in this embodiment is an electronic device used to implement a demand analysis and procurement method for new product project management in the embodiment of this application, based on the demand analysis and procurement method for new product project management introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as the technical personnel of this field implement the electronic device used in the demand analysis and procurement method for new product project management in the embodiment of this application, it belongs to the scope of protection of this application.

[0052] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0053] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A demand analysis and procurement method for new product project management, characterized in that: include: Step 1: Collect project demand data of group M; Step 2: Perform cluster analysis on the project demand data to obtain n demand clusters; Step 3: construct a demand structure relationship model for each demand cluster, and obtain a valid quality set based on the demand structure relationship model; Step 4: Optimize the demand structure relationship model to obtain the optimal demand distribution pattern; Step 5: According to the optimal demand distribution pattern, optimize the supplier selection and obtain the optimal supplier combination; Step 6: Use the optimal supplier combination as the procurement plan, evaluate procurement risks based on the effective quality set, and develop risk response measures.

2. The demand analysis and procurement method for new product project management according to claim 1, characterized in that: The project demand data includes basic project information, demand information and cost information; Basic project information includes project name, project type, project budget and project period; Demand information includes demand source, demand priority and demand category; cost information includes the estimated procurement cost and implementation cost for each demand.

3. The demand analysis and procurement method for new product project management according to claim 2, characterized in that: The cluster analysis method includes: Standardize or normalize the project demand data, define clustering features, which include demand priority, demand category, procurement cost, and implementation cost; and represent the clustering features as a multidimensional vector; Define that the project demand data is a mixture of K probability distributions, and then obtain the mixture model corresponding to the project demand data; K is an integer greater than 1; each probability distribution corresponds to a potential cluster; Then the probability density function of the mixture model is ;in, is the covariate vector, It is The mixed weights of probability distributions satisfy that the sum of the mixed weights of all probability distributions is 1; It is The density function of a probability distribution; For the The parameters of the density function of a probability distribution, The original project requirement data; Randomly initialize the parameters of the density function of each probability distribution, and for each data point y, calculate its initial posterior probability under each probability distribution ,in, represents the hidden cluster label; ;in, It is a data point Likelihood under the kth probability distribution; based on the calculated posterior probability, re-estimate the comprehensive parameters of each probability distribution, which include the parameters of the mixing weight and density function; the re-estimation formula is: ;in, is the index of the data point; is the total number of data points; is the mixing weight of the kth probability distribution after re-estimation; ; is the parameter of the density function of the kth probability distribution after re-estimation; repeat until the preset maximum number of iterations is reached; Set the number of clusters n. For different numbers of clusters n, use the number of clusters as the number of probability distributions and iteratively update the corresponding hybrid models. For each hybrid model, calculate its log-likelihood value LP and the number of parameters p. Based on the log-likelihood value and the number of parameters p to calculate the fit indices ;in, is the adjustment factor; select The number of clusters with the smallest value is taken as n; Take n at this time as the number of probability distributions, calculate the posterior probability of each data point belonging to each probability distribution, and assign each data point to the probability distribution with the largest posterior probability. The probability distribution at this time is the demand cluster.

4. The demand analysis and procurement method for new product project management according to claim 3, characterized in that: The method of obtaining the covariate vector includes: Select project type, project budget and project cycle as covariate features; map the selected covariate features to a multidimensional feature space; at this time, the project demand data is used as data points in the multidimensional feature space; calculate the distance between any two data points in the multidimensional feature space, and construct a distance matrix D; set a seed point set S, randomly select a data point as the first seed point s1, add s1 to the seed point set S, iteratively select seed points, for the current seed point set S, calculate the closest distance between each data point that is not a seed point and the seed point in S, select the data point with the largest closest distance as the new seed point, add the new seed point s_new to the seed point set S, set a distance threshold, and repeat until the closest distance of all data points that are not seed points to the seed points in S is less than the distance threshold, stop the iterative selection; starting from each seed point in the seed point set S, extend several scan lines in the multidimensional feature space, and define the direction of the scan line; Along each scan line, the data points encountered are marked as the neighborhood points of the corresponding seed point. For each seed point, all the data points in the neighborhood formed by its neighborhood points are regarded as a cluster. The statistics of the covariate features of the data points in each cluster are calculated, and these statistics are constructed into a covariate vector.

5. The demand analysis and procurement method for new product project management according to claim 4, characterized in that: The method of defining the direction of the scan line includes: For each data point , calculate its density value in the multidimensional feature space and the density gradient vector ;in, is the preset step size; for each seed point s, obtain its density value and density gradient vector; Initially, the direction of the scan line is defined as the direction of the density gradient vector of the seed point s; recorded as the initial direction; the scan line is extended from the seed point s along the initial direction, and a step length is taken on the scan line , calculate the new data point at this step size The density gradient vector ; Adjust the direction of the scan line to The scan line direction is adjusted repeatedly to extend in the direction where the density increases fastest. The maximum length of the scan line is set to L_max. When the length of the scan line exceeds L_max or encounters an area where the density gradient vector is 0, the scan line is terminated. The next scan line is restarted from the seed point, and the direction is determined by the current density gradient vector.

6. The demand analysis and procurement method for new product project management according to claim 5, characterized in that: The method of constructing the demand structure relationship model includes: The data points in the demand cluster are regarded as the demand; for each demand cluster, a two-dimensional grid structure is constructed, and the two-dimensional grid structure contains a number of regular grids; For each demand in the demand cluster, a mapping function is used to map the demand to a grid point on the grid of the two-dimensional grid structure; a grid field is obtained; the scope of interaction is defined, and the interaction between the demands is calculated; the calculation formula for the interaction is: ;in, For demand and demand the interactions between them; For demand The priority of For demand The priority of For demand The cost, For demand Costs; and is the weight coefficient; Defining the Energy Field of the Lattice Field ; ;in Grid Field operators on ; is the quality parameter; is the coupling constant; is a differential operator, which represents the difference between adjacent grid points; for each grid point , calculate the local effective mass ; ;in, is the traversal index of the grid point, traversal and grid point Interacting grid points; At this time, the two-dimensional lattice structure integrating the local effective mass and energy field is used to construct the demand structure relationship model.

7. The method for demand analysis and procurement for new product project management according to claim 6, characterized in that: The method for obtaining the effective quality set includes: Initialize the local effective mass of all grid points, use the mean field approximation or other numerical methods to solve the energy field, obtain the field operator, and repeatedly calculate the local effective mass of each grid point based on the field operator until the preset number of iterations is reached; fix the local effective mass of each grid point at this time to form an effective mass set.

8. The demand analysis and procurement method for new product project management according to claim 7, characterized in that: The method for obtaining the optimal demand distribution pattern includes: Obtain the initial demand distribution pattern from the demand structure relationship model, that is, the initial mapping position of each demand on the grid; set the initial temperature T0, temperature drop rate a1, and end temperature Tend; define the objective function as the energy field and the weighted sum of the uniformity of demand distribution; The formula for demand distribution uniformity is: ;in, Grid The demand density on To meet the demand for uniform distribution, is the average value of the demand density of all grid points; is the number of rows in the two-dimensional lattice structure, is the number of columns of the two-dimensional grid structure; the demand density is the weighted sum of the number and priority of the demand at the corresponding grid point; Calculate the value of the objective function corresponding to the initial demand distribution pattern ; Set the current temperature , set an iteration number for each temperature; at the current temperature, iterate: For the demand structure relationship model, its neighborhood structure is defined as 8 neighborhoods; a fixed radius is preset, and for each grid point, the demand density within the fixed radius around it is calculated; the maximum neighborhood radius Rmax, the minimum neighborhood radius Rmin and the density threshold are set; for grid points with demand density less than or equal to the density threshold, its neighborhood radius is set to Rmin; for grid points with demand density greater than the density threshold, its neighborhood radius is set to Rmax; for each demand, the corresponding adaptive neighborhood is constructed according to the neighborhood radius of the grid point where it is located; Randomly select a demand, and randomly select a new location based on the adaptive neighborhood of the grid point where the demand is currently located; obtain a new distribution pattern; calculate the value of the objective function of the new distribution pattern , calculate the energy difference ; like If it is less than or equal to 0, the new distribution pattern is accepted and updated. ;like If greater than 0, then the probability Accept new distribution patterns; Repeat until the number of iterations at that temperature is reached; lower the temperature and repeat the iteration, lowering the temperature to the temperature drop rate a1 multiplied by the current temperature as the temperature for the next iteration; until the current temperature is less than Tend; at this time, the new distribution pattern finally output is the optimized demand distribution pattern.

9. The demand analysis and procurement method for new product project management according to claim 8, characterized in that: Methods for obtaining the optimal supplier combination include: Collect relevant information about potential suppliers, including supplier name, supplier type, supplier capabilities, supplier rating, and supplier cost; Determine the constraints for supplier selection based on demand and budget; formalize the constraints as hard constraints; build a model for evaluating supplier quality based on relevant information of potential suppliers; the output of the model is the supplier's quality score; Treat each demand in the optimal demand distribution pattern as a molecule; assign an initial quality value to each molecule, determined based on the priority and complexity index of the demand; construct an optimization function; the optimization function is the weighted sum of the supplier's quality score and the quality value of the molecule; Combine suppliers, assign a supplier to each molecule, obtain the supplier combination, and calculate the value of the optimization function under the current supplier combination; use the optimization algorithm to iterate the optimization of the supplier combination, stop the optimization iteration when the value of the optimization function converges, and output the supplier combination at this time, which is the optimal supplier combination.

10. A demand analysis and procurement system for new product project management, which is used to implement the demand analysis and procurement method for new product project management as claimed in any one of claims 1 to 9, characterized in that: include: Data collection module, used to collect M group project demand data; The preliminary analysis module is used to cluster the project demand data and obtain n demand clusters; A model building module is used to build a demand structure relationship model for each demand cluster and obtain a valid quality set based on the demand structure relationship model; Optimization module, used to optimize the demand structure relationship model and obtain the optimal demand distribution pattern; The supplier matching module is used to optimize the supplier selection according to the optimal demand distribution pattern to obtain the optimal supplier combination; The risk prediction module is used to use the optimal supplier combination as a procurement plan, evaluate procurement risks based on the effective quality set, and formulate risk response measures; each module is connected through wired and / or wireless means to achieve data transmission between modules.

Images