Enterprise intelligent data analysis decision support system and method based on big data

Abstract

The invention discloses an enterprise intelligent data analysis decision support system and method based on big data, and relates to the technical field of enterprise data analysis and decision support, and the method comprises the steps: carrying out the clustering analysis of product sales volume data according to a fuzzy K-Prototypes algorithm; the algorithm is improved based on a particle swarm optimization algorithm, and a fitness function of the improved algorithm is calculated; the algorithm is optimized based on ReliefF, and clustering internal divergence is obtained; analyzing the improved clustering effect according to the fitness function and the clustering internal divergence; according to the clustering analysis of the enterprise product sales volume data, making a product resource decision scheme, predicting the cost of executing the decision scheme, and calculating the difference between the minimum prediction cost and the actual cost; and comparing and analyzing the prediction result of the model and the prediction cost of the experience decision scheme, and evaluating the performance of the decision scheme. The problem that enterprise product resource allocation is unreasonable is solved, and support can be provided for intelligent management and optimization of enterprises.

Description

Technical Field

[0001] The present invention relates to the technical field of enterprise data analysis and decision support, and more specifically, to an enterprise intelligent data analysis decision support system and method based on big data. Background Art

[0002] With the rapid development of information technology, enterprises have accumulated a large amount of business data in their daily operations, including sales data, customer data, supply chain data, market data, etc. If these data are not effectively analyzed and used, it will be difficult to provide strong support for decision-making. Most of the existing enterprise decision support systems (DSS) rely on traditional statistical methods or rule engines and face the following problems: limited data processing capabilities and inability to analyze large-scale data in real time; inability to handle the diversity and complexity of data, resulting in inaccurate analysis results; poor adaptability to user needs during the analysis process and lack of personalized decision support; the system lacks the mining of deep-level correlations between data, resulting in an inability to provide comprehensive decision-making basis.

[0003] Deficiencies of existing technologies:

[0004] Existing decision support systems often provide basic reporting and analysis functions, but for the needs of the big data era, there is still a lack of sufficient intelligent analysis tools. Many companies cannot make full use of the data in the system for in-depth analysis and cannot effectively support data-driven decision-making; although the application of artificial intelligence (AI) technology is becoming more and more important in modern enterprises, many traditional ERP systems have not been well integrated with AI, machine learning and other technologies, resulting in their limited capabilities in automation, predictive analysis, decision support, etc. The present invention solves the problem of unreasonable product resource allocation in enterprises by optimizing clustering algorithm analysis and formulating decision-making plans. It has important practical significance for enterprises to optimize resource allocation and improve operational efficiency, and can provide support for enterprises to achieve intelligent management and optimization.

[0005] In view of the above problems, the present invention proposes a solution. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide an enterprise intelligent data analysis decision support system and method based on big data, which solves the problems raised in the above-mentioned background technology by evaluating the enterprise decision-making performance.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The enterprise intelligent data analysis decision support system based on big data includes: a collection module, which is used to collect enterprise product sales data and perform cluster analysis on the product sales data according to the fuzzy K-Prototypes algorithm; an improvement module, which is used to improve the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm, use the solution obtained based on the particle swarm optimization algorithm as the initial value of the fuzzy K-Prototypes algorithm, and calculate the fitness function of the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm;

[0009] The optimization module is used to further optimize the fuzzy K-Prototypes algorithm based on ReliefF, calculate the feature weights, use the feature weights to weight the features of each dimension to obtain the weight-optimized data, and input the weight-optimized data into the fuzzy K-Prototypes algorithm for clustering to obtain the intra-cluster divergence; the analysis module is used to analyze the clustering effect of the improved fuzzy K-Prototypes algorithm based on the fitness function and the intra-cluster divergence; the prediction module is used to formulate product resource decision plans based on the cluster analysis of the enterprise product sales data, and to build a convolutional neural network model to predict the cost of executing the decision plan and calculate the difference between the minimized predicted cost and the actual cost; the evaluation module is used to compare and analyze the prediction results of the model with the predicted cost of the empirical decision plan, and evaluate the performance of the decision plan based on the difference between the predicted cost value and the actual cost value.

[0010] In a preferred embodiment, the specific steps of clustering the product sales data according to the fuzzy K-Prototypes algorithm are as follows: randomly select K initial cluster centers, including the mean of numerical features and the mode of categorical features; calculate the membership of each data point in the cluster to the cluster center, for numerical features, use Euclidean distance to calculate the distance from each data point to the cluster center, and calculate the membership of the numerical feature according to the membership function; for categorical features, use matching degree to measure the similarity between the data point and the cluster center, if the category of the data point is the same as the category of the cluster center, the matching degree is 1; otherwise, it is 0; weighted average the matching degrees of each categorical feature to obtain the membership of the categorical feature;

[0011] The cluster centers are updated weightedly according to the membership degree. The cluster centers of numerical features are updated by weighted average, and the cluster centers of categorical features are updated by weighted majority. It is determined whether the cluster centers converge. If so, the iteration is stopped. Otherwise, the membership degree is returned to be calculated until the cluster centers converge or the maximum number of iterations is reached.

[0012] In a preferred embodiment, the specific steps of improving the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm are as follows: training the initial cluster center based on the particle swarm optimization algorithm, taking the position vector of the particle swarm as the set of cluster centers, each particle represents a candidate solution of the cluster center, and adjusting the cluster center position by repeated iterations to minimize the fitness function value; taking the solution obtained based on the PSO algorithm as the initial value of the FKP algorithm, calculating the distance from the sample to each cluster center, using the Euclidean distance measurement for numerical data and the matching measurement for categorical data, and calculating the membership of each sample to each cluster based on the distance from the sample to each cluster center; calculating the new center position of each cluster based on the membership, updating the cluster center, and recalculating the fitness function value until the change in the fitness function is less than a threshold or the maximum number of iterations is reached.

[0013] In a preferred embodiment, the specific steps for obtaining the intra-cluster divergence are as follows: randomly select a sample x i ; Find the i The nearest neighbor instance of the same category is called a “neighbor positive example” x i +; find the value that matches x i Nearest neighbor instances of different categories, called “neighbor counterexamples” x i -; For each feature, by comparing x i The difference in features between the nearest positive example and the nearest negative example is used to calculate the weight of the feature, and each feature is weighted using the obtained weight to obtain weight-optimized data. The weighted data becomes a new feature space; the weight-optimized data is input into the fuzzy K-Prototypes algorithm for clustering, and the intra-cluster divergence is obtained by calculating the average of the weighted distances between each sample. If the intra-cluster divergence after weighting is low, it means that the weighting makes the samples in the cluster closer and the clustering effect is optimized.

[0014] In a preferred embodiment, the steps of analyzing the clustering effect of the improved fuzzy K-Prototypes algorithm based on the fitness function and the intra-cluster divergence are as follows: obtaining the fitness function and the intra-cluster divergence; calculating the clustering optimization coefficient according to the fitness function and the intra-cluster divergence weighted, comparing the clustering optimization coefficient with the historical clustering optimization coefficient, and analyzing the clustering effect of the improved fuzzy K-Prototypes algorithm; if the clustering optimization coefficient is greater than the historical clustering optimization coefficient, the clustering optimization effect is not good and needs to be optimized again; if the clustering optimization coefficient is less than the historical clustering optimization coefficient, the clustering optimization effect is good and does not need to be optimized again.

[0015] In a preferred embodiment, the process of obtaining the minimized prediction cost is as follows: the product sales data is passed into the model, the spatial features of the data are extracted through convolution operations, and important information is retained; the spatial features extracted by the convolution layer are mapped to the decision layer, and the cost prediction of the decision execution is output; the model is trained using the training set data, the hyperparameters are adjusted, and the model parameters are adjusted on the validation set to minimize the prediction cost.

[0016] In a preferred embodiment, the specific steps of evaluating the performance of the decision-making scheme are as follows: defining the mean square error between the model prediction cost and the actual cost and the determination coefficient between the model prediction cost and the empirical prediction cost as input variables, and dividing them into different fuzzy sets respectively; defining the performance of the decision-making scheme as an output variable, and dividing it into fuzzy sets; formulating fuzzy rules to describe the influence of the mean square error between the model prediction cost and the actual cost and the determination coefficient between the model prediction cost and the empirical prediction cost on the performance of the decision-making scheme; performing fuzzy reasoning according to the fuzzy rules to determine the performance of the decision-making scheme.

[0017] The enterprise intelligent data analysis decision support method based on big data includes: collecting enterprise product sales data, performing cluster analysis on the product sales data according to the fuzzy K-Prototypes algorithm; improving the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm, using the solution obtained based on the particle swarm optimization algorithm as the initial value of the fuzzy K-Prototypes algorithm, and calculating the fitness function of the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm; further optimizing the fuzzy K-Prototypes algorithm based on ReliefF, calculating the feature weights, and using the feature weights to optimize each dimension. The weighted features are weighted to obtain weight-optimized data, and the weight-optimized data is input into the fuzzy K-Prototypes algorithm for clustering to obtain the intra-cluster divergence; the clustering effect of the improved fuzzy K-Prototypes algorithm is analyzed according to the fitness function and the intra-cluster divergence; according to the clustering analysis of the enterprise product sales data, a product resource decision plan is formulated, and a convolutional neural network model is constructed to predict the cost of executing the decision plan, and calculate the difference between the minimized predicted cost and the actual cost; the prediction results of the model are compared and analyzed with the predicted cost of the empirical decision plan, and the performance of the decision plan is evaluated based on the difference between the predicted cost value and the actual cost value. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a schematic diagram of the structure of the enterprise intelligent data analysis decision support system based on big data of the present invention.

[0019] Figure 2 It is a structural schematic diagram of the enterprise intelligent data analysis decision support method based on big data of the present invention. DETAILED DESCRIPTION

[0020] 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.

[0021] Embodiment 1, Figure 1 The present invention provides an enterprise intelligent data analysis decision support system based on big data.

[0022] The acquisition module is used to collect the sales data of enterprise products and perform cluster analysis on the product sales data based on the fuzzy K-Prototypes algorithm;

[0023] The fuzzy K-Prototypes algorithm is a technique that combines the K-Means and K-Modes clustering algorithms. It is suitable for situations that contain both numerical and categorical data. It not only considers the Euclidean distance of numerical data (such as K-Means), but also combines the similarity calculation of categorical data (such as K-Modes).

[0024] The enterprise product sales data includes numerical features such as sales volume and price, and categorical features such as product type and region;

[0025] The fuzzy K-Prototypes algorithm is an improvement on K-Means and K-Modes, and is suitable for processing both numerical and categorical data. Unlike the traditional K-Means clustering algorithm, the fuzzy K-Prototypes algorithm uses membership to process the degree to which a data point belongs to a cluster, rather than rigid classification;

[0026] The specific steps of clustering analysis of product sales data based on the fuzzy K-Prototypes algorithm are as follows:

[0027] Initialize cluster centers: randomly select K initial cluster centers, which include the mean of numerical features and the mode of categorical features;

[0028] Calculate membership: For each data point, calculate its membership to each cluster center. For numerical features, use Euclidean distance to calculate the distance from each data point to the cluster center. Calculate the membership of the numerical feature based on the membership function.

[0029] For categorical features, the matching degree is used to measure the similarity between the data point and the cluster center. If the category of the data point is the same as that of the cluster center, the matching degree is 1; otherwise, it is 0. The membership degree of the categorical feature is obtained by weighted average of the matching degrees of each categorical feature.

[0030] Update cluster centers: Update cluster centers by weight according to membership. Cluster centers of numerical features are updated by weighted average.

[0031] For categorical features, the cluster center is the most frequently occurring category in the feature, and categorical features are updated by weighted mode;

[0032] Iterative update: Determine whether the cluster center converges. If so, stop the iteration; otherwise, return to calculate the membership until the cluster center converges or the maximum number of iterations is reached.

[0033] An improvement module is used to improve the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm, use the solution obtained based on the particle swarm optimization algorithm as the initial value of the fuzzy K-Prototypes algorithm, and calculate the fitness function of the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm;

[0034] Particle swarm optimization (PSO) is a group intelligence optimization method, whose goal is to find the global optimal solution to the problem by simulating the foraging process of bird flocks. In the fuzzy K-Prototypes algorithm, the role of PSO is mainly reflected in the following aspects:

[0035] Selection of initial cluster centers: PSO can be used to optimize the location of cluster centers in the fuzzy K-Prototypes algorithm to avoid the local optimal problem that may be caused by random initialization in K-Means.

[0036] Optimizing the objective function: PSO can improve the quality of clustering results by adjusting the positions of cluster centers and minimizing the objective function of the fuzzy K-Prototypes algorithm.

[0037] The specific steps of improving the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm are as follows:

[0038] In the process of improving the FKP algorithm by using the particle swarm algorithm, the PSO algorithm is first used to train the initial cluster center, and the position vector of the particle swarm is used as the set of cluster centers. Each particle represents a candidate solution of the cluster center. The PSO adjusts the position of the cluster center through repeated iterations to minimize the fitness function value.

[0039] Although the PSO algorithm can find an approximate solution close to the global solution, it cannot absolutely guarantee convergence to the global optimal solution. Therefore, the solution obtained based on the PSO algorithm is used as the initial value of the FKP algorithm, and then the FKP algorithm is used to further find the global optimal solution;

[0040] Calculate the distance from the sample to the center of each cluster. For numerical data, use the Euclidean distance metric. For categorical data, use the matching metric (such as the Hamming distance). And calculate the membership of each sample to each cluster based on the distance from the sample to the center of each cluster.

[0041] The new center position of each cluster is calculated based on the membership degree, the cluster center is updated, and the fitness function value is recalculated until the stopping condition is reached (for example, the change in the fitness function is less than a certain threshold or the maximum number of iterations is reached;

[0042] According to the distance from the sample to each cluster center and the sample's membership to the cluster, the fitness function of the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm is calculated. The specific calculation formula is as follows:

[0043] Where J is the fitness function, d n (x i , C k ) is the sample x i Numerical data distance to cluster k, d c (x i , C k ) is the sample x i The categorical data distance to cluster k, u ik is the membership of sample i to cluster k, α and β are weighting coefficients for numerical and categorical data, which are used to adjust the weights of the two distance metrics.

[0044] The optimization module is used to further optimize the fuzzy K-Prototypes algorithm based on ReliefF, calculate the feature weights, use the feature weights to weight the features of each dimension to obtain weight-optimized data, and input the weight-optimized data into the fuzzy K-Prototypes algorithm for clustering to obtain the intra-cluster divergence;

[0045] ReliefF is a feature selection algorithm that assigns weights to each feature by evaluating its importance in classification. It mainly calculates feature weights by comparing the differences between instances. Features with higher weights are more important for distinguishing samples of different categories.

[0046] The specific steps for further optimizing the fuzzy K-Prototypes algorithm based on ReliefF are as follows:

[0047] Randomly select a sample x i;

[0048] Find the value with x i The nearest neighbor instance of the same category is called a “neighbor positive example” x i +;

[0049] Find the value with x i Nearest neighbor instances of different categories, called “neighbor counterexamples” x i -;

[0050] For each feature, by comparing x i The difference in features between the nearest positive example and the nearest negative example is used to calculate the weight of the feature, and each feature is weighted using the obtained weight to obtain weight-optimized data. The weighted data becomes the new feature space;

[0051] The weighted optimized data is input into the fuzzy K-Prototypes algorithm for clustering, and the intra-cluster divergence is obtained by calculating the average value of the weighted distance between each sample to measure the similarity between samples; the specific calculation formula of the intra-cluster divergence is as follows: ;W represents the intra-cluster divergence, |C k | is cluster C k The number of samples, w(x i ) is the sample x i The weight, w(x j ) is the sample x j The weight of d(x i , x j ) is the sample x i and x j The distance between.

[0052] If the intra-cluster divergence after weighting is low, it means that weighting makes the samples within the cluster more compact and the clustering effect is optimized.

[0053] The analysis module is used to analyze the clustering effect of the improved fuzzy K-Prototypes algorithm based on the fitness function and the intra-cluster divergence;

[0054] The specific steps include:

[0055] Step C1, obtaining the fitness function J and the intra-cluster divergence W;

[0056] Step C2, weighted calculation of cluster optimization coefficient G based on fitness function J and intra-cluster divergence W,

[0057] Step C3, comparing the clustering optimization coefficient with the historical clustering optimization coefficient, and analyzing the clustering effect after the improvement of the fuzzy K-Prototypes algorithm.

[0058] Furthermore, the clustering optimization coefficient G can be calculated as: G = a1J + a2W;

[0059] Where a1 and a2 are preset proportional coefficients of the fitness function J and the intra-cluster divergence W, respectively, which can be calculated and obtained by the hierarchical analysis method or the priority diagram method according to the actual situation, and will not be elaborated here.

[0060] At the same time, it should be noted that the fitness function J, the intra-cluster divergence W, and the corresponding cluster optimization coefficient are stored in the database. After calculating the current cluster optimization coefficient, the value is compared with the historical database to analyze the cluster optimization effect;

[0061] If the clustering optimization coefficient is greater than the historical clustering optimization coefficient, the clustering optimization effect is not good and needs to be optimized again. If the clustering optimization coefficient is less than the historical clustering optimization coefficient, the clustering optimization effect is good and does not need to be optimized again. The subsequent steps can be carried out.

[0062] The prediction module is used to formulate product resource decision plans based on cluster analysis of enterprise product sales data, and to build a convolutional neural network model to predict the cost of executing the decision plan and calculate the difference between the minimized predicted cost and the actual cost;

[0063] Based on the clustering results, enterprises can make specific resource decision plans, such as:

[0064] High-selling products: Increase production, increase advertising investment, optimize supply chain, etc. to ensure that these products meet market demand;

[0065] Medium-selling products: Conduct market research, analyze the reasons for sales changes, and consider appropriate promotions or adjust product strategies;

[0066] Low-volume products: may need to be withdrawn from the market, discounted or have their product features adjusted;

[0067] To predict the cost of executing a decision plan, we can build a convolutional neural network model:

[0068] Input: including historical sales data, product category information, and marketing activity data;

[0069] Output: Cost forecast of decision execution;

[0070] The product sales data and marketing activity information are passed into the model, and the spatial features of the data are extracted through convolution operations. The dimensions of the data are reduced through the pooling layer to improve the computational efficiency of the model and retain important information. The features extracted by the convolution layer are mapped to the predicted value of the decision execution cost, and the cost prediction of the decision execution is output;

[0071] Use the training set data to train the model, adjust the hyperparameters, and adjust the model parameters on the validation set to prevent overfitting and minimize the prediction cost;

[0072] The specific process of calculating the minimum difference between the predicted cost and the actual cost is as follows:

[0073] The difference between the predicted cost and the actual cost can be calculated using a loss function. The most commonly used loss function is the mean squared error (MSE), which is calculated as follows: Where MSE is the mean square error, n is the number of samples, and y pi is the cost of the i-th sample predicted by the model, y ti is the actual cost of the i-th sample;

[0074] The evaluation module is used to compare and analyze the prediction results of the model with the predicted costs of the empirical decision-making scheme, and to evaluate the performance of the decision-making scheme based on the difference between the predicted cost value and the actual cost value.

[0075] Compare and analyze the prediction results of the model with the prediction costs of the empirical decision-making scheme, and compare and analyze the prediction costs of the model with the empirical prediction costs, and calculate the determination coefficient between the model prediction costs and the empirical prediction costs. The value range is 0 to 1. The closer to 1, the closer the model prediction results are to the prediction costs of the empirical decision-making scheme. The specific calculation formula for the determination coefficient between the model prediction costs and the empirical prediction costs is as follows: ; In the formula, R 2 is the coefficient of determination, y pi is the cost of the i-th sample predicted by the model, y di is the cost of the ith sample predicted by experience, y d The average of the empirically predicted costs;

[0076] The specific steps to evaluate the performance of a decision solution are as follows:

[0077] In step C1, the mean square error between the model prediction cost and the actual cost and the determination coefficient between the model prediction cost and the empirical prediction cost are defined as input variables, and they are divided into different fuzzy sets respectively.

[0078] For example, "Low", "Medium", and "High" are for the mean square error between the model-predicted cost and the actual cost, and "Low", "Medium", and "High" are for the coefficient of determination between the model-predicted cost and the empirically predicted cost.

[0079] Step C2, defines the performance of the decision solution as an output variable and divides it into fuzzy sets, for example, "Low", "High", for the performance of the decision solution.

[0080] Step C3, formulate a set of fuzzy rules to describe the impact of different input variables on output variables. The definition of rules can be based on professional knowledge or obtained through data analysis and experiments. For example:

[0081] The mean square error between the model prediction cost and the actual cost is marked as MSE, and the determination coefficient between the model prediction cost and the empirical prediction cost is marked as R 2 , the performance of the decision solution is marked as D, then we can define

[0082] Rule 1: IF(MSE is High)AND(R 2 is Low) THEN(D is Low)

[0083] Rule 2: IF(MSE is Low) AND(R 2 is High) THEN(D is High) ...

[0085] Step C4, fuzzy reasoning is performed according to fuzzy rules to determine the performance of the decision-making solution.

[0086] It should be noted that the division of fuzzy sets can be adjusted according to actual conditions. For example, although this embodiment takes three fuzzy sets as examples, in fact, the mean square error between the model prediction cost and the actual cost, the determination coefficient between the model prediction cost and the empirical prediction cost, and the performance of the decision-making scheme can be divided into more than three sets to facilitate better precise adjustment.

[0087] Furthermore, for the performance judgment of the intelligent air-conditioning operation and maintenance management system, a threshold can be set according to the actual situation. For example, when the mean square error between the model prediction cost and the actual cost exceeds 100, it is marked as "High"; when the determination coefficient between the model prediction cost and the empirical prediction cost is higher than 0.8, it is marked as "High", and so on. I will not go into details here.

[0088] Embodiment 2, Figure 2 It is a structural schematic diagram of the enterprise intelligent data analysis decision support method based on big data of the present invention.

[0089] Enterprise intelligent data analysis decision support methods based on big data include:

[0090] Collect enterprise product sales data and perform cluster analysis on the product sales data based on the fuzzy K-Prototypes algorithm;

[0091] The fuzzy K-Prototypes algorithm is improved based on the particle swarm optimization algorithm, the solution obtained based on the particle swarm optimization algorithm is used as the initial value of the fuzzy K-Prototypes algorithm, and the fitness function of the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm is calculated;

[0092] The fuzzy K-Prototypes algorithm is further optimized based on ReliefF, and the feature weights are calculated. The feature weights are used to weight the features of each dimension to obtain weight-optimized data. The weight-optimized data is input into the fuzzy K-Prototypes algorithm for clustering to obtain the intra-cluster divergence.

[0093] The clustering effect of the improved fuzzy K-Prototypes algorithm is analyzed based on the fitness function and intra-cluster divergence;

[0094] Based on the cluster analysis of the company's product sales data, a product resource decision plan is formulated, and a convolutional neural network model is constructed to predict the cost of executing the decision plan and calculate the difference between the predicted cost and the actual cost to minimize it;

[0095] The prediction results of the model are compared and analyzed with the predicted costs of the empirical decision-making scheme, and the performance of the decision-making scheme is evaluated by combining the difference between the predicted cost value and the actual cost value.

[0096] 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 in the formula are set by technicians in this field according to actual conditions.

[0097] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0098] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0099] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0100] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0101] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Enterprise intelligent data analysis decision support system based on big data, characterized by: include: The acquisition module is used to collect the sales data of enterprise products and perform cluster analysis on the product sales data based on the fuzzy K-Prototypes algorithm; An improvement module is used to improve the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm, use the solution obtained based on the particle swarm optimization algorithm as the initial value of the fuzzy K-Prototypes algorithm, and calculate the fitness function of the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm; The optimization module is used to further optimize the fuzzy K-Prototypes algorithm based on ReliefF, calculate the feature weights, use the feature weights to weight the features of each dimension to obtain weight-optimized data, and input the weight-optimized data into the fuzzy K-Prototypes algorithm for clustering to obtain the intra-cluster divergence; The analysis module is used to analyze the clustering effect of the improved fuzzy K-Prototypes algorithm based on the fitness function and the intra-cluster divergence; The prediction module is used to formulate product resource decision plans based on cluster analysis of enterprise product sales data, and to build a convolutional neural network model to predict the cost of executing the decision plan and calculate the difference between the minimized predicted cost and the actual cost; The evaluation module is used to compare and analyze the prediction results of the model with the predicted costs of the empirical decision-making scheme, and to evaluate the performance of the decision-making scheme based on the difference between the predicted cost value and the actual cost value.

2. The enterprise intelligent data analysis decision support system based on big data according to claim 1 is characterized in that: The specific steps of clustering analysis of product sales data according to the fuzzy K-Prototypes algorithm are as follows: Randomly select K initial cluster centers, including the mean of numerical features and the mode of categorical features; Calculate the membership of each data point in the cluster to the cluster center. For numerical features, use the Euclidean distance to calculate the distance from each data point to the cluster center, and calculate the membership of the numerical feature based on the membership function. For categorical features, the matching degree is used to measure the similarity between the data point and the cluster center. If the category of the data point is the same as that of the cluster center, the matching degree is 1; otherwise, it is 0. The membership degree of the categorical feature is obtained by weighted average of the matching degrees of each categorical feature. The cluster centers are updated weightedly according to the membership degree. The cluster centers of numerical features are updated by weighted average, and the cluster centers of categorical features are updated by weighted majority. Determine whether the cluster center converges. If so, stop the iteration; otherwise, return to calculating the membership until the cluster center converges or the maximum number of iterations is reached.

3. The enterprise intelligent data analysis decision support system based on big data according to claim 2 is characterized in that: The specific steps of improving the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm are as follows: The initial cluster center is trained based on the particle swarm optimization algorithm. The position vector of the particle swarm is used as the set of cluster centers. Each particle represents a candidate solution for a cluster center. The position of the cluster center is adjusted by repeated iterations to minimize the fitness function value. The solution obtained based on the PSO algorithm is used as the initial value of the FKP algorithm to calculate the distance from the sample to the center of each cluster. The numerical data is measured by the Euclidean distance, and the categorical data is measured by the matching measure. The membership of each sample to each cluster is calculated based on the distance from the sample to the center of each cluster. The new center position of each cluster is calculated according to the membership degree, the cluster center is updated, and the fitness function value is recalculated until the change in the fitness function is less than the threshold or the maximum number of iterations is reached.

4. The enterprise intelligent data analysis decision support system based on big data according to claim 3 is characterized in that: The specific steps for obtaining the intra-cluster divergence are as follows: Randomly select a sample x i ; Find the value with x i The nearest neighbor instance of the same category is called a "neighbor positive example" x i +; Find the value with x i The nearest neighbor instances of different categories are called "neighbor counterexamples" x i -; For each feature, by comparing x i The difference in features between the nearest positive example and the nearest negative example is used to calculate the feature weights, and each feature is weighted using the obtained weights to obtain weight-optimized data. The weighted data becomes the new feature space. The weighted optimized data is input into the fuzzy K-Prototypes algorithm for clustering, and the intra-cluster divergence is obtained by calculating the average of the weighted distances between each sample. If the intra-cluster divergence after weighting is low, it means that the weighting makes the samples in the cluster closer and the clustering effect is optimized.

5. The enterprise intelligent data analysis decision support system based on big data according to claim 4 is characterized in that: The steps of analyzing the clustering effect of the improved fuzzy K-Prototypes algorithm based on the fitness function and the intra-cluster divergence are as follows: Get the fitness function and intra-cluster divergence; The clustering optimization coefficient is calculated based on the fitness function and the weighted divergence within the cluster. The clustering optimization coefficient is compared with the historical clustering optimization coefficient, and the clustering effect after the improvement of the fuzzy K-Prototypes algorithm is analyzed; If the clustering optimization coefficient is greater than the historical clustering optimization coefficient, the clustering optimization effect is not good and needs to be optimized again. If the clustering optimization coefficient is less than the historical clustering optimization coefficient, the clustering optimization effect is good and does not need to be optimized again.

6. The enterprise intelligent data analysis decision support system based on big data according to claim 5 is characterized in that: The process of obtaining the minimized prediction cost is as follows: The product sales data is passed into the model, and the spatial features of the data are extracted through convolution operations, while retaining important information. The spatial features extracted by the convolutional layer are mapped to the decision layer, and the cost prediction of decision execution is output; Use the training set data to train the model, adjust the hyperparameters, and adjust the model parameters on the validation set to minimize the prediction cost.

7. The enterprise intelligent data analysis decision support system based on big data according to claim 6 is characterized in that: The specific steps of evaluating the performance of the decision-making scheme are as follows: The mean square error between the model prediction cost and the actual cost and the determination coefficient between the model prediction cost and the empirical prediction cost are defined as input variables, and they are divided into different fuzzy sets respectively; The performance of the decision-making scheme is defined as the output variable, which is divided into fuzzy sets; Formulate fuzzy rules to describe the impact of the mean square error between the model predicted cost and the actual cost and the coefficient of determination between the model predicted cost and the empirical predicted cost on the performance of the decision-making scheme; Perform fuzzy reasoning based on fuzzy rules to determine the performance of the decision-making solution.

8. The enterprise intelligent data analysis decision support system based on big data according to claim 7 is characterized in that: The specific calculation formula of the fitness function of the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm is as follows: Where J is the fitness function, d n (x i , C k ) is the sample x i Numerical distance to cluster k, d c (x i , C k ) is the sample x i The categorical data distance to cluster k, u ik is the membership of sample i to cluster k, α and β are weighting coefficients for numerical and categorical data, which are used to adjust the weights of the two distance metrics.

9. The enterprise intelligent data analysis decision support system based on big data according to claim 8 is characterized in that: The specific calculation formula of the intra-cluster divergence is as follows: Where W represents the intra-cluster divergence, |C k | is cluster C k The number of samples, w(x i ) is the sample x i The weight, w(x j ) is the sample x j The weight of d(x i , x j ) is the sample x i and x j The distance between.

10. An enterprise intelligent data analysis decision support method based on big data, characterized in that: include: Collect enterprise product sales data and perform cluster analysis on the product sales data based on the fuzzy K-Prototypes algorithm; The fuzzy K-Prototypes algorithm is improved based on the particle swarm optimization algorithm, the solution obtained based on the particle swarm optimization algorithm is used as the initial value of the fuzzy K-Prototypes algorithm, and the fitness function of the fuzzy K-Prototypes algorithm based on the particle swarm optimization algorithm is calculated; The fuzzy K-Prototypes algorithm is further optimized based on ReliefF, and the feature weights are calculated. The feature weights are used to weight the features of each dimension to obtain weight-optimized data. The weight-optimized data is input into the fuzzy K-Prototypes algorithm for clustering to obtain the intra-cluster divergence. The clustering effect of the improved fuzzy K-Prototypes algorithm is analyzed based on the fitness function and intra-cluster divergence; Based on the cluster analysis of the company's product sales data, a product resource decision plan is formulated, and a convolutional neural network model is constructed to predict the cost of executing the decision plan and calculate the difference between the predicted cost and the actual cost to minimize it; The prediction results of the model are compared and analyzed with the predicted costs of the empirical decision-making scheme, and the performance of the decision-making scheme is evaluated by combining the difference between the predicted cost value and the actual cost value.

Images