A business expansion project clustering method, device, equipment and storage medium based on SOM-M

Abstract

The present invention relates to a business expansion project clustering method based on SOM-M, comprising: obtaining original business expansion project data, the original business expansion project data including: project nodes, node planned time, and node budget costs; inputting the pre-processed original business expansion project data into the SOM-M model, initializing the neuron parameters of the competition layer; calculating the similarity and first-order momentum between the neurons of the input layer and the neurons of the competition layer, and updating the neuron parameters of the competition layer; and obtaining the node category of the business expansion project based on the updated neuron parameters of the competition layer. The present application measures the similarity of multiple neurons, makes the similarities of multiple neurons compete with each other, and continuously adjusts the category center of the neuron approximation object, so that the neurons automatically cluster. After adding the first-order momentum, the convergence ability of the SOM is improved, and it is easier to find the cluster center. It has strong classification ability and fast learning ability, and has a good effect in clustering business expansion projects.

Description

Technical Field

[0001] The present invention relates to the technical field of business expansion project data anomalies, and in particular to a business expansion project clustering method, device, equipment and storage medium based on SOM-M. Background Art

[0002] Due to various reasons, the data source of business expansion supporting projects often leads to abnormalities in a small amount of data. How to accurately identify abnormal data is the key to the success of clustering business expansion supporting projects and predicting the working hours of each node in the business expansion process.

[0003] To address the problem of anomaly detection in electricity consumption data, domestic and international researchers have proposed numerous methods and achieved substantial research results. These methods include statistical methods and clustering methods. From a data processing perspective, these methods can be categorized into supervised learning and unsupervised learning. Supervised learning requires the label of the dataset, that is, whether a piece of data is anomaly. In practical projects, normal data and abnormal data are distributed in the same dataset, and there is insufficient information to distinguish them. Summary of the Invention

[0004] Based on this, the present invention provides a method, apparatus, device, and storage medium for clustering business expansion projects based on SOM-M. This application focuses on anomaly detection under unsupervised learning. The SOM-M model is a combination-based unsupervised fast anomaly detection method with linear time complexity and high accuracy. It only requires a small number of samples to achieve good anomaly detection results and has the advantage of fast convergence.

[0005] According to a first aspect of some embodiments of the present application, the present application provides a business expansion project clustering method based on SOM-M, the method comprising the following steps:

[0006] Obtaining original business expansion project data, wherein the original business expansion project data includes: project nodes, node planned time, and node budget cost;

[0007] Preprocessing the original business expansion project data to remove abnormal data;

[0008] Inputting the preprocessed original business expansion project data into the SOM-M model and initializing the neuron parameters of the competition layer, wherein the SOM-M model includes an input layer and a competition layer;

[0009] Calculating the similarity and first-order momentum between the neurons in the input layer and the neurons in the competition layer, and updating the neuron parameters in the competition layer;

[0010] Obtaining a node category of the business expansion project according to the updated neuron parameters of the competitive layer;

[0011] The nodes are labeled according to the node categories.

[0012] Furthermore, the pre-processed original business expansion project data is input into the SOM-M model to initialize the competitive layer neuron parameters, including:

[0013] The node planned time and node budget cost of each of the project nodes are used as neuron parameters of the input layer;

[0014] Determining the number of neurons in the competitive layer and the size of the competitive layer;

[0015] The neuron parameters of the competitive layer are randomly initialized.

[0016] Further, calculating the similarity and first-order momentum between the neurons of the input layer and the neurons of the competitive layer, and updating the neuron parameters of the competitive layer, includes:

[0017] Select a neuron in the input layer and calculate the similarity between each neuron in the competition layer and the neuron. The Euclidean distance calculation formula of the similarity is as follows:

[0018]

[0019] Among them, X is the neuron parameter vector of the input layer, X i is the neuron parameter vector of the i-th competitive layer, and T is the vector transposition operator;

[0020] Comparing similarity calculation results of the neurons in the competition layer, and selecting the neuron with the smallest Euclidean distance as the winning neuron;

[0021] Calculate the winning neuron parameter update adjustment value m with first-order momentum. The first-order momentum formula is as follows:

[0022] m v (c) = β·m v (c-1)+(1-β)·(D(t)-W v (e))

[0023] Among them, W v is the parameter vector of the winning neuron v, D is the parameter vector of the winning neuron, m v (c) is the parameter update adjustment value with first-order momentum of the winning neuron, m v (c-1) is the last parameter update adjustment value of the winning neuron with first-order momentum, v represents the winning neuron, and β is the first-order momentum hyperparameter;

[0024] Update the parameters of the winning neuron and its neighboring neurons. The parameter update formula is as follows:

[0025] W v (e+1)=W v (e)+θ(u,v,σ)×α(e)×m

[0026]

[0027] Among them, W v is the parameter vector of neuron v, e is the current iteration number, θ(u,v,σ) is the nearest neighbor function, which can be a bubble function or a Gaussian function, u is the x-axis coordinate of the winning neuron on the competition plane, v is the y-axis coordinate of the winning neuron on the competition plane, E is the total number of iterations / 2, α is the learning rate, and D is the parameter vector of the winning neuron;

[0028] The neuron parameters of the competitive layer are updated repeatedly until the set maximum number of iterations is reached.

[0029] Furthermore, according to the updated neuron parameters of the competitive layer, the node categories of the business expansion project are obtained, including:

[0030] The Euclidean distance between the input layer heating neuron parameter vector and the competition layer neuron parameter vector is calculated respectively, and each input layer neuron is classified into the competition layer neuron with the smallest Euclidean distance to obtain the node category of the business expansion project.

[0031] Furthermore, labeling the node according to the node category includes:

[0032] While labeling the classified competition layer neurons, the input layer neurons classified into the same competition layer neurons are labeled with the same label.

[0033] According to a second aspect of the present application, a device for clustering business expansion projects based on SOM-M is provided, comprising:

[0034] An original business expansion project data acquisition module is used to acquire original business expansion project data, wherein the original business expansion project data includes: project nodes, node planned time, and node budget cost;

[0035] A data preprocessing module, used to preprocess the original business expansion project data and remove abnormal data;

[0036] A parameter initialization module, configured to input the preprocessed raw business expansion project data into the SOM-M model and initialize the neuron parameters of the competitive layer, wherein the SOM-M model includes an input layer and a competitive layer;

[0037] A parameter updating module, configured to calculate the similarity and first-order momentum between the neurons in the input layer and the neurons in the competitive layer, and update the neuron parameters in the competitive layer;

[0038] A node category determination module, configured to obtain the node category of the business expansion project based on the updated neuron parameters of the competitive layer;

[0039] The label module is used to label the node according to the node category.

[0040] Furthermore, the parameter initialization module includes:

[0041] An input parameter determination unit, configured to use the node planned time and node budget cost of each of the project nodes as neuron parameters of the input layer;

[0042] a competition parameter determination unit, configured to determine the number of neurons in the competition layer and the size of the competition layer;

[0043] An initialization unit is used to randomly initialize the neuron parameters of the competitive layer.

[0044] Furthermore, the parameter updating module includes:

[0045] The similarity calculation unit is used to select a neuron in the input layer and calculate the similarity between each neuron in the competition layer and the neuron. The Euclidean distance calculation formula of the similarity is as follows:

[0046]

[0047] Among them, X is the neuron parameter vector of the input layer, X i is the neuron parameter vector of the i-th competitive layer, and T is the vector transposition operator;

[0048] A winning neuron determining unit, configured to compare similarity calculation results of the neurons in the competition layer and select the neuron with the smallest Euclidean distance as the winning neuron;

[0049] The first-order momentum parameter determination unit is used to calculate the winning neuron parameter update adjustment value m with the first-order momentum. The first-order momentum formula is as follows:

[0050] m v (c) = β·m v (c-1)+(1-β)·(D(t)-W v (e))

[0051] Among them, W v is the parameter vector of the winning neuron v, D is the parameter vector of the winning neuron, m v(c) is the parameter update adjustment value with first-order momentum of the winning neuron, m v (c-1) is the last parameter update adjustment value of the winning neuron with first-order momentum, v represents the winning neuron, and β is the first-order momentum hyperparameter;

[0052] The parameter updating unit is used to update the parameters of the winning neuron and its neighboring neurons. The parameter updating formula is as follows:

[0053] W v (e+1)=W v (e)+θ(u,v,σ)×α(e)×m

[0054]

[0055] Among them, W v is the parameter vector of neuron v, e is the current iteration number, θ(u,v,σ) is the nearest neighbor function, which can be a bubble function or a Gaussian function, u is the x-axis coordinate of the winning neuron on the competition plane, v is the y-axis coordinate of the winning neuron on the competition plane, E is the total number of iterations / 2, α is the learning rate, and D is the parameter vector of the winning neuron;

[0056] The iteration unit is used to repeatedly update the neuron parameters of the competitive layer until a set maximum number of iterations is reached.

[0057] According to a third aspect of the present application, there is provided a device, comprising:

[0058] at least one memory and at least one processor;

[0059] The memory is used to store one or more programs;

[0060] When the one or more programs are executed by the at least one processor, the at least one processor implements the steps of a business expansion project clustering method based on SOM-M as described in any one of the first aspects.

[0061] According to a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of the first aspects are implemented.

[0062] The present application provides a business expansion project clustering method based on SOM-M, which can measure the similarity between multiple neurons and multiple objects, make the similarities of multiple neurons compete with each other, and continuously adjust the category center of the neurons approaching the object, so that the neurons are automatically clustered. After adding the first-order momentum to the self-organizing neural network, the convergence ability of the self-organizing neural network is improved, and it is easier to find the cluster center. It has strong classification ability and fast learning ability, and has a good effect in clustering business expansion projects.

[0063] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a flowchart of a method for clustering business expansion projects based on SOM-M in an embodiment of the present application;

[0065] Figure 2 This is a structural diagram of a device for a business expansion project clustering method based on SOM-M in an embodiment of the present application. DETAILED DESCRIPTION

[0066] In order to make the objectives, technical solutions and advantages of the present application clearer, the embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

[0067] It should be clear that the embodiments described are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the embodiments of the present application.

[0068] The terms used in the embodiments of the present application are for the purpose of describing specific embodiments only and are not intended to limit the embodiments of the present application. The singular forms "a," "the," and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0069] When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims. In the description of the present application, it should be understood that the terms "first", "second", "third", etc. are only used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence, nor can they be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0070] In addition, in this application, unless otherwise specified, "plurality" refers to two or more. "And / or" describes the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0071] Due to various reasons, the data source of business expansion supporting projects often leads to abnormalities in a small amount of data. How to accurately identify abnormal data is the key to the success of clustering business expansion supporting projects and predicting the working hours of each node in the business expansion process.

[0072] To solve the above problem, please refer to Figure 1 , Figure 1 This application provides a business expansion project clustering method based on SOM-M, which includes the following steps:

[0073] Step S1: Acquire original business expansion project data, which includes: project nodes, node planned time, and node budget costs.

[0074] Step S2: pre-processing the original business expansion project data to remove abnormal data.

[0075] Specifically, it includes: checking whether there are null values ​​in the original business expansion project data. If a project data contains one or more null value data, the null value data is assigned a value of zero.

[0076] The data obtained in the above steps are checked a second time. If the planned time of a node in a project data is zero, the project data is eliminated.

[0077] Step S3: inputting the pre-processed original business expansion project data into the SOM-M model and initializing the neuron parameters of the competition layer, wherein the SOM-M model includes an input layer and a competition layer.

[0078] The self-organizing map (SOM) generates a low-dimensional, discrete map by learning the data in the input space. To some extent, it can also be seen as a dimensionality reduction algorithm. SOM is an unsupervised artificial neural network. Unlike general neural networks that are trained based on the back propagation of the loss function, it uses a competitive learning strategy, relying on competition between neurons to gradually optimize the network, and uses the neighborhood function to maintain the topological structure of the input space. Maintaining the topological structure of the input space means that the two-dimensional mapping contains the relative distances between data points. Adjacent samples in the input space will be mapped to adjacent output neurons. Because it is based on unsupervised learning, this means that no human intervention is required during the training phase, and we can cluster data without knowing the categories.

[0079] In this application, the SOM-M consists of an input layer and a competitive layer. The number of neurons in the input layer is determined by the input vector dimension, with one neuron corresponding to one feature. The number of neurons in the competitive layer SOM determines the granularity and scale of the final model.

[0080] Step S4: Calculate the similarity and first-order momentum between the neurons in the input layer and the neurons in the competition layer, and update the neuron parameters in the competition layer.

[0081] The neuron parameters of the competitive layer are equivalent to randomly generating a vector, such as a vector of planned time or a vector of budget costs. Then, when calculating the similarity, the Euclidean distance between this vector and the vector of an actual project node is calculated. Updating the neuron parameters of the competitive layer means making the neuron parameters of the competitive layer "close to" an actual project node.

[0082] Step S5: Obtain the node category of the business expansion project according to the updated neuron parameters of the competition layer.

[0083] The competition layer neuron parameters that are closest to an actual project node are the updated competition layer neuron parameters. At this time, it can be considered that the category corresponding to the competition layer neuron parameters is the node category of the business expansion project.

[0084] Specifically, a business expansion project has about 14 nodes. For example, node 1 can be classified into four categories based on the planned time and budget cost of all project nodes 1, namely high budget & high time, high budget & low time, low budget & high time, and low budget & low time.

[0085] In a specific embodiment, the pre-processed original business expansion project data is input into the SOM-M model to initialize the competitive layer neuron parameters, including:

[0086] The node planned time and node budget cost of each project node are used as the neuron parameters of the input layer.

[0087] The number of neurons in the competitive layer and the size of the competitive layer are determined.

[0088] The neuron parameters of the competitive layer are randomly initialized.

[0089] In a specific example, the input layer neuron parameters are determined, including the node's planned duration and budget. For example, to classify node 1 of 100 historical projects based on their planned duration and budget, the input layer size is 100*2. The competitive layer neuron parameters are undetermined and are ultimately determined through iterative parameter updates. For example, to classify into four categories, the applicable parameters are planned duration and budget, so the competitive layer size is set to 4*2, with the parameters randomly initialized according to a range.

[0090] In a specific embodiment, calculating the similarity and first-order momentum between the neurons of the input layer and the neurons of the competitive layer, and updating the neuron parameters of the competitive layer, includes:

[0091] Select a neuron in the input layer and calculate the similarity between each neuron in the competition layer and the neuron. The Euclidean distance calculation formula of the similarity is as follows:

[0092]

[0093] Among them, X is the neuron parameter vector of the input layer, X i is the neuron parameter vector of the i-th competitive layer, and T is the vector transpose operator.

[0094] The similarity calculation results of the neurons in the competition layer are compared, and the neuron with the smallest Euclidean distance is selected as the winning neuron.

[0095] Calculate the winning neuron parameter update adjustment value m with first-order momentum. The first-order momentum formula is as follows:

[0096] m v (c) = β·m v (c-1)+(1-β)·(D(t)-W v (e))

[0097] Among them, W v is the parameter vector of the winning neuron v, D is the parameter vector of the winning neuron, m v(c) is the parameter update adjustment value with first-order momentum of the winning neuron, m v (c-1) is the last parameter update adjustment value of the winning neuron with first-order momentum, v represents the winning neuron, and β is the first-order momentum hyperparameter.

[0098] Update the parameters of the winning neuron and its neighboring neurons. The parameter update formula is as follows:

[0099] W v (e+1)=W v (e)+θ(u,v,σ)×α(e)×m

[0100]

[0101] Among them, W v is the parameter vector of neuron v, e is the current iteration number, θ(u,v,σ) is the nearest neighbor function, which can be bubble function or Gaussian function, u is the x-axis coordinate of the winning neuron on the competition plane, v is the y-axis coordinate of the winning neuron on the competition plane, E is the total number of iterations / 2, α is the learning rate, and D is the parameter vector of the winning neuron.

[0102] The neuron parameters of the competitive layer are updated repeatedly until the set maximum number of iterations is reached.

[0103] In a preferred embodiment, the node category of the business expansion project is obtained according to the updated neuron parameters of the competitive layer, including:

[0104] The Euclidean distance between the input layer heating neuron parameter vector and the competition layer neuron parameter vector is calculated respectively, and each input layer neuron is classified into the competition layer neuron with the smallest Euclidean distance to obtain the node category of the business expansion project.

[0105] Corresponding to the above-mentioned business expansion project clustering method based on SOM-M, the present application also provides a business expansion project clustering device based on SOM-M, including:

[0106] The original business expansion project data acquisition module is used to obtain original business expansion project data, and the original business expansion project data includes: project nodes, node planned time, and node budget cost.

[0107] The data preprocessing module is used to preprocess the original business expansion project data and eliminate abnormal data.

[0108] A parameter initialization module is used to input the preprocessed original business expansion project data into the SOM-M model and initialize the neuron parameters of the competition layer, wherein the SOM-M model includes an input layer and a competition layer.

[0109] The parameter updating module is used to calculate the similarity and first-order momentum between the neurons of the input layer and the neurons of the competition layer, and update the neuron parameters of the competition layer.

[0110] The node category determination module is used to obtain the node category of the business expansion project based on the updated neuron parameters of the competition layer.

[0111] In an optional example, the parameter initialization module includes:

[0112] The input parameter determination unit is used to use the node planned time and node budget cost of each of the project nodes as the neuron parameters of the input layer.

[0113] The competition parameter determination unit is used to determine the number of neurons in the competition layer and the size of the competition layer.

[0114] An initialization unit is used to randomly initialize the neuron parameters of the competitive layer.

[0115] In an optional example, the parameter updating module includes:

[0116] The similarity calculation unit is used to select a neuron in the input layer and calculate the similarity between each neuron in the competition layer and the neuron. The Euclidean distance calculation formula of the similarity is as follows:

[0117]

[0118] Among them, X is the neuron parameter vector of the input layer, X i is the neuron parameter vector of the i-th competitive layer, and T is the vector transpose operator.

[0119] The winning neuron determining unit is used to compare the similarity calculation results of the neurons in the competition layer and select the neuron with the smallest Euclidean distance as the winning neuron.

[0120] The first-order momentum parameter determination unit is used to calculate the winning neuron parameter update adjustment value m with the first-order momentum. The first-order momentum formula is as follows:

[0121] m v (c) = β·m v (c-1)+(1-β)·(D(t)-W v (e))

[0122] Among them, W v is the parameter vector of the winning neuron v, D is the parameter vector of the winning neuron, m v (c) is the parameter update adjustment value with first-order momentum of the winning neuron, m v(c-1) is the last parameter update adjustment value of the winning neuron with first-order momentum, v represents the winning neuron, and β is the first-order momentum hyperparameter.

[0123] The parameter updating unit is used to update the parameters of the winning neuron and its neighboring neurons. The parameter updating formula is as follows:

[0124] W v (e+1)=W v (e)+θ(u,v,σ)×α(e)×m

[0125]

[0126] Among them, W v is the parameter vector of neuron v, e is the current iteration number, θ(u,v,σ) is the nearest neighbor function, which can be bubble function or Gaussian function, u is the x-axis coordinate of the winning neuron on the competition plane, v is the y-axis coordinate of the winning neuron on the competition plane, E is the total number of iterations / 2, α is the learning rate, and D is the parameter vector of the winning neuron.

[0127] The iteration unit is used to repeatedly update the neuron parameters of the competitive layer until a set maximum number of iterations is reached.

[0128] In an optional example, the node category determination module includes:

[0129] The node category determination unit is used to respectively calculate the Euclidean distance between the input layer heating neuron parameter vector and the competition layer neuron parameter vector, classify each input layer neuron into the competition layer neuron with the smallest Euclidean distance, and obtain the node category of the business expansion project.

[0130] Corresponding to the above-mentioned business expansion project clustering method based on SOM-M, the present application also provides a device including at least one memory and at least one processor;

[0131] The memory is used to store one or more programs;

[0132] When the one or more programs are executed by the at least one processor, the at least one processor implements the steps of a business expansion project clustering method based on SOM-M as described in any one of the above items.

[0133] The implementation process of the functions and effects of each component in the above-mentioned device is specifically detailed in the implementation process of the corresponding steps in the above-mentioned method, which will not be repeated here. For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described above is merely schematic, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0134] Corresponding to the above-mentioned method for evaluating the process of power business expansion projects, an embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of any of the above-mentioned SOM-M-based business expansion project clustering methods.

[0135] The present disclosure may take the form of a computer program product implemented on one or more storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-usable storage media include permanent and non-permanent, removable and non-removable media, and information storage may be implemented by any method or technology. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

[0136] The present application provides a business expansion project clustering method based on SOM-M, which can measure the similarity between multiple neurons and multiple objects, make the similarities of multiple neurons compete with each other, and continuously adjust the category center of the neurons approaching the object, so that the neurons are automatically clustered. After adding the first-order momentum to the self-organizing neural network, the convergence ability of the self-organizing neural network is improved, and it is easier to find the cluster center. It has strong classification ability and fast learning ability, and has a good effect in clustering business expansion projects.

[0137] It should be understood that the embodiments of the present application are not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the embodiments of the present application is limited only by the appended claims. The embodiments described above only express several implementation methods of the embodiments of the present application, and their descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the embodiments of the present application, several variations and improvements can be made, which all fall within the scope of protection of the embodiments of the present application.

Claims

1. A business expansion project clustering method based on SOM-M, characterized by: The method comprises the following steps: Obtaining original business expansion project data, wherein the original business expansion project data includes: project nodes, node planned time, and node budget cost; Preprocessing the original business expansion project data to remove abnormal data; Inputting the preprocessed original business expansion project data into the SOM-M model and initializing the neuron parameters of the competition layer, wherein the SOM-M model includes an input layer and a competition layer; Calculating the similarity and first-order momentum between the neurons in the input layer and the neurons in the competition layer, and updating the neuron parameters in the competition layer; Obtaining a node category of the business expansion project according to the updated neuron parameters of the competitive layer; Calculating the similarity and first-order momentum between the neurons in the input layer and the neurons in the competitive layer, and updating the neuron parameters in the competitive layer, includes: Select a neuron in the input layer and calculate the similarity between each neuron in the competition layer and the neuron. The Euclidean distance calculation formula of the similarity is as follows: Among them, X is the neuron parameter vector of the input layer, X i is the neuron parameter vector of the i-th competitive layer, and Τ is the vector transposition operator; Comparing similarity calculation results of the neurons in the competition layer, and selecting the neuron with the smallest Euclidean distance as the winning neuron; Calculate the winning neuron parameter update adjustment value m with first-order momentum. The first-order momentum formula is as follows: m v (c)=β·m v (c-1)+(1-β)·(D(t)-W v (e)) Among them, W v is the parameter vector of the winning neuron v, D is the parameter vector of the winning neuron, m v (c) is the parameter update adjustment value with first-order momentum of the winning neuron, m v (c-1) is the last parameter update adjustment value of the winning neuron with first-order momentum, v represents the winning neuron, and β is the first-order momentum hyperparameter; Update the parameters of the winning neuron and its neighboring neurons. The parameter update formula is as follows: W v (e+1)DW v (e)+θ(u,v,σ)×α(e)×m Among them, W v is the parameter vector of neuron v, e is the current iteration number, θ(u,v,σ) is the nearest neighbor function, which can be a bubble function or a Gaussian function, u is the x-axis coordinate of the winning neuron on the competition plane, v is the y-axis coordinate of the winning neuron on the competition plane, E is the total number of iterations / 2, α is the learning rate, and D is the parameter vector of the winning neuron; The neuron parameters of the competitive layer are updated repeatedly until the set maximum number of iterations is reached.

2. The SOM-M-based business expansion project clustering method according to claim 1, characterized in that: The pre-processed original business expansion project data is input into the SOM-M model to initialize the competitive layer neuron parameters, including: The node planned time and node budget cost of each of the project nodes are used as neuron parameters of the input layer; Determining the number of neurons in the competitive layer and the size of the competitive layer; The neuron parameters of the competitive layer are randomly initialized.

3. The business expansion project clustering method based on SOM-M according to claim 1, characterized in that: According to the updated neuron parameters of the competitive layer, the node categories of the business expansion project are obtained, including: The Euclidean distance between the input layer heating neuron parameter vector and the competition layer neuron parameter vector is calculated respectively, and each input layer neuron is classified into the competition layer neuron with the smallest Euclidean distance to obtain the node category of the business expansion project.

4. A device for clustering business expansion projects based on SOM-M, characterized in that: include: An original business expansion project data acquisition module is used to acquire original business expansion project data, wherein the original business expansion project data includes: project nodes, node planned time, and node budget cost; A data preprocessing module, used to preprocess the original business expansion project data and remove abnormal data; A parameter initialization module, configured to input the preprocessed raw business expansion project data into the SOM-M model and initialize the neuron parameters of the competitive layer, wherein the SOM-M model includes an input layer and a competitive layer; A parameter updating module, configured to calculate the similarity and first-order momentum between the neurons in the input layer and the neurons in the competition layer, and update the parameters of the neurons in the competition layer; A node category determination module, configured to obtain the node category of the business expansion project based on the updated neuron parameters of the competitive layer; Among them, the parameter update module includes: The similarity calculation unit is used to select a neuron in the input layer and calculate the similarity between each neuron in the competition layer and the neuron. The Euclidean distance calculation formula of the similarity is as follows: Among them, X is the neuron parameter vector of the input layer, X i is the neuron parameter vector of the i-th competitive layer, and Τ is the vector transposition operator; A winning neuron determination unit, configured to compare similarity calculation results of the neurons in the competition layer and select the neuron with the smallest Euclidean distance as the winning neuron; The first-order momentum parameter determination unit is used to calculate the winning neuron parameter update adjustment value m with the first-order momentum. The first-order momentum formula is as follows: m v (c)=β·m v (c-1)+(1-β)·(D(t)-W v (e)) Among them, W v is the parameter vector of the winning neuron v, D is the parameter vector of the winning neuron, m v (c) is the parameter update adjustment value with first-order momentum of the winning neuron, m v (c-1) is the last parameter update adjustment value of the winning neuron with first-order momentum, v represents the winning neuron, and β is the first-order momentum hyperparameter; The parameter updating unit is used to update the parameters of the winning neuron and its neighboring neurons. The parameter updating formula is as follows: W v (e+1)DW v (e)+θ(u,v,σ)×α(e)×m Among them, W v is the parameter vector of neuron v, e is the current iteration number, θ(u,v,σ) is the nearest neighbor function, which can be a bubble function or a Gaussian function, u is the x-axis coordinate of the winning neuron on the competition plane, v is the y-axis coordinate of the winning neuron on the competition plane, E is the total number of iterations / 2, α is the learning rate, and D is the parameter vector of the winning neuron; The iteration unit is used to repeatedly update the neuron parameters of the competitive layer until a set maximum number of iterations is reached.

5. The device for clustering business expansion projects based on SOM-M according to claim 4, characterized in that: The parameter initialization module includes: An input parameter determination unit, configured to use the node planned time and node budget cost of each of the project nodes as neuron parameters of the input layer; a competition parameter determination unit, configured to determine the number of neurons in the competition layer and the size of the competition layer; An initialization unit is used to randomly initialize the neuron parameters of the competitive layer.

6. The device for clustering business expansion projects based on SOM-M according to claim 4, characterized in that: The node category determination module includes: The node category determination unit is used to respectively calculate the Euclidean distance between the input layer heating neuron parameter vector and the competition layer neuron parameter vector, classify each input layer neuron into the competition layer neuron with the smallest Euclidean distance, and obtain the node category of the business expansion project.

7. A device, characterized in that include: at least one memory and at least one processor; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, the at least one processor implements the steps of the business expansion project clustering method based on SOM-M as described in any one of claims 1 to 3.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images