Intelligent management system and method for cable production

Abstract

The invention discloses an intelligent management system and method for cable production, and the system builds an intelligent decision-making system for demand prediction through the dynamic fusion of historical sales data and market trend information. Specifically, a time sequence feature coding and semantic embedding technology based on deep learning is adopted, cross-modal association is carried out on a local rule of a historical sales window and market dynamic semantics, and deep collaborative decision response is further carried out on historical information and market trends to obtain historical sales-market trend collaborative decision response coding features; and based on this, short-term prediction of the cable market demand is realized. In this way, the timeliness and accuracy of demand prediction are remarkably improved, the risk of supply chain imbalance caused by sudden demand change is effectively relieved, and then dual optimization of the resource utilization rate and the market response capacity is achieved.

Description

Technical Field

[0001] This application relates to the field of intelligent management, and more specifically, to an intelligent management system and method for cable production. Background Art

[0002] In the field of cable production, traditional intelligent management systems are usually built around business management, inventory management, integration management, production management, and safety management modules. Chinese Patent CN111369383A provides an intelligent integrated management system for cable production, in which the generation of the raw material procurement plan depends on the sales order statistics data of the business management module and the storage quantity information of the inventory management module. The procurement order and production plan are formulated by integrating these two types of static statistical information through the integration management module.

[0003] However, market demand has the characteristics of dynamic fluctuations. Making procurement decisions only based on the historical statistical information of current sales orders and inventory levels has significant lag and rigidity defects. Specifically, the sales data of the business management module only reflects the results of completed transactions, while the data of the inventory management module only represents the current storage status. Neither of them incorporates the in-depth correlation analysis of market trend information and sales time series characteristics. This static decision-making mechanism makes it difficult for the procurement plan to adapt to short-term fluctuations in market demand, resulting in insufficient elasticity of the enterprise's supply chain.

[0004] Therefore, an optimized intelligent management method for cable production is expected. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide an intelligent management system and method for cable production, which constructs an intelligent decision-making system for demand prediction by dynamically integrating historical sales data and market trend information. Specifically, the time series feature encoding and semantic embedding technology based on deep learning are used to cross-modally associate the local laws of the historical sales window with the market dynamic semantics, and further perform in-depth collaborative decision-making responses on historical information and market trends to obtain the historical sales - market trend collaborative decision-making response coding features, and based on this, to achieve short-term prediction of the cable market demand volume. In this way, the timeliness and accuracy of demand prediction are significantly improved, the risk of supply chain imbalance caused by sudden demand changes is effectively alleviated, and thus the dual optimization of resource utilization rate and market response ability is realized.

[0006] According to one aspect of this application, an intelligent management method for cable production is provided, which includes:

[0007] Obtain historical sales data and market trend information;

[0008] Extract the sales time series coding features from the historical sales data to obtain the time series of the local sales volume time series coding features;

[0009] Perform semantic embedding coding on the market trend information to obtain the market trend information semantic embedding coding features;

[0010] Input the time series of the market trend information semantic embedding coding features and the local sales volume time series coding features into the historical sales - market trend semantic collaborative deep interaction network to obtain the historical sales - market trend collaborative decision - response coding features;

[0011] Based on the historical sales - market trend collaborative decision - response coding features, obtain the short - term predicted value of the cable market demand.

[0012] According to another aspect of the present application, there is provided an intelligent management system for cable production, which includes:

[0013] An information acquisition module for acquiring historical sales data and market trend information;

[0014] A sales time series coding feature extraction module for extracting the sales time series coding features from the historical sales data to obtain the time series of the local sales volume time series coding features;

[0015] A market trend information semantic analysis module for performing semantic embedding coding on the market trend information to obtain the market trend information semantic embedding coding features;

[0016] A historical sales - market trend semantic collaborative interaction module for inputting the time series of the market trend information semantic embedding coding features and the local sales volume time series coding features into the historical sales - market trend semantic collaborative deep interaction network to obtain the historical sales - market trend collaborative decision - response coding features;

[0017] A demand prediction module for obtaining the short - term predicted value of the cable market demand based on the historical sales - market trend collaborative decision - response coding features.

[0018] Compared with the prior art, an intelligent management system and method for cable production provided by the present application constructs an intelligent decision-making system for demand forecasting by dynamically integrating historical sales data and market trend information. Specifically, by using the time series feature encoding and semantic embedding technology based on deep learning, the local rules of the historical sales window and the market dynamic semantics are cross-modally associated, and further, a deep collaborative decision response is made to the historical information and market trends to obtain the collaborative decision response coding features of historical sales - market trends, and based on this, a short-term prediction of the cable market demand is realized. In this way, the timeliness and accuracy of demand forecasting are significantly improved, the risk of supply chain imbalance caused by sudden demand changes is effectively alleviated, and thus the dual optimization of resource utilization rate and market response ability is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present application will become more obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 It is a flowchart of an intelligent management method for cable production according to an embodiment of the present application;

[0021] Figure 2 It is a schematic diagram of data flow of an intelligent management method for cable production according to an embodiment of the present application;

[0022] Figure 3 It is a flowchart of sub-step S4 of an intelligent management method for cable production according to an embodiment of the present application;

[0023] Figure 4 It is a block diagram of an intelligent management system for cable production according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0025] As shown in this application and the claims, unless the context clearly indicates otherwise, words such as "a", "an", "one", and / or "the" are not specifically singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0026] Although this application makes various references to certain modules in the system according to the embodiments of this application, however, any number of different modules can be used and run on the user terminal and / or server. The modules are merely illustrative, and different aspects of the system and method can use different modules.

[0027] Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the operations before or below do not necessarily need to be executed precisely in sequence. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or several steps can be removed from these processes.

[0028] Next, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited by the exemplary embodiments described here.

[0029] In the technical solution of this application, an intelligent management method for cable production is proposed. Figure 1 It is a flowchart of an intelligent management method for cable production according to an embodiment of this application. Figure 2 It is a schematic diagram of data flow of an intelligent management method for cable production according to an embodiment of this application. As Figure 1 and Figure 2 shown, the intelligent management method for cable production according to the embodiment of this application includes the steps: S1, obtaining historical sales data and market trend information; S2, extracting the sales time series coding features in the historical sales data to obtain a time series of sales volume local time series coding features; S3, performing semantic embedding coding on the market trend information to obtain market trend information semantic embedding coding features; S4, inputting the market trend information semantic embedding coding features and the time series of sales volume local time series coding features into a historical sales volume - market trend semantic collaborative depth interaction network to obtain historical sales volume - market trend collaborative decision response coding features; S5, based on the historical sales volume - market trend collaborative decision response coding features, obtaining a short-term predicted value of the cable market demand.

[0030] Specifically, in S1, historical sales data and market trend information are obtained. Considering that traditional cable production management systems formulate procurement plans based on static statistical information (such as current inventory and completed orders), when market demand experiences short-term and drastic fluctuations due to external events (such as policy adjustments, emergencies) or internal factors (such as supply chain disruptions), the linear prediction method relying solely on historical sales data will be difficult to adapt to the short-term fluctuations of market demand in a timely manner, resulting in insufficient elasticity of the enterprise's supply chain. Therefore, in the technical solution of this application, historical sales data and market trend information are obtained. Among them, historical sales data refers to the actual sales records of cables within a certain period in the past of the enterprise, and its essence is structured transaction sequence data with time as the dimension, such as monthly sales volume, quarterly order volume, etc. After being segmented by a time window, such data can reveal the periodic laws of sales fluctuations (such as seasonal demand peaks) and local characteristics of sudden events (such as concentrated purchases by large customers); while market trend information covers a wider range of contents, such as industry dynamics, competitor conditions, macroeconomic indicators, changes in policies and regulations, and other factors affecting market demand, which can help the system identify the current market trend and possible future development trends. By integrating historical sales data and market trend information, short-term demand changes in the market can be better captured, so as to formulate more accurate and effective procurement and production plans.

[0031] Specifically, in S2, sales time series coding features in historical sales data are extracted to obtain a time series of local sales volume time series coding features. In the embodiment of the present application, first, historical sales data is segmented based on a predetermined time window to obtain a time series of historical sales window data. It should be understood that the dynamics of historical sales data are closely related to short-term fluctuations in market demand. However, traditional systems usually treat historical data as an overall statistic for processing, such as directly calculating the annual sales average or total. This coarse-grained analysis method will obscure the key short-term fluctuation patterns in sales data. If directly modeling the original long-sequence sales data, the model is vulnerable to interference from non-related cycles in historical data and cannot accurately capture the mutation inflection points within the recent sales window, resulting in the prediction model being unable to capture the subtle changes in market demand. Therefore, in the technical solution of the present application, historical sales data is segmented based on a predetermined time window to obtain a time series of historical sales window data. Specifically, by setting a sliding time window (such as a dynamic division in weeks / days), the system can extract local time series segments with business interpretability. The data sequence of historical sales window data after time window segmentation not only retains the time dependence of the sales volume but also provides a clear time series context for the model by dividing the segment boundaries. Among them, each historical sales window data in the time series of historical sales window data can independently represent the sales status of the current month, while the sequential relationship between segments implies the continuity of the evolution of market demand, thereby capturing the short-term fluctuation trend in sales data and helping to reveal the true demand pattern of the market.

[0032] Furthermore, a sales time-series feature encoder based on the LSTM model is used to perform time-series encoding on each historical sales window data in the time series of the historical sales window data to obtain a time series of sales volume local time-series feature encoding vectors as the time series of sales volume local time-series encoding features. Considering that the time-series characteristics of historical sales data contain the dynamic laws of market demand evolution, traditional neural network models often use simple moving average or aggregation statistical methods to process sales data, and thus cannot effectively capture the non-linear associations and long-term dependence relationships in sales fluctuations. As a type of recurrent neural network, LSTM (Long Short-Term Memory) has a unique gating mechanism (forget gate, input gate, output gate) that can adaptively learn the long-term dependence and short-term mutations in the time series, such as identifying the demand recovery cycle after a sales trough or the chain effect of sudden orders on subsequent sales volume. Therefore, in the technical solution of this application, a sales time-series feature encoder based on the LSTM model is used to perform time-series encoding on each historical sales window data in the time series of the historical sales window data to uncover the dynamic evolution laws within each sales window and obtain a time series of sales volume local time-series feature encoding vectors. During this process, after being encoded by LSTM, each historical sales window data not only contains the sales volume statistical values (such as mean, variance) within that window, but also implicitly contains its dynamic association with the front and back time windows. This encoding process is essentially an abstract modeling of sales dynamics. In this way, the system can more accurately understand the dynamic mechanism behind past sales behaviors, and thus provide strong support for predicting future market demands.

[0033] Specifically, in step S3, semantic embedding encoding is performed on the market trend information to obtain the semantic embedding encoding features of the market trend information. In the embodiments of the present application, the market trend information is input into a semantic encoder based on the Bert model to obtain a semantic embedding encoding vector of the market trend information as the semantic embedding encoding features of the market trend information. Since market trend information usually exists in the form of unstructured text, such as infrastructure plans issued by the government, technical route analysis in industry white papers, strategic movement reports of competitors, or interpretations of macroeconomic policies. This type of information contains potential driving factors for changes in market demand, but its semantic complexity and context relevance make it difficult for traditional keyword matching or simple statistical methods to effectively extract deep semantic features. BERT (Bidirectional Encoder Representations from Transformers), as a pre-trained language model, can capture the global dependency relationships between words in market trend information through a bidirectional attention mechanism. Therefore, in the technical solution of the present application, the market trend information is input into a semantic encoder based on the Bert model to obtain a semantic embedding encoding vector of the market trend information. Here, the Bert model can automatically learn the key elements contained in the input market trend information and convert them into a high-dimensional numerical representation form. By using a semantic encoder based on the Bert model, deep market trend information semantic features can be effectively extracted from unstructured market trend information text data, laying a solid foundation for subsequent market demand prediction and analysis.

[0034] Specifically, in step S4, the time series of the semantic embedded encoding features of the market trend information and the local time series encoding features of the sales volume are input into the historical sales volume - market trend semantic collaborative deep interaction network to obtain the historical sales volume - market trend collaborative decision response encoding features. It should be understood that the semantic embedded features of the market trend (such as policy orientation, raw material price fluctuations) are essentially high - dimensional non - linear semantic expressions, while the local time series encoding features of the sales volume (such as weekly sales volume fluctuation patterns) contain dynamic time dependencies. There are modal differences and information barriers between the two in the original feature space. Traditional linear models or shallow neural networks are difficult to model the deep interaction relationships of such cross - modal features. Therefore, to overcome the above difficulties and achieve the deep collaboration between market dynamics and sales time series, in the technical solution of this application, the time series of the semantic embedded encoding vector of the market trend information and the local time series feature encoding vector of the sales volume are input into the historical sales volume - market trend semantic collaborative deep interaction network to obtain the historical sales volume - market trend collaborative decision response encoding features. That is, by constructing the graph structure relationship in the latent space, the deep collaboration between the semantic information of the market trend and the local time series features of the sales volume is realized. Specifically, first, the semantic embedded encoding vector of the market trend information is used as the static decision point, and the time series of the local time series feature encoding vector of the sales volume is used as the dynamic decision point, and the non - linear decision response unit is used to capture the interaction pattern between the historical sales volume and the market trend in the latent space; then, the local connection relationship (such as the spatio - temporal coupling between the policy release window and the order surge window) and the global structure constraint (such as the macro - matching between the industry cycle and the sales trend) of the market and sales features are explicitly modeled through the graph structure. In this way, the system can extract low - dimensional but high - information - density historical sales volume - market trend collaborative features from heterogeneous data, enabling enterprises to achieve accurate demand perception and flexible resource allocation in a high - noise market environment. Specifically, in a specific example of this application, as Figure 3 shown, step S4 includes: S41, capturing the cross - modal dynamic association between the market trend and the sales volume time series through a non - linear interaction unit to generate the implicit encoding features of the historical sales volume - market trend decision point state; S42, performing spectral - graph - guided adaptive aggregation analysis on the implicit encoding features of the historical sales volume - market trend decision point state to obtain the historical sales volume - market trend collaborative decision response encoding features.

[0035] Specifically, in S41, a cross-modal dynamic association between the market trend and the sales volume time series is captured by a non-linear interaction unit to generate a hidden encoding feature of the historical sales volume - market trend decision point state. That is, in the technical solution of this application, each sales volume local time series feature encoding vector in the time series of the market trend information semantic embedding encoding vector and the sales volume local time series feature encoding vector is input into the historical sales volume - market trend non-linear decision response unit to obtain a sequence of hidden encoding vectors of the historical sales volume - market trend decision point state as the hidden encoding feature of the historical sales volume - market trend decision point state. It should be understood that the market trend information semantic embedding encoding feature vector reflects the long-term influence of the macro market environment, while the time series of the sales volume local time series feature encoding vector characterizes the short-term dynamics of micro transactions. Traditional linear models or shallow neural networks are difficult to model the non-linear coupling relationship between historical sales volume and market trend and cannot quantify the hidden association between the two in time series. In the technical solution of this application, the non-linear decision response unit can adaptively capture the dynamic dependence relationship of the historical sales volume - market trend cross-modal features through the construction of a hidden space. That is, by constructing a dynamic mapping of the hidden space, the high-dimensional entanglement state of market and sales features is deconstructed. Specifically, the static market trend semantic vector is used as the environmental decision anchor point, and the dynamic sales time series encoding vector sequence is used as the operation state variable, and the response trajectories of the two in the hidden space are reconstructed using non-linear mapping. In this way, the planar interaction of traditional feature superposition is broken through, and the generated sequence of hidden encoding vectors of the historical sales volume - market trend decision point state fully reflects the real state and development trend of the market, providing strong support for market demand prediction. In a specific example of this application, the following formula is used to input each sales volume local time series feature encoding vector in the time series of the market trend information semantic embedding encoding vector and the sales volume local time series feature encoding vector into the historical sales volume - market trend non-linear decision response unit to obtain a sequence of hidden encoding vectors of the historical sales volume - market trend decision point state; where the formula is:

[0036]

[0037]

[0038]

[0039]

[0040] Among them, is the time series of the sales volume local time series feature encoding vector, , , and are the 1st, 2nd, and The local time series feature encoding vectors of sales volume, is the structured mapping encoding vector of the core board parameters, is matrix multiplication, and are respectively the corresponding decision response weight matrix and decision response bias vector, is the activation function, , , , and are respectively the 1st, 2nd, th, th and th historical sales - market trend decision point state latent encoding vectors in the sequence of the historical sales - market trend decision point state latent encoding vectors, is the sequence of the historical sales - market trend decision point state latent encoding vectors.

[0041] Specifically, in step S42, an adaptive aggregation analysis of the historical sales - market trend decision point state implicit coding features based on spectrogram guidance is performed to obtain the historical sales - market trend collaborative decision response coding features. In the embodiments of the present application, first, an implicit state graph structure analysis of the historical sales - market trend decision point state implicit coding features is performed to obtain the historical sales - market trend decision point state class neighborhood matrix and the historical sales - market trend decision point state degree matrix. Since the in - depth collaborative analysis of market trends and historical sales depends on the effective modeling of the internal structure of the high - dimensional feature space, when the semantic embedding features of market trends and the sales time - series coding features are mapped to the hidden space through a non - linear decision response unit, although the formed sequence of decision point state implicit coding vectors decouples part of the noise of the historical sales - market trend features through non - linear mapping, it may still exhibit a manifold structure or clustering characteristics in the high - dimensional space. Traditional statistical methods or shallow models are difficult to directly analyze such high - dimensional non - linear structures. If a global analysis of the original hidden space is directly performed, the model is vulnerable to interference from redundant features and cannot focus on key associations. Therefore, in the technical solution of the present application, the historical sales - market trend decision point state class neighborhood matrix and the historical sales - market trend decision point state degree matrix of the sequence of historical sales - market trend decision point state implicit coding vectors are calculated; through the construction of the neighborhood matrix, the complex associations between the sequences of historical sales - market trend decision point state implicit coding vectors can be transformed into local connection relationships of the graph structure. The calculation of each connection weight in the historical sales - market trend decision point state class neighborhood matrix essentially quantifies the state association intensity of different historical sales - market trend decision points in the hidden space, while the degree matrix identifies the hub status of key decision points (such as the sales window in the week when a policy is released) in the graph structure through the statistics of the number of node connections, providing a basis for weighted structural importance for subsequent spectral analysis. In a specific example of the present application, the following formula is used to perform an implicit state graph structure analysis of the historical sales - market trend decision point state implicit coding features to obtain the historical sales - market trend decision point state class neighborhood matrix and the historical sales - market trend decision point state degree matrix; where the formula is:

[0042]

[0043]

[0044]

[0045] Wherein, is the two - norm of the calculated vector, is the inverse hyperbolic cosine function, , , , and are the eigenvalues at respective positions in the historical sales - market trend decision point state - type neighborhood matrix, is the historical sales - market trend decision point state - type neighborhood matrix, is the square of the calculation of the vector one - norm, is one less than the number of vectors in , and are the eigenvalues at respective positions on the diagonal in the historical sales - market trend decision point state - type degree matrix, is the historical sales - market trend decision point state - type degree matrix.

[0046] Next, based on the historical sales - market trend decision point state - type neighborhood matrix and the historical sales - market trend decision point state - type degree matrix, calculate the historical sales - market trend decision point state Laplacian matrix. It should be understood that although the neighborhood matrix explicitly defines the association strength between nodes, such as the coupling degree between policy text semantics and sales volume fluctuations in a specific period, its discretized adjacency relationship cannot be directly used to quantify global structural characteristics, such as the demand propagation path driven by policies; while the degree matrix calibrates the centrality of nodes, such as a core policy being associated with multiple sales time - series nodes, but the degree value of a single node cannot reveal its role in the overall graph structure, such as a hub node or an edge node. Therefore, in the technical solution of this application, based on the historical sales - market trend decision point state - type neighborhood matrix and the historical sales - market trend decision point state - type degree matrix, calculate the historical sales - market trend decision point state Laplacian matrix. That is, through the spectral characteristics of the Laplacian matrix, reveal the underlying manifold structure of historical sales - market trends. In the cable demand forecasting scenario, the synergy effect between market trends and sales time - series often lies in the low - frequency components of the graph structure. By constructing the Laplacian matrix, the system can map the complex graph topology to the frequency domain space, enabling the model to focus on low - dimensional features (such as the resonance period of policy - sales volume) that reflect the true market laws, rather than being misled by short - term perturbations. This frequency - domain - based feature screening mechanism significantly enhances the robustness of the model, enabling the production plan to accurately respond to the essential associations in market - sales dynamics, rather than being disturbed by surface fluctuations. In addition, the algebraic form of the Laplacian matrix also provides a mathematical framework for the unified processing of cross - modal features, such as through the linear combination of features to achieve the collaborative weight allocation of policy semantics and sales time - series, thereby supporting the intelligent decision - making of the flexible supply chain. In a specific example of this application, calculate the historical sales - market trend decision point state Laplacian matrix with the following formula; where the formula is:

[0047]

[0048] Among them, is the Laplacian matrix of the historical sales - market trend decision point state.

[0049] Then, perform spectral decomposition on the Laplacian matrix of the historical sales - market trend decision point state to obtain a sequence of core component coding vectors of the historical sales - market trend decision point. Specifically, in the technical solution of this application, first, perform spectral decomposition on the Laplacian matrix of the historical sales - market trend decision point state to obtain a sequence of core component coding vectors of the historical sales - market trend decision point; it should be understood that first, due to the noise sensitivity and topological incompleteness of graph - structured data, when the Laplacian matrix of the historical sales - market trend decision point state is constructed, its spectral information may be distorted due to data noise, local connection anomalies, or sparsity. For example, due to the ambiguity of text semantics of a certain policy node or the short - term fluctuations of sales nodes, isolated edges or false associations may be formed in the graph structure, interfering with the stability of the global topology. Therefore, in the technical solution of this application, use a topological closure mechanism to optimize the Laplacian matrix of the historical sales - market trend decision point state based on cut - cycle space closure to obtain an optimized Laplacian matrix of the historical sales - market trend decision point state. That is, using the connected - component reconstruction mechanism in topology to strip the distortion of local perturbations on the global structure. During this process, the topological closure mechanism extracts the connected components of the historical sales - market trend decision point state - like neighborhood matrix and the historical sales - market trend decision point state - like degree matrix through pseudo - inverse operations, and decomposes the matrix into a cut space and a cycle space. Among them, cut - space closure suppresses edge noise by strengthening the core connected paths (such as the demand propagation chain driven by policies), while cycle - space closure enhances the continuity of periodic patterns by closing loops (such as the collaborative pattern of quarterly sales fluctuations and policy cycles). Through the joint optimization of cut - cycle space closure, the Laplacian matrix of the historical sales - market trend decision point state can capture both the steady - state and dynamic components in market - sales collaboration. The optimized Laplacian matrix of the historical sales - market trend decision point state not only filters out the isolated edges generated by sales data noise but also strengthens the real market - driven mode, thereby enhancing the anti - interference ability and topological representation accuracy of the model for complex market - sales relationships.

[0050] In the above - mentioned preferred example, the specific steps for optimizing the Laplacian matrix of the historical sales - market trend decision point state based on cut - cycle space closure using the topological closure mechanism are as follows:

[0051] Specifically, first use pseudo - inverse operations to extract the connected - component information of the historical sales - market trend decision point state - like neighborhood matrix and the historical sales - market trend decision point state - like degree matrix, that is, let:

[0052]

[0053]

[0054] Among them, the matrix and can respectively represent the cut space and the cycle space of the historical sales - market trend decision point state Laplacian matrix , and represents matrix multiplication;

[0055] Thus, by closing the cut space and the cycle space, the historical sales - market trend decision point state Laplacian matrix is optimized, which is expressed as:

[0056]

[0057] Among them, is the exponential function value with the natural constant as the base, is addition by position points, is the optimized historical sales - market trend decision point state Laplacian matrix.

[0058] In this way, by performing a closure operation based on the pseudoinverse to extract the global structural invariants in the structural information of the graph, the spectral topology information of the historical sales - market trend decision point state Laplacian matrix is represented more robustly.

[0059] Furthermore, the sequence of the historical sales - market trend decision point core component coding vectors is adaptively aggregated in terms of features to obtain the historical sales - market trend collaborative decision response coding vector. That is, the market - sales collaborative relationship is compressed from a high - dimensional topological space to a low - dimensional spectral domain to retain the key historical sales - market trend decision point structural information and eliminate redundant details, where the smallest eigenvalue corresponds to the low - frequency component (such as the main trend of demand driven by policies), while the high - frequency component represents random perturbations (such as sales pulses caused by logistics disruptions). Through the feature refinement mechanism based on the spectral domain, the system can accurately capture the essential laws in the market - sales collaboration, thereby improving the accuracy of market demand prediction and enhancing the agility of the producer in responding to market fluctuations and the efficiency of resource allocation.

[0060] In a specific example of this application, the historical sales - market trend decision point state Laplacian matrix is spectrally decomposed by the following formula to obtain the sequence of the historical sales - market trend decision point core component coding vectors; among them, the formula is:

[0061]

[0062] Among them, To perform spectral decomposition operations on , is a sequence of the core component coding vectors of the historical sales - market trend decision point, , , and are respectively the 1st, 2nd, th, and th core component coding vectors of the historical sales - market trend decision point in the sequence of the core component coding vectors of the historical sales - market trend decision point, is a historical sales - market trend eigenvalue diagonal matrix with elements on the diagonal being , , and , , , and are respectively , , and corresponding eigenvalues, is the historical sales - market trend eigenvalue diagonal matrix.

[0063] Subsequently, perform feature adaptive aggregation on the sequence of the core component coding vectors of the historical sales - market trend decision point to obtain the historical sales - market trend collaborative decision response coding vector. It should be understood that although the sequence of the core component coding vectors of the historical sales - market trend decision point has filtered out noise, the influence weights of the represented feature dimensions (such as policy correlation intensity, sales cycle pattern) on decision - making may change significantly in different market environments. Traditional fixed - weight fusion methods cannot adapt to such dynamic requirements. For example, when the policy effect lags, it is necessary to delay the weight activation timing of low - frequency components to avoid premature procurement. Therefore, in the technical solution of this application, adaptive fusion of the sequence of the core component coding vectors of the historical sales - market trend decision point not only integrates the complementary information of the core components of the historical sales - market trend decision point, but also achieves semantic alignment of the historical sales - market trend decision feature space through dynamic weight allocation. The generated historical sales - market trend collaborative decision response coding vector can accurately reflect the changing trend of market demand. In this way, the system can more flexibly and accurately predict the short - term fluctuations of market demand, help enterprises timely adjust production strategies, reduce the risk of excess inventory, and at the same time ensure the stability of product supply. In a specific example of this application, the following formula is used to perform feature adaptive aggregation on the sequence of the core component coding vectors of the historical sales - market trend decision point to obtain the historical sales - market trend collaborative decision response coding vector; where, the formula is:

[0064]

[0065] Among them, For performing an adaptive fusion operation, and are respectively the corresponding fusion weight matrix and fusion bias vector, is a function, is the corresponding historical sales - market trend scoring weight vector, is the corresponding historical sales - market trend weight value, is a masking operation, is a preset threshold, is the corresponding historical sales - market trend masking weight value, is the historical sales - market trend collaborative decision - making response coding vector.

[0066] Specifically, in step S5, based on the historical sales - market trend collaborative decision - making response coding features, a short - term predicted value of the cable market demand is obtained. In the technical solution of this application, feature decoding and mapping are performed on the historical sales - market trend collaborative decision - making response coding vector to obtain the short - term predicted value of the cable market demand. It should be understood that the historical sales - market trend collaborative decision - making response coding vector has already integrated market trend information and sales time - series characteristics, reflecting the main structural direction and key factors of the market. However, it still needs to be converted into an intuitive and operable demand prediction value. Therefore, in the technical solution of this application, feature decoding and mapping are performed on the historical sales - market trend collaborative decision - making response coding vector to restore the abstract decision - making response coding features extracted by the deep - learning model into specific market demand prediction values, so as to provide direct support for the procurement and production decisions of enterprises. In a specific example of this application, the short - term predicted value of the cable market demand can be obtained by inputting the historical sales - market trend collaborative decision - making response coding vector into a short - term prediction model of the cable market demand based on a decoder. Among them, the decoder maps the high - dimensional collaborative decision - making response coding vector back to a low - dimensional actual demand prediction value by using the trained model parameters and algorithms. In this way, the system can more accurately predict the short - term fluctuations of market demand and help enterprises make more scientific and reasonable procurement and production decisions.

[0067] In summary, the intelligent management method for cable production according to the embodiments of the present application is elucidated. It constructs an intelligent decision-making system for demand forecasting by dynamically integrating historical sales data and market trend information. Specifically, it adopts time series feature encoding and semantic embedding techniques based on deep learning to cross-modally associate the local patterns of the historical sales window with the market dynamic semantics, and further makes a deep collaborative decision response to the historical information and market trends to obtain the historical sales volume - market trend collaborative decision response coding features, and based on this, realizes the short-term prediction of the cable market demand. In this way, the timeliness and accuracy of demand forecasting are significantly improved, the risk of supply chain imbalance caused by sudden demand changes is effectively alleviated, and the dual optimization of resource utilization rate and market response ability is realized.

[0068] Furthermore, an intelligent management system for cable production is also provided.

[0069] Figure 4 FIG. is a block diagram of the intelligent management system for cable production according to the embodiments of the present application. As Figure 4 shown, the intelligent management system 300 for cable production according to the embodiments of the present application includes: an information acquisition module 310 for acquiring historical sales data and market trend information; a sales time series coding feature extraction module 320 for extracting the sales time series coding features in the historical sales data to obtain a time series of sales volume local time series coding features; a market trend information semantic analysis module 330 for performing semantic embedding coding on the market trend information to obtain market trend information semantic embedding coding features; a historical sales volume - market trend semantic collaborative interaction module 340 for inputting the time series of market trend information semantic embedding coding features and sales volume local time series coding features into a historical sales volume - market trend semantic collaborative deep interaction network to obtain historical sales volume - market trend collaborative decision response coding features; and a demand forecasting module 350 for obtaining a short-term forecast value of the cable market demand based on the historical sales volume - market trend collaborative decision response coding features.

[0070] As described above, the intelligent management system 300 for cable production according to the embodiments of the present application can be implemented in various wireless terminals, such as a server with an intelligent management algorithm for cable production. In a possible implementation manner, the intelligent management system 300 for cable production according to the embodiments of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the intelligent management system 300 for cable production can be a software module in the operating system of the wireless terminal, or can be an application program developed for the wireless terminal; of course, the intelligent management system 300 for cable production can also be one of the numerous hardware modules of the wireless terminal.

[0071] Alternatively, in another example, the intelligent management system 300 for cable production and the wireless terminal may also be separate devices, and the intelligent management system 300 for cable production can be connected to the wireless terminal through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format.

[0072] The embodiments of the present disclosure have been described above. The above description is exemplary and not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of the terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An intelligent management method for cable production, characterized in that: include: Get historical sales data and market trend information; Extract sales time series coding features from historical sales data to obtain a time series of local time series coding features of sales volume; Performing semantic embedding coding on the market trend information to obtain semantic embedding coding features of the market trend information; The time series of the semantic embedding coding features of market trend information and the local temporal coding features of sales volume are input into the historical sales volume-market trend semantic collaborative deep interaction network to obtain the historical sales volume-market trend collaborative decision response coding features, including: capturing the cross-modal dynamic association between market trends and sales volume time series through nonlinear interaction units to generate the historical sales volume-market trend decision point state implicit coding features; performing spectral-guided historical sales volume-market trend decision point adaptive aggregation analysis on the historical sales volume-market trend decision point state implicit coding features to obtain the historical sales volume-market trend collaborative decision response coding features; Based on the historical sales volume-market trend collaborative decision-making response coding features, the short-term forecast value of the cable market demand is obtained.

2. The intelligent management method for cable production according to claim 1, characterized in that: Extract the sales time series coding features from the historical sales data to obtain the time series of local time series coding features of sales volume, including: Segment the historical sales data into data segments based on a predetermined time window to obtain a time series of the historical sales window data; A sales time series feature encoder based on the LSTM model is used to time series encode each historical sales window data in the time series of the historical sales window data to obtain a time series of local time series feature encoding vectors of sales volume as a time series of local time series encoding features of sales volume.

3. The intelligent management method for cable production according to claim 2, characterized in that: The market trend information is semantically embedded and encoded to obtain the semantic embedding encoding features of the market trend information, including: The market trend information is input into the semantic encoder based on the Bert model to obtain the market trend information semantic embedding coding vector as the market trend information semantic embedding coding feature.

4. The intelligent management method for cable production according to claim 3, characterized in that: The cross-modal dynamic association between market trends and sales time series is captured through nonlinear interaction units to generate implicit encoding features of historical sales-market trend decision point states, including: The market trend information semantic embedding coding vector and each sales volume local temporal feature coding vector in the time series of the sales volume local temporal feature coding vector are input into the historical sales volume-market trend nonlinear decision response unit to obtain a sequence of historical sales volume-market trend decision point state implicit coding vectors as the historical sales volume-market trend decision point state implicit coding features.

5. The intelligent management method for cable production according to claim 4, characterized in that: The implicit coding features of the historical sales-market trend decision point states are analyzed based on the spectrum-guided adaptive aggregation of historical sales-market trend decision points to obtain the historical sales-market trend collaborative decision response coding features, including: Performing a structural analysis of the implicit state graph of the historical sales volume-market trend decision point state on the implicit coding features of the historical sales volume-market trend decision point state to obtain a historical sales volume-market trend decision point state class neighborhood matrix and a historical sales volume-market trend decision point state class degree matrix; Based on the historical sales-market trend decision point state class neighborhood matrix and the historical sales-market trend decision point state class degree matrix, calculate the historical sales-market trend decision point state Laplace matrix; The historical sales-market trend decision point state Laplace matrix is ​​spectrally decomposed and feature reconstructed to obtain the historical sales-market trend collaborative decision response coding vector as the historical sales-market trend collaborative decision response coding feature.

6. The intelligent management method for cable production according to claim 5, characterized in that: The historical sales-market trend decision point state Laplace matrix is ​​spectrally decomposed and feature reconstructed to obtain the historical sales-market trend collaborative decision response encoding vector, including: Perform spectral decomposition on the state Laplace matrix of the historical sales volume-market trend decision point to obtain a sequence of core component encoding vectors of the historical sales volume-market trend decision point; The sequence of historical sales-market trend decision point core component encoding vectors is adaptively aggregated to obtain the historical sales-market trend collaborative decision response encoding vector.

7. The intelligent management method for cable production according to claim 6, characterized in that: The historical sales-market trend decision point state Laplace matrix is ​​spectrally decomposed to obtain a sequence of core component encoding vectors of the historical sales-market trend decision point, including: Use the topological closure mechanism to optimize the historical sales volume-market trend decision point state Laplace matrix based on the cut-circular space closure to obtain the optimized historical sales volume-market trend decision point state Laplace matrix; The optimized historical sales-market trend decision point state Laplace matrix is ​​spectrally decomposed to obtain a sequence of core component encoding vectors of the historical sales-market trend decision points.

8. The intelligent management method for cable production according to claim 7, characterized in that: Based on the historical sales volume-market trend collaborative decision-making response coding features, the short-term forecast value of the cable market demand is obtained, including: The historical sales volume-market trend collaborative decision response coding vector is feature decoded and mapped to obtain the short-term forecast value of the cable market demand.

9. An intelligent management system for cable production, characterized in that: include: Information acquisition module, used to obtain historical sales data and market trend information; A sales time series coding feature extraction module is used to extract sales time series coding features from historical sales data to obtain a time series of local time series coding features of sales volume; A market trend information semantic analysis module is used to perform semantic embedding coding on the market trend information to obtain semantic embedding coding features of the market trend information; The historical sales volume-market trend semantic collaborative interaction module is used to input the time series of the market trend information semantic embedding coding features and the sales volume local temporal coding features into the historical sales volume-market trend semantic collaborative deep interaction network to obtain the historical sales volume-market trend collaborative decision response coding features; The demand forecasting module is used to obtain the short-term forecast value of the cable market demand based on the historical sales volume-market trend collaborative decision-making response coding characteristics.

Images