Electrical equipment sales customer demand prediction method and system based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based electrical equipment sales customer demand prediction method and system, and relates to the technical field of electrical equipment sales. The method comprises the following steps: multi-source data acquisition: docking a client-side data source, a product-side data source and a market-side data source through a preset data interface, and acquiring multi-source data related to electrical equipment sales; data preprocessing: performing standardization processing on the multi-source data to obtain feature data adaptive to an artificial intelligence prediction model; and artificial intelligence prediction model construction and prediction: constructing a fusion type artificial intelligence prediction model based on the feature data, and outputting a demand prediction result of the electrical equipment sales customer through the fusion type artificial intelligence prediction model. According to the invention, through multi-source data integration, industry adaptation AI model construction and prediction decision linkage, the core problems of low demand prediction precision, data island, disjunction between the model and the industry scene, and disjunction between the prediction result and the business decision in the electrical equipment sales field are effectively solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrical equipment sales, in particular to an electrical equipment sales customer demand prediction method and system based on artificial intelligence. BACKGROUND

[0002] Under the background of deep integration of digital economy and real economy, electrical equipment, as a core supporting product of industrial production and infrastructure engineering, its sales demand is influenced by multiple factors such as customer industry attribute, equipment life cycle, industry policy, market trend, etc. The industry has an increasingly urgent demand for accurate and dynamic customer demand prediction - accurate demand prediction can support inventory optimization, precise marketing and order priority sorting, which is the key to improving the efficiency of electrical equipment sales and reducing operating costs. Currently, demand prediction in the field of electrical equipment sales relies on traditional methods or simple adaptation of general artificial intelligence models: traditional methods mainly rely on sales staff's experience and judgment or basic statistical analysis (such as historical mean calculation), without fully considering key influencing factors such as industry-specific equipment replacement cycle and energy efficiency policy; although general AI models introduce data processing and algorithm prediction, they are not customized for the characteristics of the electrical equipment industry, only use basic time series models to process historical procurement data, without building a "customer-product-policy" correlation logic, and without realizing the linkage of prediction results with inventory, marketing, and order sales links.

[0003] The existing technology generally has the core defects of "demand prediction and industry scene disconnection, and prediction results difficult to land": on the one hand, without considering industry key factors and correlation logic, it is difficult to accurately tap customer's implicit demand (such as policy-driven equipment replacement demand), resulting in insufficient prediction accuracy and easy inventory and demand mismatch; on the other hand, the prediction results are output in the form of basic data, without being converted into decision-making information that can directly guide business, which is disconnected from the actual sales process, making it difficult to support inventory optimization, marketing conversion and other core business actions, and unable to meet the integrated demand of "accurate prediction + business landing" for electrical equipment sales. SUMMARY

[0004] The present application aims to provide an electrical equipment sales customer demand prediction method and system based on artificial intelligence to solve the problems raised in the background.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solution: an electrical equipment sales customer demand prediction method based on artificial intelligence, comprising the following steps: Multi-source data acquisition: through a pre-set data interface, the customer-side data source, product-side data source and market-side data source are connected to collect multi-source data related to electrical equipment sales, including customer-side data, product-side data and market-side data; Data preprocessing: standardizing the multi-source data to obtain feature data suitable for the artificial intelligence prediction model, the standardization process including data cleaning, feature engineering and data normalization; Artificial intelligence prediction model construction and prediction: based on the feature data, a fusion artificial intelligence prediction model is constructed, and the demand prediction results of the electrical equipment sales customers are output through the fusion artificial intelligence prediction model, the demand prediction results including explicit demand prediction results and implicit demand prediction results; The construction and prediction process of the fusion artificial intelligence prediction model includes: The historical purchase time series data in the feature data is processed to output the explicit demand prediction results; an electrical equipment sales knowledge graph is constructed and customer implicit demand is mined based on the electrical equipment sales knowledge graph to output the implicit demand prediction results; the explicit demand prediction results and the implicit demand prediction results are dynamically adjusted with the inventory cost reduction rate and the policy change adaptation degree as the optimization target; Demand prediction result output and sales decision linkage: the demand prediction results are converted into sales decision support information, and are linked with inventory management, precision marketing and order management links, the sales decision support information including quantitative prediction reports, stock preparation suggestions, marketing plans and order priority sorting results.

[0006] Preferably, the multi-source data collection step includes: Customer-side data collection: collecting the customer-side data from a customer relationship management system; Product-side data collection: collecting the product-side data from an enterprise resource planning system; Market-side data collection: collecting the market-side data from an industry policy platform and an industry database; The customer-side data is used to reflect the attributes and behavior information related to the customer and electrical equipment procurement, the product-side data is used to reflect the product characteristics and inventory status information of electrical equipment, and the market-side data is used to reflect the electrical equipment industry policy and market trend information.

[0007] Preferably, the customer-side data collection step includes: Customer basic information collection: collecting the industry type and project scale data of the customer, the industry type including industrial manufacturing type, new energy infrastructure type and commercial building type, and the project scale including annual output value and infrastructure investment amount; Customer historical purchase data collection: collecting the device type, purchase quantity, purchase time and service life data of the purchased equipment of the customer's past purchase of electrical equipment; Customer feedback data collection: collecting the fault record and equipment replacement intention data of the customer on the purchased electrical equipment.

[0008] Preferably, the data preprocessing step comprises: Data cleaning: removing outliers in the multi-source data using an outlier removal algorithm based on statistical rules, and filling missing values in the multi-source data using a missing value filling algorithm based on industry average values; Feature engineering: extracting customer features, product features and market features from the cleaned multi-source data, the customer features including associated features of industry type and equipment service life, associated features of project size and purchase frequency, the product features including associated features of energy efficiency level and applicable scene, associated features of inventory turnover rate and replenishment cycle, and the market features including associated features of policy validity period and subsidy ratio, associated features of market growth rate and competitor inventory; Data normalization: mapping the extracted customer features, product features and market features to a preset numerical interval using a preset normalization algorithm to obtain the feature data.

[0009] Preferably, the processing of the historical purchase time series data in the feature data to output the explicit demand prediction result comprises: Time series model construction: constructing a time series prediction model using a deep learning model, the input of the time series prediction model being the historical purchase time series data in the feature data, and the output being the electrical equipment purchase quantity and purchase time in a future preset period; Model training and verification: taking 70% of the historical purchase time series data as a training set and 30% as a verification set, training the time series prediction model using a preset optimizer, taking the mean square error of the predicted value and the actual value as the loss function, and iteratively training until the loss function value is less than a preset threshold; Explicit demand output: inputting the feature data into the trained time series prediction model to obtain the explicit demand prediction result.

[0010] Preferably, the construction of the electrical equipment sales knowledge graph and the mining of customer implicit demand based on the electrical equipment sales knowledge graph to output the implicit demand prediction result comprises: Knowledge graph node construction: constructing nodes of the electrical equipment sales knowledge graph, the nodes including customer nodes, electrical equipment nodes and market factor nodes, the customer nodes being associated with customer basic information, the electrical equipment nodes being associated with equipment parameter information, and the market factor nodes being associated with industry policy information; Node association relationship definition: defining the association relationship between the customer nodes, electrical equipment nodes and market factor nodes, the association relationship including matching relationship between customer industry type and electrical equipment applicable scene, and triggering relationship between industry policy and electrical equipment demand; Inferential demand extraction: extracting equipment demands not explicitly expressed by the customer from the feature data based on the association relationship to form the inferential demand prediction result.

[0011] Preferably, the demand prediction result output and sales decision linkage step includes: Quantitative prediction report generation: organizing the demand prediction result into the quantitative prediction report containing customer demand priority, equipment demand type proportion, and demand time distribution; Inventory linkage processing: synchronizing the quantitative prediction report to an inventory management system to generate the restocking suggestion in combination with existing electrical equipment inventory quantity, the restocking suggestion including equipment type and restocking quantity to be replenished; Precise marketing scheme generation: generating the marketing scheme adapting to potential demands of the customer based on the inferential demand prediction result, the marketing scheme including equipment replacement scheme and policy interpretation content; Order priority sorting: sorting customer orders according to the demand urgency in the demand prediction result to generate the order priority sorting result, the demand urgency being determined based on customer project progress and equipment service life.

[0012] The application also provides an electrical equipment sales customer demand prediction system based on artificial intelligence, including: Perception layer: connecting customer-side data sources, product-side data sources, and market-side data sources through a preset data interface, collecting multi-source data related to electrical equipment sales, the multi-source data including customer-side data, product-side data, and market-side data; Data layer: performing standardization processing on the multi-source data to obtain feature data adapting to the artificial intelligence prediction model, the standardization processing including data cleaning, feature engineering, and data normalization; AI calculation layer: based on the feature data, constructing a fusion-type artificial intelligence prediction model, outputting demand prediction results of electrical equipment sales customers through the fusion-type artificial intelligence prediction model, the demand prediction results including explicit demand prediction results and inferential demand prediction results; The construction and prediction process of the fusion-type artificial intelligence prediction model includes: processing historical purchase time series data in the feature data to output explicit demand prediction results; constructing an electrical equipment sales knowledge graph and mining customer inferential demands based on the electrical equipment sales knowledge graph to output inferential demand prediction results; and dynamically adjusting the explicit demand prediction results and the inferential demand prediction results with inventory cost reduction rate and policy change adaptation degree as optimization targets; An application layer: transform the demand prediction result into sales decision support information, and link with inventory management, precision marketing and order management links, the sales decision support information including quantitative prediction report, warehousing suggestion, marketing scheme and order priority sorting result.

[0013] The application further provides an electronic device, which is a physical device, and the electronic device comprises: a processor and a memory connected in communication with the processor; The memory is used to store executable instructions executed by the processor, and the processor is used to execute the executable instructions to realize the AI-based electrical equipment sales customer demand prediction method as described above.

[0014] The application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by the processor to realize the AI-based electrical equipment sales customer demand prediction method as described above.

[0015] Compared with the prior art, the application has the following beneficial effects: The application effectively solves the core problems of low demand prediction accuracy, data silos, disconnection between models and industry scenarios, and disconnection between prediction results and business decisions in the electrical equipment sales field through multi-source data integration, industry-adapted AI model construction and prediction decision linkage. Multi-source data collection covers the full dimensions of customers, products and markets, industry-specific associated features are extracted in the preprocessing link, and a fusion AI model accurately mines explicit and implicit demands, and can dynamically adapt to policy and market changes. The prediction results are further linked with inventory management, precision marketing and order management links, which not only improves the comprehensiveness and dynamic adaptability of demand prediction, but also reduces the risk of inventory and demand mismatch, improves sales conversion efficiency, and can also be expanded to adapt to other industrial product sales prediction scenarios, with both industry-specificity and general adaptability, providing full-process precision decision support for electrical equipment sales business. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A main flowchart of an AI-based electrical equipment sales customer demand prediction method provided by the application; Figure 2 A structural schematic diagram of an AI-based electrical equipment sales customer demand prediction system provided by the application; Figure 3 A structural schematic diagram of an electronic device provided by the application. DETAILED DESCRIPTION

[0017] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0018] The terminal can be a mobile phone, a tablet computer, a palm computer PDA, a notebook computer or a desktop computer, and of course, can also be other devices with similar functions, and the present embodiment is not limited thereto.

[0019] Please refer to Figure 1 The present application provides an electrical equipment sales customer demand prediction method based on artificial intelligence, which is applied to, comprising: Step 100, multi-source data acquisition: through the preset data interface to the customer side data source, the product side data source and the market side data source, collect the multi-source data related to the electrical equipment sales, the multi-source data includes customer side data, product side data and market side data.

[0020] Among them, "preset data interface" refers to the interface pre-configured by the system, which conforms to the industrial data transmission protocol, including hardware interface and software data interaction interface supporting industrial Ethernet protocol and time sensitive network (TSN) protocol, which can realize stable docking with different data sources; "customer side data source" refers to the system or platform storing customer related information, specifically including customer relationship management system (CRM), which records all information related to customer and electrical equipment purchase; "product side data source" refers to the system storing electrical equipment product information and enterprise internal operation data, specifically including enterprise resource planning system (ERP), which covers product parameters, inventory status and other core data; "market side data source" refers to the platform providing external information such as electrical equipment industry policy and market trend, specifically including industry policy platform (such as the official website of the State Energy Administration) and industry database (such as electrical equipment market analysis database); "multi-source data" refers to all data related to electrical equipment sales collected from customer side data source, product side data source and market side data source, which covers customer, product and market three dimensions, and is the basis for subsequent data processing and AI prediction; "customer side data" refers to the attribute and behavior information reflecting customer and electrical equipment purchase collected from customer relationship management system, including customer basic information, historical purchase data and customer feedback data; "product side data" refers to the information reflecting electrical equipment product characteristics and inventory status collected from enterprise resource planning system, including product parameter data, inventory data and price data; "market side data" refers to the information reflecting electrical equipment industry policy and market trend collected from industry policy platform and industry database, including policy data, market trend data and competitive product data.

[0021] In addition, it should be noted that the key of step S100 is not simply collecting basic data such as "purchase quantity and product model", but for the characteristics of electrical equipment industry, the industry key influencing factors such as "equipment service life, energy efficiency policy and infrastructure project scale" are included in the collection range - electrical equipment has a fixed replacement cycle of 8-10 years, energy efficiency policy (such as GB20052-2020 "Power Transformer Energy Efficiency Limiting Value and Energy Efficiency Grade") directly determines the customer's selection tendency of equipment, and the scale of infrastructure project is related to the magnitude of equipment demand. The inclusion of these data makes the subsequent prediction more accurately match the actual demand of the industry, avoiding the problem of "insufficient scene adaptability" of general data collection scheme.

[0022] In a possible implementation, the preset data interface can support dynamic expansion, in addition to connecting with the customer relationship management system, the enterprise resource planning system, and the industry policy platform, the interface can also access an e-commerce platform (collect customer equipment browsing records), a logistics system (collect equipment delivery cycle data), and further enrich the data dimensions; meanwhile, the interface can configure data filtering rules to preliminarily remove duplicate data (such as repeated purchase records of the same customer) in the collection stage, thereby reducing the workload of subsequent data cleaning.

[0023] Specifically, the step 100 further includes: Step 110, collecting customer-side data: collecting the customer-side data from the customer relationship management system; Step 120, collecting product-side data: collecting the product-side data from the enterprise resource planning system; Step 130, collecting market-side data: collecting the market-side data from the industry policy platform and the industry database; The customer-side data is used to reflect the attributes and behavior information of the customer related to the electrical equipment purchase, the product-side data is used to reflect the product characteristics and inventory state information of the electrical equipment, and the market-side data is used to reflect the industry policy and market trend information of the electrical equipment.

[0024] Among them, "customer relationship management system" refers to the system for managing the interaction between enterprises and customers, storing the whole cycle information of customers from the first contact to long-term cooperation, which is the core source of customer-side data; "enterprise resource planning system" refers to the system integrating the operation links of enterprise production, inventory, procurement, etc., which can reflect the product status and inventory level of electrical equipment in real time, and is the only source of product-side data; "industry policy platform" refers to the official or authoritative platform (such as the official website of the State Energy Administration, the platform of the Equipment Industry Department of the Ministry of Industry and Information Technology) that releases relevant policies of electrical equipment (such as energy efficiency standards, subsidy policies), and "industry database" refers to the professional database (such as the electrical equipment market report database released by industry research institutions) that collects data such as the market size, growth rate and competitive product dynamics of electrical equipment, which together constitute the source of market-side data; "customer attributes and behavior information related to electrical equipment procurement" refers to customer information that directly affects customer procurement decisions, including customer industry affiliation (determining equipment type demand), project size (determining procurement quantity), historical procurement habits (determining procurement cycle), etc.; "electrical equipment product characteristics and inventory status information" refers to information describing the parameters and inventory level of electrical equipment, product characteristics including equipment model, energy efficiency level, applicable scenario (determining the matching degree of equipment and customer demand), and inventory status including existing inventory quantity, inventory turnover rate (determining the urgency of inventory); "electrical equipment industry policy and market trend information" refers to external information that affects the overall demand of the electrical equipment market, industry policy directly guiding customer selection (such as energy efficiency policy eliminating high-energy-consuming equipment), and market trend reflecting the direction of demand growth (such as new energy infrastructure driving demand growth of distribution equipment). From the implementation effect, the above steps are designed through "data source collection + data positioning clear", which not only ensures the comprehensiveness of data (covering customers, products, and markets), but also ensures the relevance of data (all data sources are related to "electrical equipment sales demand"), solving the problem of "data source confusion and weak relevance of data and prediction target" in the existing collection scheme.

[0025] In addition, it needs to be explained that its subdivision logic is highly consistent with the business process of electrical equipment sales: customer-side data corresponds to "who buys" (customer attributes), product-side data corresponds to "what to sell" (device characteristics), and market-side data corresponds to "why to buy" (policy and trend driven), which form a "customer-product-market" demand-driven closed loop, enabling the subsequent AI model to predict demand from the "root of demand generation", rather than just based on the surface association of historical data.

[0026] In one possible implementation, the collection of each data source can set "priority and update frequency": the update frequency of customer-side data (especially historical purchase data and customer feedback data) is set to "real-time" (synchronized within 1 hour after the customer generates a purchase behavior or feedback), ensuring the capture of the latest customer demand dynamics; the update frequency of product-side data (inventory data and price data) is set to "every 4 hours", balancing real-time performance and system resource consumption; the update frequency of market-side data (policy data) is set to "once a day" (as policy release has periodicity, there is no need for high-frequency update), and the update frequency of market trend data is set to "once a week"; at the same time, when the industry policy platform releases an emergency policy (such as the advance implementation of energy efficiency standards), the system can trigger an "emergency update mechanism" to complete policy data collection and synchronization within 15 minutes.

[0027] For example, in one possible implementation, collecting multi-source data includes: S110 (collection of customer-side data): collecting the industry type (new energy infrastructure type), project scale (annual output value of 50 billion, new energy power station construction project investment of 1 billion), historical purchase data (purchased 10 10kV transformers each year from 2021 to 2022, and the purchase time was concentrated in Q3), and customer feedback data (the purchased transformer in 2022 operates normally, and it is planned to add equipment in 2023 for expansion) of customer A (a new energy vehicle enterprise) from a customer relationship management system; S120 (collection of product-side data): collecting the product characteristics (model S13-1000kVA, energy efficiency level one, suitable for new energy power station, unit price 220,000 yuan) of 10kV level one energy efficiency transformer and inventory status (existing inventory 8, inventory turnover rate 0.8 in the last 3 months) from an enterprise resource planning system; S130 (collection of market-side data): collecting 2023 new energy infrastructure subsidy related policies (subsidy ratio 15%, covering 10kV and above distribution equipment, effective period 2023-2028) from the official website of the National Energy Administration (industry policy platform), and collecting 2023 Q1 electrical equipment market trends (new energy supporting distribution equipment market growth rate 25%, competitor's 10kV transformer inventory 12, unit price 23,000 yuan) from an industry database, forming a classified multi-source data set.

[0028] Further, step 110 further includes: Step 111, collection of customer basic information: collecting the industry type and project scale data of the customer, wherein the industry type includes industrial manufacturing type, new energy infrastructure type, and commercial building type, and the project scale includes annual output value and infrastructure investment; Step 112, collection of customer historical purchase data: collecting the equipment type, purchase quantity, purchase time, and service life data of the purchased equipment of the customer's past purchase of electrical equipment; Step 113, customer feedback data collection: collect customer's fault records and equipment replacement intention data of purchased electrical equipment.

[0029] Among them, "industry type" refers to the core business field of the customer, which is divided into industrial manufacturing type (such as automobile factory, steel plant, which needs high-power electrical equipment), new energy infrastructure type (such as photovoltaic power station, wind power station, which needs special equipment adapted to new energy scene), commercial building type (such as shopping mall, office building, which needs low-voltage power distribution equipment), different industry types correspond to completely different equipment demand types; "project scale" refers to the data reflecting the business volume of the customer, "annual output value" is directly related to the equipment procurement budget of the customer (the higher the annual output value, the more sufficient the procurement budget), "infrastructure investment" is for customers with infrastructure projects (such as new factory construction of new energy automobile enterprises), which directly determines the equipment procurement quantity; "equipment type of customer's past procurement of electrical equipment" refers to the specific electrical equipment model (such as 10kV transformer, 0.4kV switch cabinet) purchased by the customer in history, which reflects the equipment selection preference of the customer; "procurement quantity" refers to the equipment quantity purchased by the customer each time, "procurement time" refers to the specific quarter or month of the customer's historical procurement, which can extract the customer's procurement cycle rule (such as a customer purchases every Q4) in combination; "service life of purchased equipment" refers to the time of the current equipment used by the customer, which can be combined with the replacement cycle of electrical equipment of 8-10 years to predict the replacement demand node of the customer; "customer's fault record of purchased electrical equipment" refers to the device running fault information (such as transformer oil leakage, switch cabinet tripping) feedback by the customer, and the customer with high fault frequency is more likely to produce replacement demand in advance; "equipment replacement intention" refers to the explicit expression of the customer's equipment update plan (such as "plan to replace old equipment in Q3 2023"), which is a direct signal for the conversion of implicit demand to explicit demand. The design idea of the above steps is to "collect data from the 'cause' of customer demand", rather than just collect surface data such as "procurement quantity", so that the subsequent model can predict "what to buy and how much to buy" based on the logic of "why to buy", greatly improving the rationality of the underlying logic of prediction.

[0030] In a possible implementation, the customer basic information collection can increase the "customer credit level" dimension (such as AAA level, AA level), the customer with high credit level (AAA level) has more stable procurement demand and low performance risk, and can be given higher demand priority in subsequent prediction; the customer historical procurement data collection can increase the "use scene subdivision of the procurement equipment" (such as 10kV transformer purchased by customer A is used for power station main transformer, and 10kV transformer purchased by customer B is used for workshop power distribution), the scene subdivision can further accurately match the equipment model (such as the main transformer needs a 10kV-2000kVA transformer with higher power, and the workshop power distribution needs a 10kV-1000kVA transformer); the customer feedback data collection can increase the "equipment maintenance record" (such as the transformer of customer C is maintained once every half year, and the transformer of customer D is maintained once every year), and the customer with low maintenance frequency and no failure can prolong the equipment replacement demand cycle (such as from 8 years to 9 years).

[0031] For example, in a possible implementation, the customer-side data collection includes: S111 (customer basic information collection): collect the industry type (new energy infrastructure type) and project scale (annual output value of 50 billion, new power station construction investment of 1.2 billion) of customer A (new energy vehicle enterprise), the industry type (industrial manufacturing type) and project scale (annual output value of 20 billion, no new infrastructure project) of customer B (automobile parts factory); S112 (customer historical procurement data collection): collect the past procurement equipment type (10kV-1000kVA transformer purchased in 2021-2022), procurement quantity (10 units per year), procurement time (every year from September to October, Q4), and equipment service life (the equipment purchased in 2021 has been used for 2 years, and the equipment purchased in 2022 has been used for 1 year) of customer A, and the past procurement equipment type (S9 type 8kV transformer purchased in 2020), procurement quantity (5 units), procurement time (June 2020), and equipment service life (used for 3 years) of customer B; S113 (customer feedback data collection): collect the failure record (no failure) and equipment replacement intention (“12 new 10kV transformers are needed for the new power station in 2023”) of customer A, and the failure record (switch cabinet tripping once in 2022, and has been repaired) and equipment replacement intention (“no plan to replace, but pay attention to higher energy-efficient equipment”) of customer B, to form a complete customer-side data dimension.

[0032] In step 200, data preprocessing: performing standardization processing on the multi-source data to obtain feature data suitable for an artificial intelligence prediction model, and the standardization processing includes data cleaning, feature engineering and data normalization.

[0033] "Standardization processing" refers to a series of standardized operations performed on multi-source data, specifically including three sub-steps: data cleaning, feature engineering, and data normalization. Its core objectives are to eliminate data defects, strengthen the correlation between data and prediction targets, and unify data formats. "Data cleaning" refers to the operation of processing outliers, missing values, and duplicate values ​​in multi-source data, with the aim of removing data noise and ensuring data accuracy. "Feature engineering" refers to the operation of extracting key features related to the prediction of customer demand for electrical equipment from the cleaned multi-source data, with the aim of strengthening the data's support for the prediction results and reducing redundant information. "Data normalization" refers to the operation of mapping the extracted feature data to a preset numerical range, with the aim of eliminating the interference of differences in the magnitude of different feature data on AI model training. "Artificial intelligence prediction model" refers to the algorithm model used to realize the prediction of customer demand for electrical equipment sales, specifically a fusion model that integrates time series prediction, knowledge graph association, and reinforcement learning optimization. "Feature data" refers to the structured data that has been standardized and can be directly input into the artificial intelligence prediction model, which includes three major categories of key information: customer characteristics, product characteristics, and market characteristics.

[0034] In one possible implementation, the outlier removal algorithm for data cleaning can be adaptively adjusted: for historical customer purchase data, if the customer is a large industrial enterprise (annual output value exceeding 1 billion), the standard deviation coefficient based on the 3σ principle can be adjusted to 1.2 (allowing for greater fluctuations in purchase volume); for small and medium-sized commercial customers (annual output value less than 100 million), the standard deviation coefficient can be adjusted to 0.8 (strictly controlling the range of outliers); at the same time, missing value filling can adopt the method of "industry average + customer-specific correction". For example, if the project scale data of customer B (an industrial manufacturing enterprise) is missing, the average project scale of the industrial manufacturing industry (annual output value of 2 billion) is first used to fill it, and then it is corrected to an annual output value of 1.8 billion based on customer B's historical purchase volume (an average of 5 transformers purchased per year), to ensure that the filled data is more in line with the actual situation of the customer.

[0035] Specifically, step 200 also includes: Step 210, Data Cleaning: Outliers in the multi-source data are removed using an outlier removal algorithm based on statistical rules, and missing values ​​in the multi-source data are filled using a missing value filling algorithm based on industry averages. Step 220, Feature Engineering: Extract customer features, product features, and market features from the cleaned multi-source data. The customer features include the correlation between industry type and equipment service life, and the correlation between project scale and purchase frequency. The product features include the correlation between energy efficiency level and applicable scenarios, and the correlation between inventory turnover rate and replenishment cycle. The market features include the correlation between policy validity period and subsidy ratio, and the correlation between market growth rate and competitor inventory. Step 230, Data Normalization: The extracted customer features, product features and market features are mapped to a preset numerical range using a preset normalization algorithm to obtain the feature data.

[0036] Among them, the "outlier removal algorithm based on statistical rules" refers to an algorithm that uses statistical characteristics of data (such as mean and standard deviation) to identify and remove data that deviates from the normal range. In this embodiment, considering the characteristics of electrical equipment procurement data being "stable in magnitude and with small fluctuations," the 3σ principle is adopted (outliers are defined as data exceeding the mean ± 3 times the standard deviation) to avoid misjudging normal procurement fluctuations; the "missing value imputation algorithm based on industry mean" refers to an algorithm that uses the average level of the industry to which the data belongs to imputation when a certain data is missing, ensuring that the imputation data conforms to the general industry rules; "cleaned multi-source data" refers to noise-free data after outlier removal and missing value imputation, which is the input basis for feature engineering; "customer characteristics" refers to the correlation information extracted from customer-side data that is directly related to demand forecasting. "Correlation characteristics between industry type and equipment service life" reflects the customer's replacement demand probability (e.g., industrial manufacturing + 8 years of equipment service life → high replacement demand), and "correlation characteristics between project scale and procurement frequency" reflects the customer's procurement volume potential (e.g., annual output value of 5 billion + 2 procurements per year → high procurement volume); "product characteristics" refers to The correlation information extracted from product-side data includes: "Correlation characteristics between energy efficiency level and applicable scenarios" reflecting the matching degree between products and customer needs (e.g., Level 1 energy efficiency + new energy power station → matching new energy customers); and "Correlation characteristics between inventory turnover rate and replenishment cycle" reflecting the ability of inventory to support demand (e.g., inventory turnover rate of 0.8 + replenishment cycle of 1 month → need to prepare inventory in advance). "Market characteristics" refer to the correlation information extracted from market-side data, such as "Correlation characteristics between policy validity period and subsidy ratio" reflecting the duration and strength of the policy's driving effect on demand (e.g., 5-year validity period + 15% subsidy). →Long-term high demand), "Correlation characteristics between market growth rate and competitor inventory" reflects the impact of market competition on demand (e.g., 20% growth rate + low competitor inventory → intense demand competition); "Preset normalization algorithm" refers to a pre-configured algorithm used to unify the data volume. In this embodiment, Min-Max normalization (mapping the data to the [0,1] interval) is used. Because the volume of electrical equipment data (such as annual output value and subsidy ratio) varies greatly (annual output value from 100 million to 10 billion, subsidy ratio from 0% to 20%), normalization can avoid the model being overly biased towards data with large volume. The above steps ensure that the pre-processed data can support the accuracy of subsequent AI predictions to the greatest extent by "adapting each preprocessing step to the characteristics of electrical equipment data" rather than using a general data processing scheme.

[0037] In one possible implementation, data cleaning can add a "duplicate data merging" operation: for multiple duplicate purchase records of the same customer (e.g., customer A has two records of "purchasing 1 transformer" entered on the same day), the system can automatically merge them into one record of "purchasing 2 transformers" and mark the merging time to avoid duplicate data interfering with the model; feature engineering can add "feature weight allocation": assign different weights to different features based on their impact on demand forecasting (e.g., "equipment service life" weight 0.3, "subsidy ratio" weight 0.25, "industry type" weight 0.2, ensuring that the core influencing factors have a higher proportion); data normalization can support "dimension-specific interval adjustment": for "equipment service life" (range 0-10 years), the mapping interval is set to [0,1]; for "subsidy ratio" (range 0%-20%), the mapping interval is set to [0,0.8], avoiding uneven data distribution after normalization due to an excessively large range of data in a certain dimension.

[0038] For example, in one feasible implementation, the preprocessed data includes: S210 (Data Cleaning): For Customer B's procurement data (historical average procurement volume of 5 units, standard deviation of 1 unit), remove one abnormal record of "procuring 15 transformers" (exceeding the range of 5±3×1=8 units); For Customer C's missing project scale data (commercial building customer), fill it with the average annual output value of the commercial building industry (800 million). S220 (Feature Engineering): Extract customer features from the cleaned data (Customer A: New energy infrastructure type + 2-year equipment service life + 5 billion annual output value → associated feature "New energy + 2 years + 5 billion"; Customer B: Industrial manufacturing + 3-year equipment service life + 2 billion annual output value → associated feature "Industry + 3 years + 2 billion"), product features (10kV transformer: Level 1 energy efficiency + applicable scenarios for new energy power stations + 8 units in stock → associated feature "Level 1 energy efficiency + new energy scenarios + 0.8 turnover rate"), and market features (New energy subsidies: 15% ratio + 5-year validity period + 25% growth rate → associated feature "15% subsidy + 5 years + 25% growth rate"). S230 (Data Normalization): Using Min-Max normalization, Customer A's "5 billion annual output value" is mapped to 0.8 (industry annual output value range 0-6 billion), "2-year equipment service life" is mapped to 0.2 (equipment service life range 0-10 years), "15% subsidy ratio" is mapped to 0.75 (subsidy ratio range 0%~20%), and "25% market growth rate" is mapped to 0.8 (growth rate range 0%~30%), forming feature data that can be input into AI models.

[0039] Step 300, Artificial Intelligence Prediction Model Construction and Prediction: Based on the feature data, a fusion artificial intelligence prediction model is constructed, and the demand prediction results of electrical equipment sales customers are output through the fusion artificial intelligence prediction model. The demand prediction results include explicit demand prediction results and implicit demand prediction results.

[0040] The construction and prediction process of the fusion-based artificial intelligence prediction model includes: The historical procurement time series data in the feature data is processed to output explicit demand forecast results; an electrical equipment sales knowledge graph is constructed and implicit customer needs are mined based on the electrical equipment sales knowledge graph to output implicit demand forecast results; the explicit demand forecast results and implicit demand forecast results are dynamically adjusted with inventory cost reduction rate and policy change adaptability as optimization objectives.

[0041] Among them, the "fusion-type AI prediction model" refers to an AI model that integrates three technical paths: time series prediction, knowledge graph association, and reinforcement learning optimization. Unlike single models, it can simultaneously cover explicit and implicit demand prediction and dynamic optimization. "Feature data" refers to the standardized data output from step 200, which is the input basis for model training and prediction. "Historical procurement time series data" refers to the time dimension data of customers' past purchases in the feature data, including time-related records of purchase time, purchase quantity, and purchased equipment type, which is the core basis for predicting explicit demand. "Explicit demand prediction results" refer to the customer's clearly defined equipment demand results that can be directly derived from historical data, specifically including the purchase quantity, purchase time, and equipment type of electrical equipment within a future preset period. "Electrical equipment sales knowledge graph" refers to... It is a semantic network built around the sales scenario of electrical equipment, containing nodes and relationships, used to mine implicit connections between data; "implicit demand prediction results" refer to potential equipment demand results that customers have not explicitly expressed but can be inferred through industry relationships, specifically including equipment replacement demand and suggestions for suitable equipment types; "inventory cost reduction rate" refers to the proportion of inventory reduction that can be achieved after the prediction results are implemented, which is one of the core objectives of reinforcement learning optimization; "policy change adaptability" refers to the degree to which the prediction results adapt to changes in industry policies (such as the degree of matching of prediction results after the adjustment of subsidy policies), which is another core objective of reinforcement learning optimization; "demand prediction results" refer to the comprehensive demand results finally output after explicit prediction, implicit mining, and dynamic optimization, including explicit demand prediction results and implicit demand prediction results.

[0042] In one possible implementation, time series forecasting can employ a hybrid model of "LSTM + Transformer": LSTM is used to capture the long-term time dependence of procurement data (such as the peak procurement season in Q4 of each year), while the attention mechanism of Transformer is used to focus on key time nodes (such as the month when subsidy policies take effect or the month when customer projects are completed). During model training, the joint loss function is "mean squared error (MSE) between predicted and actual values ​​+ demand time matching degree" (MSE weight 0.6, time matching degree weight 0.4), further improving the accuracy of explicit demand forecasting. At the same time, the association relationships of the knowledge graph can support dynamic updates. When a new energy efficiency policy is introduced (such as GB20052-2024 introduced in 2024), the system automatically adds the association relationship of "new energy efficiency standard - equipment elimination" (such as "GB20052-2024 → elimination of level 2 energy efficiency equipment from 2025"), ensuring the timeliness of implicit demand mining.

[0043] The process of processing the historical procurement time-series data in the feature data to output explicit demand forecast results includes: Step 311, Time Series Model Construction: A time series prediction model is constructed using a deep learning model. The input of the time series prediction model is the historical procurement time series data in the feature data, and the output is the procurement quantity and procurement time of electrical equipment within a future preset period. Step 312, Model Training and Validation: Using 70% of the historical procurement time series data as the training set and 30% as the validation set, the time series prediction model is trained using a preset optimizer, with the mean square error between the predicted value and the actual value as the loss function, and iterative training is performed until the loss function value is less than a preset threshold. Step 313, Explicit Demand Output: Input the feature data into the trained time series prediction model to obtain the explicit demand prediction result.

[0044] Here, "deep learning model" refers to a machine learning model based on neural networks with a deep structure. In this embodiment, a Long Short-Term Memory (LSTM) network is used because LSTM can effectively capture the long-term dependencies in time series data (such as the cyclical nature of customer purchases in Q4 of each year), avoiding the "large long-term prediction bias" problem of traditional time series models (such as ARIMA). "Time series prediction model" refers to a model specifically designed to process time-dimensional data and predict future trends. Its input is "historical purchase time series data" (purchase records arranged in chronological order), and its output is "the purchase quantity and purchase time of electrical equipment within a future preset period" (explicit requirements). "Training set" refers to the dataset used to train model parameters, and "validation set" refers to the dataset used to validate the model training effect and adjust model parameters. In this embodiment, they are divided in a 7:3 ratio (7... The data allocation (0% historical data for training, 30% for validation) conforms to the typical data partitioning ratio for machine learning models, ensuring that the model fully learns historical patterns while validating its generalization ability. The "preset optimizer" refers to a pre-configured algorithm used to optimize the model training process; in this embodiment, the Adam optimizer is used because it converges quickly and is less sensitive to the learning rate, making it suitable for scenarios with "limited sample size" in electrical equipment procurement data. The "mean squared error between predicted and actual values" refers to the average of the squared differences between the model's prediction and the actual historical data, serving as the core loss function for measuring prediction accuracy. The "preset threshold" refers to a pre-defined pass / fail standard for the loss function; in this embodiment, it is set to 5% (i.e., model training is complete when the mean squared error is <5%), ensuring that the model's prediction accuracy meets the actual business needs of electrical equipment sales (e.g., a deviation of ±0.5 units is allowed when predicting 10 units). These steps, by ensuring that "model selection and parameter settings are adapted to the characteristics of electrical equipment procurement data," avoid the "scenario mismatch" problem of general models and ensure the accuracy of explicit demand predictions.

[0045] In one possible implementation, the time series model construction can include a "feature fusion layer": in addition to inputting historical procurement time series data, the "market features" extracted in step 200 (such as subsidy ratio, market growth rate) can be used as auxiliary input. The feature fusion layer combines time features with market features (such as "Q4 procurement cycle + 15% subsidy") to further improve prediction accuracy. Model training and validation can adopt a "cross-validation" method: historical data is divided into 5 batches, with 4 batches used as the training set and 1 batch used as the validation set each time, and the model accuracy is validated 5 times in a loop to avoid model overfitting caused by a single data partition. The explicit demand output can be labeled with "demand confidence level" (such as labeling the 13 procurement forecasts for customer A as "confidence level 92%" and the 5 procurement forecasts for customer B as "confidence level 85%"). Prediction results with a confidence level below 80% need to be manually reviewed to ensure the reliability of the output results.

[0046] For example, in one feasible implementation, achieving explicit demand forecasting includes: S311 (Time Series Model Construction): The LSTM model is used to construct a time series forecasting model. The input layer is set to "customer's quarterly purchase volume, equipment service life, and quarterly subsidy ratio for the past 3 years", and the output layer is set to "purchase volume and purchase month for the next 6 months (Q3-Q4)". S312 (Model Training and Validation): Using customer A's historical data from 2020 to 2022 (70%, i.e., data from 2020 to 2021) as the training set and the 2022 data (30%) as the validation set, the Adam optimizer is used for training, with mean squared error as the loss function. After 50 rounds of iterative training, the loss function value drops to 4.2% (<5% preset threshold), and the model training is complete. S313 (Explicit Demand Output): Input the characteristic data of customer A (the service life of the equipment in Q2 2023, the subsidy ratio of 15%) into the trained model, and output the explicit demand prediction result: "Purchase 13 10kV-1000kVA transformers in September-October 2023 (Q4), with a confidence level of 93%", thus completing the explicit demand prediction.

[0047] Furthermore, the step of constructing a knowledge graph for electrical equipment sales and mining latent customer needs based on the knowledge graph to output latent demand prediction results includes: Step 321, Knowledge Graph Node Construction: Construct the nodes of the electrical equipment sales knowledge graph. The nodes include customer nodes, electrical equipment nodes, and market factor nodes. The customer nodes are associated with basic customer information, the electrical equipment nodes are associated with equipment parameter information, and the market factor nodes are associated with industry policy information. Step 322, Node Relationship Definition: Define the relationship between the customer node, electrical equipment node, and market factor node. The relationship includes the matching relationship between customer industry type and electrical equipment applicable scenarios, and the triggering relationship between industry policies and electrical equipment demand. Step 323, Implicit Demand Extraction: Based on the aforementioned relationship, extract the customer's unexpressed equipment requirements from the feature data to form the implicit demand prediction result.

[0048] The "nodes of the electrical equipment sales knowledge graph" refer to the basic units carrying information in the knowledge graph, divided into three categories: customer nodes, electrical equipment nodes, and market factor nodes. Each type of node is associated with its core attributes (customer nodes are associated with industry and scale, equipment nodes with model and energy efficiency, and market factor nodes with policies and trends), ensuring that the node information can support the definition of the relationships. "Customer basic information" refers to the core attribute data of customer nodes (such as industry type and project scale), "equipment parameter information" refers to the core attribute data of electrical equipment nodes (such as model, energy efficiency rating, and applicable scenarios), and "industry policy information" refers to the core attribute data of market factor nodes (such as policy type, validity period, and subsidies). The "proportion" refers to the logical connection between different types of nodes. The "matching relationship between customer industry type and electrical equipment application scenario" reflects "who the customer is → what equipment they need" (e.g., new energy industry → equipment applicable to new energy power plants). The "triggering relationship between industry policy and electrical equipment demand" reflects "policy-driven → what equipment the customer needs" (e.g., energy efficiency policy upgrade → phasing out S9 type → demand for S13 type). "Equipment demand not explicitly expressed by the customer" refers to demand that the customer has not directly mentioned but can be deduced through the node relationship (e.g., customer B uses S9 type equipment + energy efficiency policy phases out S9 type → needs to replace with S13 type). This type of demand is an important supplement to explicit demand and directly affects the sales conversion rate.

[0049] In one possible implementation, the knowledge graph node construction can include "competitor nodes": collecting parameter information of competitor equipment (such as the energy efficiency and price of a competitor's S13 transformer), forming a four-node network of "customer-own equipment-competitor equipment-market factors," which can further uncover the implicit demand of "customer using competitor equipment + own equipment is better → replacement with own equipment"; the node relationship definition can add "time dimension relationship" (such as "3 months before policy expiration → customer purchases in advance"), making the relationship more in line with the time pattern of electrical equipment procurement; the extraction of implicit demand can add "demand urgency classification" (such as "replacement demand 1 year before policy elimination → high urgency, 3 years before policy elimination → low urgency"), providing a basis for subsequent order ranking.

[0050] For example, in one feasible implementation, implicit requirement mining includes: S321 (Knowledge Graph Node Construction): Construct customer nodes (Customer B: Industrial manufacturing type, annual output value of 2 billion), electrical equipment nodes (S9 type 8kV transformer: Level 2 energy efficiency, applicable to industrial workshops; S13 type 10kV transformer: Level 1 energy efficiency, applicable to industrial + new energy), and market factor nodes (Energy efficiency policy: Level 2 energy efficiency equipment will be phased out in 2024, no subsidies). S322 (Node Relationship Definition): Defines the relationship between "Industrial Manufacturing Type → Applicable Equipment in Industrial Workshops", "Elimination of Level 2 Energy Efficiency in 2024 → S9 Type Equipment Needs to be Replaced", and "Level 1 Energy Efficiency → Suitable for Industrial Manufacturing Customers". S323 (Implicit Demand Extraction): Based on the correlation analysis of Customer B's data—Customer B uses S9 type level 2 energy efficiency equipment (electrical equipment node) + 2024 phase-out policy for level 2 energy efficiency equipment (market factor node) + industrial manufacturing type (customer node), the implicit demand is derived: "5 S13 type 10kV level 1 energy efficiency transformers need to be replaced in Q4 of 2023 to avoid policy risks in 2024", and the implicit demand forecast results are output.

[0051] Step 400, Output of demand forecast results and linkage with sales decisions: The demand forecast results are transformed into sales decision support information and linked with inventory management, precision marketing and order management. The sales decision support information includes quantitative forecast reports, inventory preparation suggestions, marketing plans and order priority ranking results.

[0052] Among them, "demand forecasting results" refers to the comprehensive demand results output by step S300, including both explicit and implicit customer demands; "sales decision support information" refers to structured information generated based on demand forecasting results to support sales business decisions, specifically including quantitative forecast reports, inventory preparation suggestions, marketing plans, and order priority ranking results; "inventory management" refers to the enterprise's management of electrical equipment inventory, including planning, replenishment, and allocation, with the core objective of achieving "inventory-demand matching"; "precision marketing" refers to targeted marketing activities based on customer demand characteristics, with the core objective of improving customer conversion rates; and "order management" refers to the enterprise's receipt, sorting, and execution of customer equipment orders. In the management phase, the core objective is to optimize order response efficiency. The "Quantitative Forecast Report" refers to a report that presents demand forecast results in a data-driven and structured format, including core information such as customer demand priority, equipment demand type proportions, and demand time distribution. "Stocking Recommendations" refer to inventory replenishment suggestions generated based on demand forecast results and existing inventory, including the type of equipment requiring replenishment, replenishment quantity, and replenishment time. "Marketing Plans" refer to targeted promotional plans developed based on customers' implicit needs, including equipment replacement plans, policy interpretations, and pricing strategies. "Order Priority Ranking Results" refers to the results of ranking orders according to the urgency of customer needs, used to guide the order of order execution.

[0053] In one possible implementation, sales decision support information can be customized: for enterprise sales management departments, quantitative forecast reports can include a "regional demand distribution" module (e.g., the demand from new energy customers in East China accounts for 40%) for regional resource allocation; for frontline sales staff, marketing plans can include a "customer communication script library" (e.g., explaining the subsidy application process for S13 transformers to customers); and order priority ranking can introduce "customer value weight," increasing the urgency weight of long-term cooperative customers (cooperation for more than 5 years) by 0.3, prioritizing the delivery of orders from high-value customers, and further improving customer satisfaction and repurchase rate.

[0054] Specifically, step 400 also includes: Step 410, Quantitative forecast report generation: The demand forecast results are organized into a quantitative forecast report that includes customer demand priority, equipment demand type proportion and demand time distribution. Step 420, Inventory Linkage Processing: Synchronize the quantitative forecast report to the inventory management system, and generate the replenishment suggestion based on the existing electrical equipment inventory quantity. The replenishment suggestion includes the type of equipment to be replenished and the replenishment quantity. Step 430, Precision Marketing Plan Generation: Based on the implicit demand prediction results, generate the marketing plan that matches the potential needs of customers. The marketing plan includes equipment replacement solutions and policy interpretation content. Step 440, Order Priority Ranking: Sort customer orders according to the urgency of demand in the demand forecast results to generate the order priority ranking results. The urgency of demand is determined based on the customer's project progress and the service life of the equipment.

[0055] Among them, "Quantitative Forecast Report" refers to a structured report that presents demand forecast results with specific data; "Customer Demand Priority" refers to the importance level of customers ranked based on the scale and urgency of their demand (e.g., customers with large demand and high urgency have higher priority); "Equipment Demand Type Proportion" refers to the proportion of demand for different equipment models in the total demand (e.g., 10kV transformers account for 65%); and "Demand Time Distribution" refers to the distribution of demand in different time periods (e.g., 13 units demanded in Q3, 5 units demanded in Q4). These three together constitute the "data foundation" for sales decisions. "Inventory Management System" refers to a system used to manage electrical equipment inventory levels and generate replenishment plans; "Existing Electrical Equipment Inventory Quantity" refers to the actual inventory of each equipment model in the current warehouse; "Equipment Types Requiring Replenishment" refers to equipment models whose demand forecast exceeds the existing inventory; and "Replenishment Quantity" refers to the demand... The difference between the forecast quantity and the existing inventory provides direct guidance for the inventory department's purchasing actions; "Marketing plans adapted to potential customer needs" refers to promotional plans designed for implicit needs; "Equipment replacement plans" refers to recommended equipment upgrade plans to customers (such as the specific model and quantity of replacing S9 with S13); "Policy interpretation content" refers to explaining the impact of policies on their equipment selection to customers (such as "S9 will be phased out in 2024, replacing now can avoid policy risks"); the combination of these two aspects enhances customers' acceptance of implicit needs; "Demand urgency" refers to the time urgency of customer needs, determined by "customer project progress" (such as project completion time) and "equipment service life" (such as whether the equipment is nearing its replacement cycle); "Order priority ranking results" refers to the order of order execution ranked by demand urgency, used to guide the production and delivery departments to prioritize high-urgency orders. The above steps involve "industry-specific customization of the linkage process"—for example, given the "high value and long delivery cycle" characteristics of electrical equipment, inventory recommendations are generated one month in advance (to ensure sufficient time for procurement), marketing plans include a dual-dimensional interpretation of "policy + cost" (to persuade customers to make long-term decisions), and order sorting incorporates the "policy validity period" factor (to avoid policy risks), ensuring that the linkage effect aligns with the actual needs of the industry.

[0056] In one possible implementation, the quantitative forecasting report could include a "Regional Demand Analysis" module (e.g., East China accounts for 40% of demand, North China 30%) to guide the allocation of regional sales resources; inventory linkage could include "Replenishment Time Window" suggestions (e.g., 10kV transformers need to be replenished before August 2023 to ensure delivery in September) to avoid replenishment delays; the precision marketing plan could include a "Cost Comparison Table" (e.g., the initial investment for customer B to replace S13 equipment vs. the penalty cost after policy elimination) to strengthen customers' willingness to replace; and order priority ranking could include a weight for "Years of Customer Cooperation" (customers with more than 5 years of cooperation have a 20% higher priority) to maintain customer relationships.

[0057] For example, in one feasible implementation, linking prediction results with decision-making includes: S410 (Quantitative Forecast Report Generation): Organizes demand forecast results into a quantitative report, including customer demand priority (Customer A: 13 units demanded + Q4 completion → Priority 1; Customer B: 5 units demanded + policy elimination → Priority 2), equipment demand type percentage (10kV-1000kVA transformers account for 65%, S13 type 10kV transformers account for 35%), and demand time distribution (Q3 2023: Customer A 13 units, Q4: Customer B 5 units). S420 (Inventory Linkage Processing): Synchronizes the report to the inventory management system, and generates a restocking recommendation based on the existing inventory (8 units of 10kV-1000kVA transformers and 12 units of S13 type 10kV transformers) (5 units of 10kV-1000kVA transformers need to be restocked, to be completed before August 2023; no restocking is required for S13 type transformers). S430 (Precision Marketing Plan Generation): Targeting the implicit needs of Customer B, a marketing plan is generated (equipment replacement plan: 5 S13 type 10kV-800kVA transformers; policy interpretation content: "Level 2 energy efficiency will be phased out in 2024, and replacement now can enjoy manufacturer installation subsidies, avoiding subsequent rectification costs"). S440 (Order Priority Ranking): Based on the urgency of demand (Customer A's project is completed in September, urgency 0.9; Customer B's policy will take effect in 2024, urgency 0.7), generate order ranking results (Customer A's 13-unit order is prioritized for production, with stocking in August and delivery in September; Customer B's 5-unit order is produced in October and delivered in November), thus realizing the commercialization of the forecast results.

[0058] In this embodiment, the present invention effectively solves the core problems of low demand forecasting accuracy, data silos, model disconnect from industry scenarios, and forecasting results disconnect from business decisions in the electrical equipment sales field by integrating multi-source data, constructing industry-adaptive AI models, and linking prediction and decision-making. Multi-source data collection covers all dimensions of customers, products, and markets. The preprocessing stage extracts industry-specific correlation features. The integrated AI model accurately mines explicit and implicit needs and can dynamically adapt to policy and market changes. The forecasting results are further linked to inventory management, precision marketing, and order management, which not only improves the comprehensiveness and dynamic adaptability of demand forecasting, but also reduces the risk of inventory and demand mismatch and improves sales conversion efficiency. At the same time, it can be extended to adapt to other industrial product sales forecasting scenarios, combining industry-specificity and general adaptability, and providing accurate decision-making support for the entire process of electrical equipment sales business.

[0059] Based on the above embodiments, such as Figure 2As shown, the present invention also provides an artificial intelligence-based customer demand forecasting system for electrical equipment sales, used to support the artificial intelligence-based customer demand forecasting method for electrical equipment sales described in the above embodiments. The artificial intelligence-based customer demand forecasting system for electrical equipment sales includes: Perception Layer: Connects to customer-side data sources, product-side data sources, and market-side data sources through preset data interfaces to collect multi-source data related to the sales of electrical equipment. The multi-source data includes customer-side data, product-side data, and market-side data. Data layer: The multi-source data is standardized to obtain feature data that is adapted to the artificial intelligence prediction model. The standardization process includes data cleaning, feature engineering and data normalization. AI computing layer: Based on the feature data, a fusion artificial intelligence prediction model is constructed, and the demand prediction results of electrical equipment sales customers are output through the fusion artificial intelligence prediction model. The demand prediction results include explicit demand prediction results and implicit demand prediction results. The construction and prediction process of the fusion-based artificial intelligence prediction model includes: The historical procurement time series data in the feature data is processed to output explicit demand forecast results; an electrical equipment sales knowledge graph is constructed and implicit customer needs are mined based on the electrical equipment sales knowledge graph to output implicit demand forecast results; the explicit demand forecast results and implicit demand forecast results are dynamically adjusted with inventory cost reduction rate and policy change adaptability as optimization objectives. Application layer: The demand forecast results are transformed into sales decision support information and linked with inventory management, precision marketing and order management. The sales decision support information includes quantitative forecast reports, inventory preparation suggestions, marketing plans and order priority ranking results.

[0060] In an optional embodiment, the perception layer is further configured to: collect customer-side data from a customer relationship management system; collect product-side data from an enterprise resource planning system; and collect market-side data from an industry policy platform and an industry database. The customer-side data reflects customer attributes and behavioral information related to electrical equipment procurement, the product-side data reflects electrical equipment product characteristics and inventory status information, and the market-side data reflects electrical equipment industry policies and market trends.

[0061] In an optional embodiment, the data layer is further configured to: data cleaning: remove outliers from the multi-source data using an outlier removal algorithm based on statistical rules, and fill missing values ​​in the multi-source data using a missing value filling algorithm based on industry averages; feature engineering: extract customer features, product features, and market features from the cleaned multi-source data, wherein the customer features include correlation features between industry type and equipment service life, and correlation features between project scale and purchase frequency; the product features include correlation features between energy efficiency level and applicable scenarios, and correlation features between inventory turnover rate and replenishment cycle; and the market features include correlation features between policy validity period and subsidy ratio, and correlation features between market growth rate and competitor inventory; and data normalization: map the extracted customer features, product features, and market features to a preset numerical range using a preset normalization algorithm to obtain the feature data.

[0062] In an optional embodiment, the AI ​​computing layer is further used for: time series model construction: constructing a time series prediction model using a deep learning model, wherein the input of the time series prediction model is historical procurement time series data in the feature data, and the output is the procurement quantity and procurement time of electrical equipment within a future preset period; model training and validation: using 70% of the historical procurement time series data as the training set and 30% as the validation set, training the time series prediction model using a preset optimizer, using the mean square error between the predicted value and the actual value as the loss function, iteratively training until the loss function value is less than a preset threshold; explicit demand output: inputting the feature data into the trained time series prediction model to obtain the explicit demand prediction result.

[0063] In an optional embodiment, the AI ​​computing layer is further used for: knowledge graph node construction: constructing nodes of the electrical equipment sales knowledge graph, the nodes including customer nodes, electrical equipment nodes, and market factor nodes, the customer nodes being associated with basic customer information, the electrical equipment nodes being associated with equipment parameter information, and the market factor nodes being associated with industry policy information; node relationship definition: defining the relationships between the customer nodes, electrical equipment nodes, and market factor nodes, the relationships including the matching relationship between customer industry type and electrical equipment applicable scenarios, and the triggering relationship between industry policies and electrical equipment demand; implicit demand extraction: extracting unexpressed equipment needs from the feature data based on the relationships, forming the implicit demand prediction results.

[0064] In an optional embodiment, the application layer is further configured to: generate a quantitative forecast report: organize the demand forecast results into a quantitative forecast report containing customer demand priority, equipment demand type proportion, and demand time distribution; synchronize the quantitative forecast report to the inventory management system, and generate the restocking suggestion based on the existing electrical equipment inventory quantity, the restocking suggestion including the type of equipment to be replenished and the replenishment quantity; generate a precise marketing plan: generate a marketing plan adapted to the potential needs of customers based on the implicit demand forecast results, the marketing plan including equipment replacement plan and policy interpretation content; prioritize orders: sort customer orders according to the demand urgency in the demand forecast results, and generate the order priority ranking result, the demand urgency being determined based on the customer project progress and equipment service life.

[0065] In this embodiment, the present invention effectively solves the core problems of low demand forecasting accuracy, data silos, model disconnect from industry scenarios, and forecasting results disconnect from business decisions in the electrical equipment sales field by integrating multi-source data, constructing industry-adaptive AI models, and linking prediction and decision-making. Multi-source data collection covers all dimensions of customers, products, and markets. The preprocessing stage extracts industry-specific correlation features. The integrated AI model accurately mines explicit and implicit needs and can dynamically adapt to policy and market changes. The forecasting results are further linked to inventory management, precision marketing, and order management, which not only improves the comprehensiveness and dynamic adaptability of demand forecasting, but also reduces the risk of inventory and demand mismatch and improves sales conversion efficiency. At the same time, it can be extended to adapt to other industrial product sales forecasting scenarios, combining industry-specificity and general adaptability, and providing accurate decision-making support for the entire process of electrical equipment sales business.

[0066] Furthermore, the AI-based electrical equipment sales customer demand forecasting system can run the aforementioned AI-based electrical equipment sales customer demand forecasting method. For specific implementation details, please refer to the method embodiment, which will not be repeated here.

[0067] Based on the above embodiments, such as Figure 3 As shown, the present invention also provides an electronic device, the electronic device comprising: The processor 22 includes at least one processor 22, at least one memory 21, a communication interface 23, and a communication bus 24, wherein the processor 22 is communicatively connected to the memory 21. In this embodiment, the memory 21 can be implemented in any suitable manner, for example, the memory 21 can be a read-only memory, a hard disk drive, a solid-state drive, or a USB flash drive, etc.; the memory 21 is used to store at least one executable instruction executed by the processor; In this embodiment, the processor 22 can be implemented in any suitable manner. For example, the processor 22 can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers, etc.; the processor is used to execute the executable instructions to implement the artificial intelligence-based customer demand forecasting method for electrical equipment sales as described above.

[0068] Based on the above embodiments, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described artificial intelligence-based method for predicting customer demand for electrical equipment sales.

[0069] Those skilled in the art will recognize that the modules and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0070] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, equipment, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0071] In the several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or units may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or equipment, and may be electrical, mechanical, or other forms.

[0072] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0073] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0074] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program instructions, such as USB flash drives, portable hard drives, read-only storage servers, random access storage servers, magnetic disks, or optical disks.

[0075] Furthermore, it should be noted that the combination of the various technical features in this case is not limited to the combination methods described in the claims of this case or the combination methods described in the specific embodiments. All technical features described in this case can be freely combined or combined in any way, unless they contradict each other.

[0076] It should be noted that the above examples are merely specific embodiments of the present invention, and the present invention is obviously not limited to the above embodiments, with many similar variations. All modifications that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should fall within the protection scope of this invention.

[0077] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting customer demand for electrical equipment sales based on artificial intelligence, characterized in that, Includes the following steps: Multi-source data acquisition: By connecting to customer-side data sources, product-side data sources, and market-side data sources through preset data interfaces, multi-source data related to the sales of electrical equipment is collected. The multi-source data includes customer-side data, product-side data, and market-side data. Data preprocessing: The multi-source data is standardized to obtain feature data that is adapted to the artificial intelligence prediction model. The standardization process includes data cleaning, feature engineering and data normalization. Artificial intelligence prediction model construction and prediction: Based on the feature data, a fusion artificial intelligence prediction model is constructed, and the demand prediction results of electrical equipment sales customers are output through the fusion artificial intelligence prediction model. The demand prediction results include explicit demand prediction results and implicit demand prediction results. The construction and prediction process of the fusion-based artificial intelligence prediction model includes: The historical procurement time series data in the feature data is processed to output explicit demand forecast results; an electrical equipment sales knowledge graph is constructed and implicit customer needs are mined based on the electrical equipment sales knowledge graph to output implicit demand forecast results; the explicit demand forecast results and implicit demand forecast results are dynamically adjusted with inventory cost reduction rate and policy change adaptability as optimization objectives. Linking demand forecast results with sales decisions: The demand forecast results are transformed into sales decision support information and linked with inventory management, precision marketing, and order management. The sales decision support information includes quantitative forecast reports, inventory preparation suggestions, marketing plans, and order priority ranking results.

2. The method for predicting customer demand for electrical equipment sales based on artificial intelligence according to claim 1, characterized in that, The multi-source data acquisition steps include: Customer-side data collection: Collect the customer-side data from the customer relationship management system; Product-side data collection: Collect the product-side data from the enterprise resource planning system; Market-side data collection: The aforementioned market-side data is collected from industry policy platforms and industry databases; The customer-side data reflects customer attributes and behaviors related to electrical equipment procurement; the product-side data reflects electrical equipment product characteristics and inventory status; and the market-side data reflects electrical equipment industry policies and market trends.

3. The method for predicting customer demand for electrical equipment sales based on artificial intelligence according to claim 2, characterized in that, The customer-side data collection steps include: Customer basic information collection: Collect customer industry type and project scale data. The industry type includes industrial manufacturing, new energy infrastructure and commercial building. The project scale includes annual output value and infrastructure investment amount. Customer historical purchase data collection: Collect data on the type of electrical equipment purchased by customers in the past, the purchase quantity, the purchase time, and the service life of the purchased equipment; Customer feedback data collection: Collect customer records of faults in purchased electrical equipment and data on their intention to replace the equipment.

4. The method for predicting customer demand for electrical equipment sales based on artificial intelligence according to claim 1, characterized in that, The data preprocessing steps include: Data cleaning: Outlier removal algorithms based on statistical rules are used to remove outliers from the multi-source data, and missing value filling algorithms based on industry averages are used to fill missing values ​​in the multi-source data; Feature engineering: Extracting customer features, product features, and market features from cleaned multi-source data. The customer features include the correlation between industry type and equipment service life, and the correlation between project scale and purchase frequency. The product features include the correlation between energy efficiency level and applicable scenarios, and the correlation between inventory turnover rate and replenishment cycle. The market features include the correlation between policy validity period and subsidy ratio, and the correlation between market growth rate and competitor inventory. Data normalization: The extracted customer features, product features and market features are mapped to a preset numerical range using a preset normalization algorithm to obtain the feature data.

5. The method for predicting customer demand for electrical equipment sales based on artificial intelligence according to claim 1, characterized in that, The process of processing the historical procurement time series data in the feature data to output explicit demand forecast results includes: Time series model construction: A time series prediction model is constructed using a deep learning model. The input of the time series prediction model is the historical procurement time series data in the feature data, and the output is the procurement quantity and procurement time of electrical equipment within a future preset period. Model training and validation: Using 70% of the historical procurement time series data as the training set and 30% as the validation set, the time series prediction model is trained using a preset optimizer. The mean square error between the predicted value and the actual value is used as the loss function, and the training is iterated until the loss function value is less than a preset threshold. Explicit demand output: Input the feature data into the trained time series prediction model to obtain the explicit demand prediction result.

6. The method for predicting customer demand for electrical equipment sales based on artificial intelligence according to claim 1, characterized in that, The process of constructing a knowledge graph for electrical equipment sales and mining latent customer needs based on this knowledge graph to output latent need prediction results includes: Knowledge graph node construction: Construct the nodes of the electrical equipment sales knowledge graph. The nodes include customer nodes, electrical equipment nodes, and market factor nodes. The customer nodes are associated with basic customer information, the electrical equipment nodes are associated with equipment parameter information, and the market factor nodes are associated with industry policy information. Node Relationship Definition: Define the relationship between the customer node, electrical equipment node, and market factor node. The relationship includes the matching relationship between customer industry type and electrical equipment application scenario, and the triggering relationship between industry policy and electrical equipment demand. Implicit demand extraction: Based on the aforementioned relationships, extract the equipment requirements that customers have not explicitly expressed from the feature data to form the implicit demand prediction results.

7. The method for predicting customer demand for electrical equipment sales based on artificial intelligence according to claim 1, characterized in that, The steps for linking demand forecasting results output with sales decisions include: Quantitative forecast report generation: The demand forecast results are compiled into a quantitative forecast report that includes customer demand priority, equipment demand type proportion, and demand time distribution. Inventory linkage processing: The quantitative forecast report is synchronized to the inventory management system, and the replenishment suggestion is generated by combining the existing electrical equipment inventory quantity. The replenishment suggestion includes the type of equipment to be replenished and the replenishment quantity. Precision marketing plan generation: Based on the implicit demand prediction results, a marketing plan that adapts to the potential needs of customers is generated. The marketing plan includes equipment replacement plan and policy interpretation content. Order Priority Ranking: Customer orders are ranked according to the urgency of demand in the demand forecast results to generate the order priority ranking results. The urgency of demand is determined based on the customer's project progress and the service life of the equipment.

8. A customer demand forecasting system for electrical equipment sales based on artificial intelligence, characterized in that, Includes the following steps: Perception Layer: Connects to customer-side data sources, product-side data sources, and market-side data sources through preset data interfaces to collect multi-source data related to the sales of electrical equipment. The multi-source data includes customer-side data, product-side data, and market-side data. Data layer: The multi-source data is standardized to obtain feature data that is adapted to the artificial intelligence prediction model. The standardization process includes data cleaning, feature engineering and data normalization. AI Computing Layer: Based on the aforementioned feature data, a fusion-based artificial intelligence prediction model is constructed. The fusion-based artificial intelligence prediction model outputs demand prediction results for electrical equipment sales customers. The demand prediction results include explicit demand prediction results and implicit demand prediction results. The construction and prediction process of the fusion-based artificial intelligence prediction model includes: The historical procurement time series data in the feature data is processed to output explicit demand forecast results; an electrical equipment sales knowledge graph is constructed and implicit customer needs are mined based on the electrical equipment sales knowledge graph to output implicit demand forecast results; the explicit demand forecast results and implicit demand forecast results are dynamically adjusted with inventory cost reduction rate and policy change adaptability as optimization objectives. Application layer: The demand forecast results are transformed into sales decision support information and linked with inventory management, precision marketing and order management. The sales decision support information includes quantitative forecast reports, inventory preparation suggestions, marketing plans and order priority ranking results.

9. An electronic device, characterized in that, The electronic device includes: The processor and the memory are communicatively connected. The memory is used to store at least one executable instruction executed by the processor, the processor being used to execute the executable instruction to implement the artificial intelligence-based customer demand forecasting method for electrical equipment sales as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the artificial intelligence-based customer demand forecasting method for electrical equipment sales as described in any one of claims 1 to 7.

Images