Intelligent mining and precision marketing method for small and medium-sized enterprise demand in electronic information

Abstract

This application relates to the interdisciplinary fields of digital marketing and electronic information technology, and specifically to a method for intelligent demand mining and precision marketing for small and medium-sized enterprises (SMEs) in the electronic information industry. The method includes: generating anonymized simulated customer data for SMEs in the electronic information industry, with the customer data generated based on customer characteristics in the electronic information industry, including customer attribute information and demand expression information; constructing a mapping model between customer attributes and product characteristics based on the customer data; analyzing the customer data based on the mapping model to mine potential customer needs and quantify the intensity of those needs; and generating marketing solutions adapted to the electronic information industry scenario based on these potential needs. This solves the problems of difficulty in demand mining, extensive marketing methods, and lack of digital marketing capabilities among SMEs in the electronic information industry. While ensuring data compliance, it lowers the technical application threshold and achieves accurate demand identification, automatic generation of marketing solutions, and continuous model evolution.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of market digitalization and electronic information cross technology, in particular to a method for intelligent mining of small and medium-sized enterprise demand in electronic information and precision marketing. BACKGROUND

[0002] With the rapid development of the Internet of Things, industrial control, and intelligent manufacturing, the electronic information industry has become an important force driving the development of the digital economy. Among them, small and medium-sized enterprises as the main body of the electronic information industry, the number is huge, the market participation is high, but in the marketing link, there are many problems such as difficulty in identifying customer demand, low precision of marketing, and lack of data support.

[0003] Currently, the customer demand of small and medium-sized enterprises in electronic information has the characteristics of high fragmentation and diversification. Customers are widely distributed in security and protection, industrial control, consumer electronics, Internet of Things and other sub-sectors, and their purchasing preferences, product parameter requirements, and budget ranges differ significantly. However, most small and medium-sized enterprises still rely on manual experience and traditional marketing methods, making it difficult to achieve large-scale identification and precise matching of customer demand, resulting in low customer conversion rate, high customer acquisition cost, and restricting the development of enterprises.

[0004] In terms of data, small and medium-sized enterprises generally lack professional data collection and analysis capabilities, and customer information is scattered in unstructured data such as inquiry records, communication texts, and contract fragments, making it difficult to form structured data sets for analysis. At the same time, due to the protection of business secrets and data security considerations, enterprises are often reluctant to open their core business data to external platforms, further increasing the technical difficulty of demand mining. Although some large enterprises have established a complete customer relationship management and data mining system, the high construction cost and complex operation process make it difficult to adapt to the actual ability and resources of small and medium-sized enterprises.

[0005] There are many methods for identifying customer demand and precision marketing in related technologies. For example, patent CN119107160A discloses an online e-commerce recommendation method based on big data, which generates a recommendation result according to the preferred goods in the user model through user collaborative filtering and item collaborative filtering. This method can achieve personalized recommendation, but it relies on real user behavior data to build a recommendation model and does not consider the adaptability of technical parameters and industry scenarios, which is difficult to apply directly to small and medium-sized enterprises that lack user behavior accumulation and scattered data.

[0006] Another patent CN112070519B discloses a prediction method based on data global search and feature classification, which extracts multi-dimensional features of brand crowd and non-brand crowd, and establishes a potential customer identification model based on an improved random forest algorithm. This method improves the accuracy of customer identification, but still needs a large amount of real user historical data for model training, and does not specifically model the technical parameter characteristics of electronic information products, making it difficult to meet the business needs of small and medium-sized enterprises in the electronic information industry for precise matching of product parameters and customer demand.

[0007] In summary, there is currently a lack of a demand mining and precision marketing method that does not require enterprises to provide core data and participate in complex system construction, and that can adapt to the characteristics of the electronic information industry. It cannot meet the actual marketing needs of small and medium-sized enterprises under limited resources. Therefore, it is urgent to design a low-threshold, high-adaptation, and independently-operable electronic information small and medium-sized enterprise demand intelligent mining and precision marketing method to fill the gap in industry adaptability and feasibility of related technologies. SUMMARY

[0008] The present application provides an electronic information small and medium-sized enterprise demand intelligent mining and precision marketing method to solve the technical problems of electronic information small and medium-sized enterprise customer demand mining difficulty, extensive marketing method, and lack of digital marketing capability. The method of the present application generates desensitized simulated customer data of electronic information small and medium-sized enterprises, constructs a four-dimensional mapping model of product parameters-customer industry-purchasing behavior-demand characteristics, mines customer demand and quantifies the intensity, generates a precision marketing scheme, and optimizes based on feedback data. The closed loop realizes the precise identification of customer demand, the automatic generation of marketing scheme, and the iterative optimization of marketing effect. The whole process does not require enterprises to provide core data, and the landing threshold is low.

[0009] The first aspect of the present application provides an electronic information small and medium-sized enterprise demand intelligent mining and precision marketing method, comprising the following steps:

[0010] Desensitized simulated customer data of electronic information small and medium-sized enterprises is generated, which is generated based on customer characteristics of the electronic information industry and does not contain real enterprise business secrets, including customer industry, purchasing category, purchasing frequency, inquiry keywords, and demand description;

[0011] Based on the desensitized simulated customer data, a four-dimensional mapping model of product parameters-customer industry-purchasing behavior-demand characteristics is constructed;

[0012] Based on the four-dimensional mapping model, customer data is analyzed to mine potential customer demand, and a demand activation index is generated to quantify the demand intensity, and a potential demand list is output;

[0013] Based on the potential demand list, automatically generate a precise marketing scheme adapted to the electronic information industry scene;

[0014] Based on the marketing execution feedback data, update the weights of the four-dimensional mapping model, complete the iteration optimization of demand mining and marketing scheme.

[0015] Optionally, in some embodiments, the generated number of desensitized simulated customer data is not less than 1000, and the industry to which the customer belongs includes intelligent manufacturing, security, Internet of Things, industrial control and consumer electronics.

[0016] Optionally, in some embodiments, the four-dimensional mapping model includes product parameter dimension, customer industry dimension, procurement behavior dimension and demand feature dimension; wherein,

[0017] The product parameter dimension includes the function, interface, power consumption, size, stability and price range of the electronic information product;

[0018] The customer industry dimension includes the classification of the subfield of the electronic information industry;

[0019] The procurement behavior dimension includes procurement batch, procurement cycle, decision chain length and budget interval;

[0020] The demand feature dimension includes replacement demand, upgrade demand, matching demand and new project demand.

[0021] Optionally, in some embodiments, the four-dimensional mapping model adopts an improved weighted collaborative filtering algorithm combined with a lightweight random forest classifier algorithm, which specifically includes:

[0022] In the preprocessing layer, the desensitized simulated customer data is standardized, unstructured inquiry keywords and demand descriptions are converted into structured feature vectors, discrete data is one-hot encoded, and continuous data is normalized;

[0023] In the association layer, an improved weighted collaborative filtering algorithm is used to calculate the association weight between the four dimensions, the improved weighted collaborative filtering algorithm expands the calculation range to the cross association between the product parameter dimension, the customer industry dimension, the procurement behavior dimension and the demand feature dimension, and sets a weighting coefficient for different dimension combinations according to the characteristics of the electronic information industry, the weight calculation formula is:

[0024] ;

[0025] Wherein, is the association weight between dimension i and dimension j, is the frequency of the simultaneous appearance of dimension i and dimension j, and These represent the frequencies of occurrence of dimension i and dimension j individually, respectively.

[0026] In the classification layer, a lightweight random forest classifier is introduced, which takes customer industry and purchasing behavior as input features and product parameters and demand features as output labels to complete the four-dimensional mapping classification.

[0027] Optionally, in some embodiments, the number of decision trees in the lightweight random forest classifier is 30-100, and the maximum depth is 5-10 layers;

[0028] When constructing the model, the anonymized simulated customer data is divided into training set, validation set and test set according to the proportion. Gradient descent method is used for iterative training. Training stops when the accuracy of the validation set reaches a preset threshold.

[0029] Precision, recall, and F1 score are used as model evaluation metrics, and the test set evaluation metrics are required to reach the preset thresholds.

[0030] Optionally, in some embodiments, the demand activation index is quantified and scored based on customer inquiry behavior, purchasing records and industry driving factors, and demand priority levels are divided according to the scoring results.

[0031] Optionally, in some embodiments, the precision marketing scheme includes product recommendation information and marketing execution strategy information, wherein the marketing execution strategy information is generated based on customer attributes and demand characteristics.

[0032] Optionally, in some embodiments, the iterative optimization further includes adjusting the mapping relationship of the four-dimensional mapping model based on the marketing execution feedback data.

[0033] Optionally, in some embodiments, the demand activation index is calculated by weighted summation based on inquiry behavior factors, procurement record factors and industry driving factors, with a score range of 0-100 points, and demand priority levels are divided according to the score.

[0034] A second aspect of this application provides a device for intelligent demand mining and precision marketing for small and medium-sized enterprises in the electronics and information industry, comprising:

[0035] The data generation module is used to generate anonymized simulated customer data for small and medium-sized enterprises in the electronics and information industry. The anonymized simulated customer data is generated based on the characteristics of customers in the electronics and information industry and does not contain any real business secrets of the enterprises. It includes: the customer's industry, purchase category, purchase frequency, inquiry keywords and demand description.

[0036] The modeling module is used to construct a four-dimensional mapping model of product parameters, customer industry, purchasing behavior, and demand characteristics based on the de-identified simulated customer data.

[0037] The intelligent mining module is used to analyze customer data based on the four-dimensional mapping model, mine potential customer needs, generate a demand activation index to quantify demand intensity, and output a list of potential needs.

[0038] The solution generation module is used to automatically generate precise marketing solutions adapted to the electronic information industry scenario based on the potential demand list.

[0039] The optimization module is used to update the weights of the four-dimensional mapping model based on marketing execution feedback data, and to perform iterative optimization of demand mining and marketing plans.

[0040] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement a method for intelligent mining and precise marketing of the needs of small and medium-sized enterprises in the field of electronic information, as described in the above embodiments.

[0041] The fourth aspect of this application provides a computer program product having a computer program stored thereon, which is executed by a processor to implement a method for intelligent mining and precise marketing of the needs of small and medium-sized electronic information enterprises as described in the above embodiments.

[0042] The beneficial effects of the embodiments of this application are as follows:

[0043] (1) The implementation of the method in this application does not require enterprises to provide any core data or participate in the implementation process. By generating a model of completely desensitized simulated customer data, it solves the problem that small and medium-sized enterprises in the electronics and information industry cannot apply digital marketing technology due to data dispersion and trade secret protection restrictions. It lowers the threshold for technology application and has strong practicality and scalability.

[0044] (2) This application constructs a four-dimensional mapping model of product parameters, customer industry, purchasing behavior and demand characteristics, which deeply binds the technical parameters of electronic information products with customer demand, breaks through the limitation of the lack of industry-specificity of related general marketing models, and improves the accuracy of demand mining; at the same time, it uses the demand activation index to quantify the demand intensity, realizes the scientific division of customer priority, and effectively solves the problems of fuzzy demand identification and mismatch of marketing resources in traditional methods.

[0045] (3) This application automatically generates a precise marketing plan based on a potential demand list, which includes product recommendations, selling points, outreach strategies and business suggestions. This achieves automated connection from demand identification to marketing execution and improves marketing response efficiency. At the same time, through a closed-loop optimization mechanism, the model weights are dynamically updated based on marketing execution feedback data, so that the model has adaptive evolution capabilities and continuously improves prediction accuracy and plan adaptability.

[0046] (4) The method steps of this application are clear and reproducible. It is based entirely on simulated data and does not rely on any real enterprise's business system or trade secrets. The ownership is clear and unambiguous. It can be used directly for the transformation of scientific research results and project application as an independent research result of universities. It is in line with the orientation of scientific and technological achievement transformation and has good social benefits and promotion and application value.

[0047] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0048] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0049] Figure 1 This is a flowchart illustrating a method for intelligent demand mining and precision marketing for small and medium-sized enterprises in the electronics and information industry, provided according to an embodiment of this application.

[0050] Figure 2 This is a block diagram of an intelligent demand mining and precision marketing device for small and medium-sized enterprises in the field of electronic information, provided according to an embodiment of this application.

[0051] Figure 3 This is a block diagram of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0052] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0053] The following description, with reference to the accompanying drawings, illustrates an embodiment of a method for intelligent demand mining and precision marketing for small and medium-sized enterprises (SMEs) in the electronic information industry. Addressing the problems mentioned in the background art, such as difficulties in demand mining, extensive marketing methods, and a lack of digital marketing capabilities among SMEs in the electronic information industry, this application first provides a method for intelligent demand mining and precision marketing for SMEs in the electronic information industry. Specifically… Figure 1 This is a flowchart illustrating a method for intelligent demand mining and precision marketing for small and medium-sized enterprises in the electronics and information industry, as provided in an embodiment of this application. Figure 1 As shown, this method for intelligent demand mining and precision marketing for small and medium-sized enterprises in the electronics and information industry includes the following steps:

[0054] In step S101, de-identified simulated customer data for small and medium-sized enterprises in the electronics and information industry is generated. The de-identified simulated customer data is generated based on the characteristics of customers in the electronics and information industry and does not contain any real business secrets of the enterprises. It includes: the industry to which the customer belongs, the category of goods purchased, the frequency of purchase, the keywords for inquiry and the description of demand.

[0055] Step S101 of this application embodiment aims to solve the technical problem in the related technology that the demand mining model cannot be built due to the dispersion of data of small and medium-sized enterprises and the inability to obtain confidential data. By generating completely desensitized simulated customer data, a compliant, sufficient and reproducible training foundation is provided for subsequent model building.

[0056] Specifically, in step S101, this embodiment of the application employs a data simulation generation method based on customer characteristics in the electronics and information industry. The generated anonymized simulated customer data includes the following core fields: customer's industry, product category, purchase frequency, inquiry keywords, and demand description. All data is fictitious and does not contain any real company's name, contact information, transaction prices, or other trade secret information, ensuring data compliance.

[0057] Optionally, in some embodiments, the number of de-identified simulated customer data generated is no less than 1,000, and the customer's industry includes: smart manufacturing, security, Internet of Things, industrial control and consumer electronics.

[0058] To ensure the representativeness of simulated data to real business scenarios, this application constructs a customer profile from the following five dimensions: customer industry, product category, purchase frequency, inquiry keywords, and demand description. The customer industry covers major sub-sectors of the electronics and information industry, including intelligent manufacturing, security, IoT, industrial control, and consumer electronics. Different industries correspond to different product needs and technology preferences; for example, the intelligent manufacturing industry emphasizes the stability and power consumption of industrial controllers, while the security industry focuses on the resolution and night vision capabilities of monitoring modules. The product category refers to the types of electronic information products the customer has historically purchased or inquired about, including but not limited to controllers, sensors, IoT modules, monitoring modules, and embedded motherboards. Each category is associated with a set of predefined product parameter templates for parameter matching in the subsequent four-dimensional mapping model. The purchase frequency describes the customer's purchasing patterns, categorized into three levels: high frequency (more than once per month), medium frequency (once per quarter), and low frequency (once every six months or more). The purchase frequency reflects the customer's business activity and the continuity of demand. Inquiry keywords are unstructured texts that simulate the inquiries made by customers, such as "low-power controller," "industrial-grade stability," and "high-definition night vision," used to extract explicit customer needs. Requirement descriptions are implicit requirements that simulate the communication made by customers, such as "existing controllers have high power consumption; a low-power product is needed as a replacement," used to uncover potential customer needs (such as replacement or upgrade requirements).

[0059] This application employs a data generation method combining rule-based and random sampling. Specifically, based on the actual distribution ratio of SMEs in the electronics and information industry, this application sets the data proportion for each industry, such as 30% for intelligent manufacturing, 25% for security, 20% for IoT, 15% for industrial control, and 10% for consumer electronics, ensuring data diversity. Typical procurement category combinations are pre-set for each industry; for example, the intelligent manufacturing industry is associated with "industrial controller + sensor," and the security industry with "monitoring module + video capture card," randomly selected according to a preset probability during generation. Procurement frequency is generated based on industry attributes; for example, project-based industries (intelligent manufacturing) tend to have low-to-medium frequency procurement, while operation-based industries (security) tend to have high-frequency procurement. Then, inquiry keywords and demand descriptions are generated using a combination of template filling and natural language. For example, for the demand for "low-power controllers," a demand description template is generated: "Needs to replace existing [high-power] products, requiring [low power consumption] and [industrial-grade stability]," where the words within square brackets are randomly selected from a pre-set thesaurus.

[0060] To meet the basic requirements for training data volume for machine learning models, this step generates no fewer than 1,000 anonymized simulated customer data entries. All data generation processes do not involve the collection, storage, or use of real enterprise information, fully complying with the Personal Information Protection Law and trade secret protection requirements, ensuring data compliance for subsequent model training and applications.

[0061] In step S102, a four-dimensional mapping model of product parameters, customer industry, purchasing behavior, and demand characteristics is constructed based on de-identified simulated customer data.

[0062] Step S102 of this application embodiment aims to solve the technical problem that general marketing models in related technologies lack industry specificity and cannot adapt to the parameter characteristics of electronic information products. By constructing an industry-specific four-dimensional mapping model, a quantitative correlation between customer attributes and product needs is established, providing core algorithm support for subsequent demand mining.

[0063] Optionally, in some embodiments, the four-dimensional mapping model includes four core dimensions: product parameter dimension, customer industry dimension, purchasing behavior dimension, and demand characteristic dimension. The product parameter dimension characterizes the technical features of electronic information products, including functions (such as control, acquisition, and transmission), interface types (such as RS485, Ethernet, USB (Universal Serial Bus)), power consumption range (such as low power ≤5W, standard power 5-15W), physical size, stability level (industrial grade, commercial grade), and price range. This dimension is the output target of demand mining, determining the specific product specifications recommended to customers. The customer industry dimension identifies the customer's specific electronic information sub-sector, including smart manufacturing, security, IoT, industrial control, and consumer electronics. Different industries have different preferences for product parameters; for example, the smart manufacturing industry emphasizes industrial-grade stability, while the consumer electronics industry focuses on size and cost. The purchasing behavior dimension is used to characterize customers' purchasing patterns and decision-making characteristics, including purchase volume (small, medium, large), purchasing cycle (monthly, quarterly, annual), decision chain length (single-person decision-making, multi-person decision-making, committee decision-making), and budget range. This dimension reflects customers' purchasing capabilities and decision-making patterns, influencing the formulation of marketing strategies. The demand characteristics dimension is used to classify customers' demand types, including replacement needs (upgrading existing products), upgrade needs (performance or functional improvements), complementary needs (system expansion or parts procurement), and new project addition needs (completely new procurement). This dimension determines the strategic direction of product recommendations in marketing plans.

[0064] The four dimensions mentioned above together constitute the mapping relationship from customer attributes to product needs. Customer industry and purchasing behavior serve as input features, while product parameters and demand features serve as output labels, forming a trainable learning framework.

[0065] Optionally, in some embodiments, to achieve accurate mapping between the four dimensions, the four-dimensional mapping model employs a combination algorithm of an improved weighted collaborative filtering algorithm and a lightweight random forest classifier, including the following three levels.

[0066] The preprocessing layer is responsible for transforming the original anonymized simulated customer data into structured feature vectors that the model can process, transforming unstructured inquiry keywords and demand descriptions into structured feature vectors, performing one-hot encoding on discrete data, and normalizing continuous data.

[0067] Specifically, in the preprocessing layer, unstructured query keywords and demand descriptions are transformed into structured feature vectors. This embodiment employs an industry-dictionary-based matching method, for example, decomposing "low-power controller" into key-value pairs of [power consumption: low, product type: controller] to form a sparse feature vector. Discrete categorical variables such as customer industry and purchase category are one-hot encoded. Taking industry as an example, categories such as intelligent manufacturing, security, and IoT are mapped to independent binary feature bits to avoid the model misinterpreting the ordinal relationships between categories. Furthermore, continuous values ​​such as purchase frequency and budget range are subjected to Min-Max normalization, linearly mapping them to the [0,1] interval. The normalization formula is:

[0068] ;

[0069] in, and These are the minimum and maximum values ​​of the feature in the training set, respectively.

[0070] At the association layer, this embodiment employs an improved weighted collaborative filtering algorithm to calculate the association weights among the four dimensions, quantifying the contribution of different dimension combinations to demand forecasting. The core logic of the algorithm is to calculate the association strength between dimensions based on co-occurrence frequency, and the weight calculation formula is as follows:

[0071] ;

[0072] in, Let i be the association weight between dimension i and dimension j. Let i be the frequency of the simultaneous occurrence of dimension i and dimension j. and These represent the frequencies of occurrence of dimension i and dimension j individually, respectively.

[0073] Unlike traditional collaborative filtering, which only calculates user-item similarity, this application extends the calculation scope to the cross-correlation between four business dimensions and allows for the setting of weighting coefficients tailored to the characteristics of the electronics and information industry. For example, when calculating the correlation between product parameters and customer industries, different product parameters can be assigned higher weights based on industry technical standards. For instance, the weight of the stability dimension in the intelligent manufacturing industry can be set to 1.2 times that of ordinary industries to enhance the identification capability of key industry parameters.

[0074] The classification layer introduces a lightweight random forest classifier to complete the mapping classification from input features to output labels. Specifically, the input features are: customer industry (encoded one-hot), purchasing behavior (including features such as purchase volume, purchase cycle, decision chain length, and budget range); output labels are: product parameters (functions, interfaces, power consumption, size, stability, and price range), and demand features (replacement demand, upgrade demand, supporting demand, and new project demand).

[0075] The random forest classifier in this embodiment has 30-100 decision trees. This range strikes a balance between model accuracy and computational efficiency. Fewer than 30 trees may lead to underfitting, while more than 100 trees increase training overhead with limited accuracy improvement. The maximum depth of the random forest classifier is 5-10 layers. Limiting the tree depth prevents overfitting while ensuring sufficient learning from the training data.

[0076] Compared to traditional random forests, the lightweight random forest used in this application features a lightweight design in the following aspects: First, the number of decision trees is limited to 30-100 to avoid model bloat due to excessive tree numbers; second, the tree depth is limited to 5-10 layers to prevent overfitting and reduce model complexity; third, a pre-screening mechanism is used in the feature selection stage to retain only features strongly related to the needs of the electronics and information industry, reducing redundant computation. These designs enable the model to be trained and inferred on ordinary servers (such as Intel Core i5, 8GB RAM) without the need for high-performance computing power.

[0077] When building the model, the anonymized simulated customer data is divided into training set, validation set and test set according to the proportion. Gradient descent method is used for iterative training. Training stops when the accuracy of the validation set reaches the preset threshold.

[0078] Specifically, in this embodiment of the application, the de-identified simulated customer data generated in step S101 is divided into three subsets according to the proportions: the training set is used for model parameter learning, accounting for 70%; the validation set is used for hyperparameter tuning and early stopping judgment, accounting for 20%; and the test set is used for final model performance evaluation, accounting for 10%.

[0079] In this application, precision, recall, and F1 score are used as model evaluation metrics, and the test set evaluation metrics are required to reach a preset threshold.

[0080] Specifically, this application embodiment uses gradient descent for iterative training. First, the parameters of each decision tree in the random forest classifier are initialized; then, training samples are input in batches, and the prediction error is calculated; next, backpropagation updates the model weights; after each round of training, the accuracy is calculated on the validation set; finally, training is stopped when the accuracy on the validation set reaches a preset threshold (e.g., 85%) or no longer improves after several consecutive rounds, to prevent overfitting.

[0081] Optionally, this embodiment employs mini-batch gradient descent for iterative training, with a batch size of 32 and a learning rate of 0.01. Mini-batch gradient descent combines the stability of batch gradient descent with the speed of stochastic gradient descent, improving training efficiency while ensuring convergence stability.

[0082] This application uses general evaluation metrics for classification tasks to measure model performance. Among them, precision is the proportion of samples predicted as positive that are actually positive, measuring the accuracy of the prediction; recall is the proportion of samples that are actually positive that are correctly predicted, measuring the comprehensiveness of the prediction; and the F1 score is the harmonic mean of precision and recall, which comprehensively evaluates the model performance.

[0083] After the model is trained, it is required to achieve the following performance standards on the test set: precision ≥ 80%, recall ≥ 78%, and F1 score ≥ 79%. This standard ensures that the model has practical application value and can accurately uncover customer needs.

[0084] The model dimensions in this application are designed to closely align with the characteristics of the electronics and information industry, deeply binding product parameters with customer needs, and achieving higher prediction accuracy than general marketing models. The association weights output by the weighted collaborative filtering algorithm in this application can intuitively reflect the strength of the relationship between each dimension, making it easier for business personnel to understand the basis of model decision-making. Furthermore, the random forest classifier has a moderate number of parameters and can be trained and inferred on ordinary servers without the need for high-performance computing power, which is in line with the technical capabilities of small and medium-sized enterprises.

[0085] In step S103, customer data is analyzed based on the four-dimensional mapping model to uncover potential customer needs, and a demand activation index is generated to quantify demand intensity, outputting a list of potential needs.

[0086] Step S103 of this application embodiment aims to solve the technical problems of ambiguous customer demand identification and inability to distinguish demand priorities in related technologies. By analyzing customer data through a trained four-dimensional mapping model, it achieves accurate mining of potential demands and quantitative assessment of demand intensity, providing accurate input basis for the generation of subsequent marketing plans.

[0087] Based on the four-dimensional mapping model trained in step S102, this step S103 analyzes customer data to uncover potential customer needs. Specifically, firstly, known information about the target customer is input into the four-dimensional mapping model, including: the customer's industry, such as intelligent manufacturing, security, etc.; purchasing behavior, such as historical purchase categories, purchase frequency, budget range, etc.; inquiry keywords, i.e., the keyword text used by the customer in recent inquiry requests; and demand description, i.e., the demand text expressed by the customer in communication.

[0088] In the above information, the inquiry keywords and demand descriptions need to be converted into structured feature vectors and then input into the model according to the preprocessing layer method in step S102.

[0089] The four-dimensional mapping model performs association layer matching based on input features. It uses trained weighted collaborative filtering weights to calculate the similarity between current customers and historical customers in each dimension, thus identifying the demand patterns of similar customer groups. The classification layer performs prediction. The random forest classifier takes the current customer's industry and behavioral characteristics as input and outputs corresponding product parameter predictions and demand feature classifications. Furthermore, it identifies demand types. Based on the demand feature labels output by the classification layer, it identifies the customer's potential demand types, including replacement demand, upgrade demand, supporting demand, or new project addition demand.

[0090] For the identified potential needs, the model further refines the parameter requirements of the required products. For example, for smart manufacturing customers identified as having "upgrade needs", the model can output: "Required product parameters - low power consumption (≤5W), industrial-grade stability, RS485 interface", clearly specifying the product specifications.

[0091] Optionally, in some embodiments, the demand activation index is quantitatively scored based on customer inquiry behavior, purchasing records and industry driving factors, and demand priority levels are divided according to the scoring results.

[0092] Understandably, in order to quantify the urgency of customer needs, this application embodiment constructs a demand activation index to achieve a quantitative assessment of demand intensity.

[0093] The demand activation index is calculated comprehensively based on inquiry behavior factors, purchasing record factors, and industry driving factors. Inquiry behavior factors reflect customers' recent proactive demand expression, including: the number of inquiries in the past 30 days (weight 30%), the matching degree between inquiry content and historical purchase categories (weight 15%), and the presence of demand intensity words in inquiry keywords (such as "urgently needed," "as soon as possible," "replacement," etc.) (weight 10%). Purchasing record factors reflect customers' historical purchasing patterns, including: whether the historical repurchase intervals for similar products are close (weight 15%), the regularity of the purchasing cycle and the current stage of the cycle (weight 10%), and whether historical purchasing volume shows an upward trend (weight 5%). Industry driving factors reflect the stimulating effect of the external environment on demand, including: whether there are technical upgrade standards or policy requirements in the customer's industry (such as energy efficiency standard improvements) (weight 5%), whether there is a trend of new technology substitution within the industry (weight 5%), and the impact of recent industry exhibitions, technology forums, and other activities (weight 5%).

[0094] The demand activation index in this embodiment is calculated using a weighted summation method, with a value range of 0-100 points. The calculation formula is as follows:

[0095] ;

[0096] in, The normalized score (0-100) for the i-th factor. The weight of this factor (the sum of the weights of all factors is 100%).

[0097] The scores for each factor are calculated based on actual data and preset rules. For example: Number of inquiries: 0 times = 0 points, 1 time = 60 points, 2 times = 80 points, 3 times or more = 100 points; Repurchase interval: >3 months from the expected repurchase time = 0 points, 1-3 months = 50 points, <1 month = 90 points, and more than the expected time = 100 points; Policy driving factor: No relevant policy = 0 points, industry initiative = 50 points, and mandatory policy = 100 points.

[0098] Based on the demand activation index score, customer demands are divided into the following three priority levels:

[0099] High Priority (80-100 points): Urgent needs require immediate attention. Typical characteristics include: multiple recent inquiries, an approaching window for repeat purchases of similar products, and being driven by mandatory industry policies. These customers have a high conversion rate and should be prioritized for marketing.

[0100] Medium Priority (50-79 points): Potential demand exists, requiring nurturing and follow-up. Typical characteristics include: long-term cooperation but no recent inquiries, industry-wide product upgrade trends, and the procurement cycle is about to enter the decision-making phase. These clients require regular contact to maintain focus.

[0101] Low priority (0-49 points): Needs are not urgent and can be maintained routinely. Typical characteristics include: stable procurement of existing products, no need for upgrades or replacements, and no near-term procurement plans. These customers do not require excessive marketing resources.

[0102] Based on the above analysis results, step S103 can automatically generate a structured and executable list of potential requirements. Table 1 is the requirement list table of this application embodiment. As shown in Table 1, the list includes the following core fields:

[0103] Table 1

[0104] Field Name Field Description Example Customer Identification De-identified Customer Number CUST_001 Industry Affiliation Customer Industry Classification Smart Manufacturing Demand Type Identified Demand Type Upgrade Demand Required Product Parameters Specific Product Specification Requirements Power consumption ≤ 5W, industrial-grade stability, RS485 interface Demand Activation Index Quantified Score (0-100) 92 Priority Level High / Medium / Low High Optimal Touch Point Suggested Marketing Intervention Time Point Before next quarter's procurement Associated Product Recommendations Matching Product Models or Categories Industrial-grade low-power controller (Model: XXX)

[0105] Table 1 is stored in a structured data format and can be directly imported into the marketing plan generation module in step S104 to achieve automated connection from demand mining to marketing execution.

[0106] Based on the reasoning capabilities of the four-dimensional mapping model, this application embodiment can identify potential needs that are not explicitly expressed by customers but objectively exist, resulting in a higher recall rate than traditional keyword matching methods. The demand activation index transforms the vague "urgency of demand" into a comparable value, facilitating the efficient allocation of marketing resources. The potential demand list in this application embodiment adopts a standardized format, which can be directly used in downstream automated processes, reducing manual intervention. Furthermore, the components of the demand activation index are clearly defined, and the source of the score can be traced, enhancing business personnel's trust in the system's recommendations.

[0107] In step S104, a precise marketing plan adapted to the electronic information industry scenario is automatically generated based on the potential demand list.

[0108] Step S104 of this application embodiment aims to solve the technical problems in the related art that marketing plans rely on manual planning, are inefficient and lack pertinence. Based on the potential demand list output in step S103, a precise marketing plan adapted to the electronic information industry scenario is automatically generated, realizing the automated connection from demand identification to marketing execution.

[0109] In step S104, using the potential demand list as the core input, and combining customer attributes and demand characteristics, a personalized and precise marketing plan is automatically generated through a combination of rule engine and template filling. The generation of the marketing plan in this embodiment follows the principles of industry adaptation, demand matching, behavioral patterns, and budget adaptation. Specifically, the content and style of the marketing plan must be adapted to the business scenario and communication habits of the customer's industry; product recommendations and selling point design must accurately correspond to the demand types and parameter requirements identified by the customer; the selection of outreach channels and timing must conform to the customer's historical purchasing behavior patterns; and business strategies must be designed within the customer's budget range to ensure the feasibility of the plan.

[0110] Optionally, in some embodiments, the precision marketing scheme includes product recommendation information and marketing execution strategy information, the marketing execution strategy information being generated based on customer attributes and demand characteristics.

[0111] The product recommendation information is based on the customer's needs and required product parameters, matching the most suitable combination of electronic information products from the product library. Specifically, parameter matching is performed first, comparing the "required product parameters" field in the potential demand list with the product specifications in the product library to filter out product models that meet all key parameter requirements. For example, if the customer's requirements are "power consumption ≤ 5W, industrial-grade stability, RS485 interface," then products that meet all three requirements will be selected. Next, combination recommendations are made, suggesting suitable product combinations based on the demand type. For example, for replacement needs: a single replacement product is recommended, emphasizing compatibility and performance improvement with the original product; for upgrade needs: products or product combinations with enhanced performance are recommended, emphasizing functional expansion and improved user experience; for complementary needs: combinations of main products and accessories are recommended to ensure system integrity and compatibility; for new project additions: overall solutions are recommended, covering core equipment and auxiliary accessories. When multiple matching products exist, they are comprehensively ranked according to dimensions such as matching degree (number of parameter matches), profit margin, and inventory status, placing the optimal choice first in the recommendation.

[0112] Marketing execution strategy information is generated based on customer attributes and needs, aiming to guide marketers to reach customers and drive conversions in the most effective way. This application's embodiment combines customer industry characteristics and demand types to generate product selling points tailored to specific application scenarios. Specifically, for different industries, it highlights the product's core value in that scenario. For example, in the smart manufacturing industry: emphasizing "industrial-grade stability, adaptability to industrial control scenarios, and reduced energy consumption"; in the security industry: emphasizing "high-definition night vision, low latency, and reliable operation around the clock"; and in the Internet of Things (IoT) industry: emphasizing "low power consumption, remote management, and easy integration."

[0113] The marketing execution strategy also considers the adaptation to different types of needs, that is, designing differentiated communication entry points for different types of needs. For example, for replacement needs: highlight "seamless compatibility with existing equipment and significant performance improvement"; for upgrade needs: emphasize "empowered by new technologies and not outdated for the next 3 years"; for supporting needs: highlight "system integrity and plug-and-play"; for new project needs: emphasize "overall solutions and one-stop service".

[0114] Marketing execution strategies also involve quantifying value propositions, which means transforming product advantages into quantifiable customer value, such as "30% lower energy consumption compared to existing products" or "lowest failure rate in the industry (below 0.5%)", to enhance persuasiveness.

[0115] Furthermore, based on customer purchasing behavior patterns, we recommend the best channels and timing for reaching out. According to customer historical interaction preferences and industry practices, we recommend suitable initial outreach channels. Specifically, for customers with short decision-making chains and mature relationships, we recommend telephone outreach; for scenarios requiring detailed technical information, we recommend email outreach; for younger decision-makers and scenarios prioritizing communication efficiency, we recommend WeChat / instant messaging; for complex product sales requiring in-depth technical communication, we recommend technical exchange meetings; and for high-value products requiring physical verification, we recommend on-site demonstrations. Additionally, based on customer purchasing cycle patterns and external events, we recommend optimal outreach times, such as the early stages of the purchasing cycle, after industry exhibitions, after policy announcements, and before quarterly / annual budgets are finalized.

[0116] This application's embodiments combine customer budget ranges and purchase volumes to recommend reasonable price ranges and preferential strategies. Specifically, based on customer budget ranges and product pricing systems, it recommends competitive price ranges. For example, for budget-sensitive customers, it recommends basic configuration products or standard prices; for value-oriented customers, it recommends high-end products or premium prices, highlighting long-term value. Differentiated preferential policies are designed based on purchase volume and customer value, such as bulk purchase discounts, long-term cooperation discounts, first-time cooperation incentives, and upgrade programs for existing customers. Furthermore, suitable payment methods can be recommended based on customer credit and purchase amount, such as standard payments, credit payments, and installment payments.

[0117] The precision marketing plan generated in step S104 is output in the form of a structured document, supporting multiple formats to adapt to different use cases. For marketers, a complete plan is output for sales personnel to communicate and follow up with customers; for customers, a simplified version is output, which can be directly sent to customers for reference; for system integration, a structured data format (JSON / XML) is output for CRM (Customer Relationship Management) and other systems to use directly.

[0118] It should be noted that the precision marketing solutions generated by this application support multiple output formats. In addition to structured documents and JSON / XML data formats, they can also be seamlessly integrated with mainstream CRM systems (such as SalesEasy, Fenxiang Sales, and WeChat Enterprise CRM) through standard API (Application Programming Interface). The marketing solutions can be directly imported into the CRM system to generate sales tasks, realizing full-process automation from demand discovery to marketing execution.

[0119] Therefore, this application reduces the time required for traditional marketing plan development (which typically takes hours) to minutes through automatic generation, significantly improving marketing response speed, and ensuring the plan content is highly aligned with customer needs. The output plan in this application includes specific products, scripts, channels, timing, and business strategies, which marketers can directly execute without secondary processing. Marketing execution results can be fed back to the closed-loop optimization module in step S105 to continuously optimize the plan generation rules.

[0120] In step S105, based on marketing execution feedback data, the weights of the four-dimensional mapping model are updated to complete the demand mining and iterative optimization of the marketing plan.

[0121] Step S105 of this application embodiment aims to solve the technical problem in the related art that the marketing model is static and fixed and cannot learn and evolve from the execution results. By collecting marketing execution feedback data, the weights and mapping relationships of the four-dimensional mapping model are dynamically updated to continuously improve the accuracy of demand prediction and the adaptability of marketing solutions.

[0122] Step S105 uses the performance data generated during marketing execution as feedback signals. Through quantitative analysis, it identifies model prediction deviations and adjusts the parameter configuration of the four-dimensional mapping model to achieve continuous model evolution. Specifically, after the marketing plan is executed, the system automatically or manually inputs outreach response data, intent conversion data, and sales data. Outreach response data reflects the customer's initial reaction to the marketing outreach, including: outreach success rate (call connection rate, email open rate, WeChat reply rate), interest expression (whether they requested materials, inquired about prices, or requested a demonstration), and response timeliness (how long after outreach the customer responded). Intent conversion data reflects the degree of conversion from interest to intent, including: needs confirmation (whether they approve of the recommended product's suitability), solution acceptance (whether they accept the recommended solution), and business negotiation progress (whether they have entered the quotation and negotiation stage). Sales data reflects the final marketing results, including: whether a deal was closed (yes / no), transaction amount (actual contract amount), transaction cycle (number of days from outreach to signing), and purchased products (actual product model and quantity purchased). After the feedback data is collected, it needs to be standardized and converted into a structured feature vector with the same format as the training data, so as to facilitate comparison and analysis with the model prediction results.

[0123] Then, the actual marketing results are compared with the prediction results of step S103 and the recommended solutions of step S104 to identify the sources of deviation. Specifically, this includes: demand prediction deviation, i.e., whether the actual demand type matches the predicted demand type? If not, analyze the misjudgment patterns of the model at the demand classification layer; parameter matching deviation, i.e., whether there are differences between the actual purchased product parameters and the recommended product parameters? If so, analyze which parameter dimensions deviate (e.g., the customer ultimately chose products with different power consumption levels); marketing strategy deviation, i.e., whether the actual effective outreach channels, timing, and wording match the recommended strategy? If there are differences, identify which strategy elements failed to predict; intensity scoring deviation, i.e., whether the distribution of demand activation index among the actual converted customers matches the predicted priority? Analyze whether the conversion rate of high-priority customers is significantly higher than that of low-priority customers to verify the quantitative effectiveness of demand intensity.

[0124] Based on the deviation analysis results, this application embodiment uses an incremental learning approach to update the weights of the four-dimensional mapping model. Specifically, for the weighted collaborative filtering algorithm of the correlation layer, the correlation weights between dimensions are adjusted according to the actual transformation data. When a certain dimension combination (such as "intelligent manufacturing industry" and "industrial-grade stability parameters") frequently shows a positive correlation in the actual transformation, its correlation weight is increased; when the predicted weight of a certain dimension combination deviates systematically from the actual transformation result, the weight value is corrected according to the gradient descent direction. The weight update formula is:

[0125] ;

[0126] in, The updated association weights between dimension i and dimension j. The association weights before the update. The learning rate controls the step size of weight adjustment. This represents the actual relevance of this dimension combination in actual conversion. This represents the relevance predicted by the model.

[0127] For the random forest model in the classification layer, this application embodiment uses incremental training or weight adjustment for updating. Incremental training uses newly generated feedback data (with labeled actual results) as new training samples to incrementally train the random forest, adjusting the node splitting parameters of the decision trees. This is suitable for scenarios with large amounts of feedback data and significant distribution variations. Weight adjustment keeps the tree structure unchanged, only adjusting the voting weights of each decision tree. For decision trees with high prediction accuracy, their voting weights are increased; for decision trees with large prediction bias, their weights are decreased or pruning is performed. This is suitable for scenarios with limited feedback data and a need for rapid response.

[0128] Furthermore, based on the correlation analysis between the actual conversion rate and the demand activation index, the weight allocation of the index components is optimized. Specifically, the correlation coefficient between each component (inquiry behavior, purchase records, industry drivers) and the actual conversion rate is calculated; the weight of components with high correlation coefficients is increased, and the weight of components with low correlation coefficients is decreased; the weights are then renormalized to ensure that the sum of the weights of all components is 100%. For example, if the analysis finds that the impact of "industry driving factors" on the conversion rate is higher than the preset weight, its weight is increased from 15% to 20%, and the weights of other factors are decreased accordingly.

[0129] Optionally, in some embodiments, iterative optimization further includes adjusting the mapping relationship of the four-dimensional mapping model based on marketing execution feedback data.

[0130] Specifically, when it is found that the existing four-dimensional division cannot fully characterize certain new demand patterns, the dimension definition can be dynamically expanded or subdivided. For example, if the emerging demand for "edge computing" is difficult to accurately classify within the existing "Internet of Things" and "industrial control" industries, a new subcategory of "edge computing" can be added; if customers' demand for "programmability" parameters becomes increasingly prominent, it can be added to the product parameter dimension; if a new demand pattern of "leasing instead of purchasing" emerges, "leasing demand" can be added to the demand characteristic dimension.

[0131] Furthermore, regarding the correlation between product parameters and demand characteristics, the correlation threshold can be adjusted based on actual conversion data. For example, if the power consumption of the products actually purchased by customers is generally higher than the "low power" threshold (≤5W) recommended by the model, the definition threshold for low power consumption can be adjusted to ≤7W; if there is a discrepancy between customers' understanding of "industrial-grade stability" and their actual purchasing behavior, the quantification standard of stability parameters can be adjusted.

[0132] The rules for generating marketing execution strategies are also incorporated into a closed-loop optimization system, continuously optimized based on performance feedback. Specifically, for selling point script templates, the common characteristics of high-conversion scripts are analyzed to optimize script generation rules; for reach channel weights, channel recommendation priorities are adjusted based on the actual conversion rates of each channel; and for business strategy parameters, the pricing range calculation model is optimized based on the deviation between the transaction price and the initial quote.

[0133] The iterative optimization process of this application embodiment can be summarized as follows: First, collect feedback data during the marketing execution process to ensure data integrity, compare the predicted results with the actual results, and identify the sources and patterns of deviation; then, update the model weights and mapping relationships based on the analysis results; next, apply the optimized model to the next batch of customers to verify the improved effect; finally, solidify the verified effective optimization scheme into the model to complete this round of iteration.

[0134] It should be noted that the above cycle can be executed automatically according to a preset period (such as weekly / monthly), or it can be triggered after the feedback data accumulates to a certain scale.

[0135] It should also be noted that the method of this application supports multiple lightweight deployment methods: First, it supports local server deployment, with a minimum hardware configuration of Intel Core i5 processor, 8GB memory, and 100GB storage space, allowing SMEs to quickly implement it using existing IT resources; Second, it supports cloud SaaS deployment, eliminating the need for enterprises to build their own servers, and can be accessed and used through a browser; Third, it supports lightweight container deployment (such as Docker), enabling rapid packaging and one-click deployment, reducing the technical implementation threshold.

[0136] Therefore, the model in this application embodiment can continuously learn from actual marketing results, adapt to market changes and evolving customer needs, and avoid the decline in accuracy caused by model aging. As feedback data accumulates, the accuracy of demand forecasting and the adaptability of marketing plans show a continuous upward trend. In other words, even if the initial model is trained based on simulated data, it can gradually approach the optimal state of real business scenarios through closed-loop optimization.

[0137] The following specific embodiments illustrate the execution process of a method for intelligent mining and precise marketing of the needs of small and medium-sized enterprises in the field of electronic information, as described in this application.

[0138] This embodiment focuses on intelligent demand mining and precision marketing for an electronic information engineering company.

[0139] Step 1: Generate anonymized simulated customer data.

[0140] Specifically, this embodiment generates 1,000 anonymized customer data entries for small and medium-sized enterprises in the electronics information industry. These entries cover industry, product categories, inquiry keywords, purchase frequency, and budget range. The data is anonymized and contains no real company information. Some examples are shown below:

[0141] Customer 1: Industry: Intelligent Manufacturing, Company Size: Small, Purchase Category: Industrial Controllers, Purchase Frequency: Once per quarter, Budget Range: 5,000~10,000 RMB, Inquiry Keywords: Low Power Consumption, Industrial Grade, High Stability, Requirement Description: "The existing controller has high power consumption and needs to be replaced with a low-power product that is suitable for industrial scenarios".

[0142] Customer 2: Industry: Security, Company size: Micro, Purchase category: Monitoring module, Purchase frequency: Once every six months, Budget range: 2000~5000 yuan, Inquiry keywords: High-definition, night vision, low latency, Requirement description: "Need high-definition monitoring module suitable for small security scenarios, with cost controlled within 5000 yuan";

[0143] The remaining 998 data entries were generated using this standard to ensure data diversity and industry relevance.

[0144] Step 2: Construct a four-dimensional mapping model.

[0145] Specifically, based on 1000 simulated data points, an improved weighted collaborative filtering and lightweight random forest algorithm were used to construct a four-dimensional mapping model of product parameters, customer industry, purchasing behavior, and demand characteristics. Training was completed under the following conditions: First, the data was divided into a training set of 700 data points, a validation set of 200 data points, and a test set of 100 data points, using a 7:2:1 ratio. Then, data preprocessing was performed: unstructured text such as "low-power controller" was converted into feature vectors [power consumption: 0.2 (normalized), product type: controller]; industry tags such as "intelligent manufacturing" and "security" were one-hot encoded; and the budget range of "5000~10000 yuan" was normalized to 0.5. Model training was conducted on a server with an Intel Core i7-12700H processor and 16GB of memory, using batch processing. With a size of 32 and a learning rate of 0.01, the model was trained for 100 iterations. Training stopped when the accuracy on the validation set reached 86%. Model evaluation was then performed, with a test set accuracy of 81%, recall of 79%, and F1 score of 80%, meeting the preset evaluation criteria. Finally, dimensions were defined and associated. In the final model, the correlation weight between the stability / power consumption parameters of the intelligent manufacturing industry and the industrial controller was 0.85, and the correlation weight between the security industry and the resolution / night vision parameters of the monitoring module was 0.82, which meets the requirements of the electronic information industry.

[0146] Step 3: Discover demand and predict its intensity.

[0147] Based on the trained model and simulated customer data, we can uncover potential customer needs and generate a demand activation index and a list of potential needs. The output results are as follows:

[0148] Customer 1: Demand activation index 92 points (high), potential demand: industrial controller replacement demand, required product parameters: low power consumption (≤5W), industrial-grade stability, RS485 interface, best time to reach out: early stage of procurement in the next quarter;

[0149] Customer 2: Demand activation index 75 (medium), potential demand: monitoring module upgrade demand, required product parameters: 1080P high definition, night vision distance ≥50 meters, latency ≤0.5 seconds, best time to reach out: within 1 month (after the exhibition).

[0150] Generate a complete "Potential Requirements List" containing all 1,000 customer requirements.

[0151] Step 4: Generate a precise marketing plan.

[0152] Based on the above list of needs, customized and precise marketing plans were generated for Client 1 and Client 2, tailored to the electronic information industry scenario. The specific plans are as follows:

[0153] Customer 1 Marketing Plan: Recommended Product: Industrial-grade low-power controller (Model: XXX, 4W power consumption, RS485 interface, industrial-grade stability); Selling Points: "This controller is suitable for smart manufacturing scenarios. Its low-power design reduces enterprise energy costs, and its industrial-grade stability meets the requirements for 24-hour continuous operation, perfectly matching your replacement needs"; Reach Channels: Telephone + technical information email; Reach Timing: 15 days before the next quarter's purchase; Business Strategy: Enjoy a 10% discount for bulk purchases of ≥5 units.

[0154] Customer 2 Marketing Plan: Recommended Product: High-definition night vision monitoring module (Model: XXX, 1080P high-definition, night vision distance 60 meters, latency 0.3 seconds); Selling Points: "This module is suitable for small-scale security scenarios, offering high-definition night vision and low latency, with a cost controlled within 4500 yuan, meeting your upgrade needs. Installation is convenient and requires no modification to existing equipment"; Reach Channels: WeChat + technical exchange; Reach Timing: Within 7 days after the exhibition; Business Strategy: First-time cooperation enjoys an 8.5% discount.

[0155] Step 5: Implement closed-loop optimization.

[0156] The simulated marketing execution results showed that Customer 1 had a response rate of 80% and a conversion rate of 50%, while Customer 2 had a response rate of 60% and a conversion rate of 30%. Based on the feedback data, the mapping weight of "demand characteristics - product parameters" in the model was adjusted (the weight of the low power consumption parameter was increased from 0.85 to 0.88) to improve the accuracy of subsequent demand forecasting and complete closed-loop optimization. If there is actual execution data in the future, only manual entry is required to automatically complete the optimization, without the need for enterprise cooperation throughout the process.

[0157] This embodiment fully replicates all the steps of this application. All steps can be implemented in practice without requiring any data or cooperation from the enterprise. It is highly practical and can be directly replicated and applied to other small and medium-sized enterprises in the electronics and information industry. This demonstrates the practicality and scalability of this application.

[0158] This application proposes a method for intelligent demand mining and precision marketing for small and medium-sized enterprises (SMEs) in the electronics and information industry. This method addresses the technical challenges of demand mining and low marketing precision faced by SMEs in the electronics and information industry due to data fragmentation and a lack of digital capabilities. It involves five core steps: generating anonymized simulated customer data, constructing a four-dimensional mapping model specific to the electronics and information industry, mining and quantifying customer needs, automatically generating precision marketing plans, and optimizing the model through a closed-loop feedback loop. The method requires no core data from the enterprise and no enterprise participation in the execution process, significantly lowering the technical application threshold while ensuring data compliance. It achieves the technical effects of accurate demand identification, automatic marketing plan generation, and continuous model evolution.

[0159] Secondly, referring to the accompanying drawings, an intelligent demand mining and precision marketing device for electronic information SMEs is described according to an embodiment of this application.

[0160] Figure 2 This is a block diagram of an electronic information small and medium-sized enterprise demand intelligent mining and precision marketing device according to an embodiment of this application.

[0161] like Figure 2 As shown, the intelligent mining and precision marketing device 10 for the needs of small and medium-sized enterprises in the electronics and information industry includes: a data generation module 100, a modeling module 200, an intelligent mining module 300, a solution generation module 400, and an optimization module 500.

[0162] The data generation module 100 is connected to the modeling module 200, the modeling module 200 is connected to the intelligent mining module 300, the intelligent mining module 300 is connected to the solution generation module 400, the solution generation module 400 and the optimization module 500 are bidirectionally connected, and the optimization module 500 is also connected to the modeling module 200 and the intelligent mining module 300 respectively, for feedback and updating of model weights and mapping relationships.

[0163] Specifically, the data generation module 100 is used to generate de-identified simulated customer data for small and medium-sized enterprises in the electronics and information industry. The de-identified simulated customer data is generated based on the characteristics of customers in the electronics and information industry and does not contain any real business secrets of the enterprises. It includes: the industry to which the customer belongs, the category of goods purchased, the frequency of purchase, the keywords of inquiry and the description of demand.

[0164] Modeling module 200 is used to build a four-dimensional mapping model of product parameters, customer industry, purchasing behavior, and demand characteristics based on de-identified simulated customer data.

[0165] The intelligent mining module 300 is used to analyze customer data based on a four-dimensional mapping model, mine potential customer needs, generate a demand activation index to quantify demand intensity, and output a list of potential needs.

[0166] The solution generation module 400 is used to automatically generate precise marketing solutions adapted to the electronic information industry scenario based on a list of potential needs.

[0167] The optimization module 500 is used to update the weights of the four-dimensional mapping model based on marketing execution feedback data, and to perform iterative optimization of demand mining and marketing plans.

[0168] It should be noted that the foregoing explanation of an embodiment of a method for intelligent mining and precise marketing of the needs of small and medium-sized enterprises in the electronic information industry also applies to an embodiment of an intelligent mining and precise marketing device for the needs of small and medium-sized enterprises in the electronic information industry, which will not be repeated here.

[0169] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0170] The memory 301, the processor 302, and the computer program stored on the memory 301 and capable of running on the processor 302.

[0171] When the processor 302 executes the program, it implements the method for intelligent mining and precise marketing of the needs of small and medium-sized electronic information enterprises provided in the above embodiments.

[0172] Furthermore, electronic devices also include:

[0173] Communication interface 303 is used for communication between memory 301 and processor 302.

[0174] The memory 301 is used to store computer programs that can run on the processor 302.

[0175] The memory 301 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0176] If the memory 301, processor 302, and communication interface 303 are implemented independently, then the communication interface 303, memory 301, and processor 302 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0177] Optionally, in a specific implementation, if the memory 301, processor 302, and communication interface 303 are integrated on a single chip, then the memory 301, processor 302, and communication interface 303 can communicate with each other through an internal interface.

[0178] Processor 302 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0179] This application also provides a computer program product, on which a computer program is stored, which, when executed by a processor, implements the above-described method for intelligent mining and precise marketing of the needs of small and medium-sized electronic information enterprises.

[0180] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0181] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0182] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0183] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0184] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0185] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for intelligent demand mining and precision marketing for small and medium-sized enterprises in the electronics and information industry, characterized in that, Includes the following steps: Generate anonymized simulated customer data for small and medium-sized enterprises in the electronics and information industry. The anonymized simulated customer data is generated based on the characteristics of customers in the electronics and information industry and does not contain any real business secrets of the enterprises. It includes: the industry to which the customer belongs, the category of goods purchased, the frequency of purchase, the keywords for inquiry and the description of demand. Based on the anonymized simulated customer data, a four-dimensional mapping model of product parameters, customer industry, purchasing behavior, and demand characteristics is constructed. Based on the aforementioned four-dimensional mapping model, customer data is analyzed to uncover potential customer needs, and a demand activation index is generated to quantify demand intensity, outputting a list of potential needs. Based on the list of potential needs, a precise marketing plan adapted to the electronic information industry scenario is automatically generated; Based on marketing execution feedback data, the weights of the four-dimensional mapping model are updated to complete the demand mining and iterative optimization of the marketing plan.

2. The method according to claim 1, characterized in that, The number of anonymized simulated customer data generated shall not be less than 1,000, and the industries to which the customers belong include: intelligent manufacturing, security, Internet of Things, industrial control and consumer electronics.

3. The method according to claim 1, characterized in that, The four-dimensional mapping model includes: product parameter dimension, customer industry dimension, purchasing behavior dimension, and demand characteristic dimension; among which... The product parameter dimensions include: the functions, interfaces, power consumption, size, stability, and price range of electronic information products; The customer industry dimension includes: sub-sectors of the electronic information industry; The procurement behavior dimensions include: procurement volume, procurement cycle, decision chain length, and budget range; The required characteristics include: replacement requirements, upgrade requirements, supporting requirements, and new project requirements.

4. The method according to claim 3, characterized in that, The four-dimensional mapping model employs a combination of an improved weighted collaborative filtering algorithm and a lightweight random forest classifier, specifically including: In the preprocessing layer, the de-identified simulated customer data is standardized, unstructured inquiry keywords and demand descriptions are transformed into structured feature vectors, discrete data is one-hot encoded, and continuous data is normalized. In the association layer, an improved weighted collaborative filtering algorithm is used to calculate the association weights among the four dimensions. This improved algorithm extends the calculation scope to include cross-associations among the product parameter dimension, customer industry dimension, purchasing behavior dimension, and demand characteristic dimension. Furthermore, weighting coefficients are set for different dimension combinations based on the characteristics of the electronics and information industry. The weight calculation formula is as follows: ； in, Let i be the association weight between dimension i and dimension j. Let i be the frequency of the simultaneous occurrence of dimension i and dimension j. and These represent the frequencies of occurrence of dimension i and dimension j individually, respectively. In the classification layer, a lightweight random forest classifier is introduced, which takes customer industry and purchasing behavior as input features and product parameters and demand features as output labels to complete the four-dimensional mapping classification.

5. The method according to claim 4, characterized in that, The lightweight random forest classifier has 30-100 decision trees and a maximum depth of 5-10 layers. When constructing the model, the anonymized simulated customer data is divided into training set, validation set and test set according to the proportion. Gradient descent method is used for iterative training. Training stops when the accuracy of the validation set reaches a preset threshold. Precision, recall, and F1 score are used as model evaluation metrics, and the test set evaluation metrics are required to reach the preset thresholds.

6. The method according to claim 1, characterized in that, The demand activation index is based on a quantitative score of customer inquiry behavior, purchasing records and industry driving factors, and the demand priority level is divided according to the score results.

7. The method according to claim 1, characterized in that, The precision marketing solution includes product recommendation information and marketing execution strategy information, which is generated based on customer attributes and needs.

8. The method according to claim 1, characterized in that, The iterative optimization also includes adjusting the mapping relationship of the four-dimensional mapping model based on the marketing execution feedback data.

9. The method according to claim 1, characterized in that, The demand activation index is calculated by weighting and summing factors such as inquiry behavior, procurement records, and industry drivers. The score ranges from 0 to 100 points, and demand priority levels are determined based on the score.

10. A device for intelligent demand mining and precision marketing for small and medium-sized electronic information enterprises, characterized in that, include: The data generation module is used to generate anonymized simulated customer data for small and medium-sized enterprises in the electronics and information industry. The anonymized simulated customer data is generated based on the characteristics of customers in the electronics and information industry and does not contain any real business secrets of the enterprises. It includes: the customer's industry, purchase category, purchase frequency, inquiry keywords and demand description. The modeling module is used to construct a four-dimensional mapping model of product parameters, customer industry, purchasing behavior, and demand characteristics based on the anonymized simulated customer data. The intelligent mining module is used to analyze customer data based on the four-dimensional mapping model, mine potential customer needs, generate a demand activation index to quantify demand intensity, and output a list of potential needs. The solution generation module is used to automatically generate precise marketing solutions adapted to the electronic information industry scenario based on the potential demand list. The optimization module is used to update the weights of the four-dimensional mapping model based on marketing execution feedback data, and to perform iterative optimization of demand mining and marketing plans.

Images