An information management system and management method for business customers

Abstract

The application discloses an enterprise customer information management system and a management method, and relates to the technical field of enterprise information management; the system comprises a customer basic information module, a space-time sequence analysis module, a geographic information correlation module, a data storage module and an interactive display module; the customer basic information module accurately processes missing values and abnormal values, and improves data quality and management efficiency in combination with distributed storage; an LSTM model integrates multi-dimensional features, monthly updates guarantee procurement time prediction accuracy, and helps enterprises to reasonably plan; the geographic information correlation module directly presents regional value and the relationship between related enterprises through coordinate conversion, clustering analysis and GIS visualization, and optimizes resource allocation; in customer management, the system outputs procurement prediction and triggers collaborative early warning, and enhances customer interaction; a data verification feedback mechanism forms a closed loop, and continuously optimizes the model; each module flexibly communicates, the interface is friendly, and the enterprise customer control, market prediction and decision-making ability are comprehensively improved.

Description

Technical Field

[0001] This invention relates to the field of enterprise information management technology, specifically to an information management system and management method for enterprise customers. Background Technology

[0002] With increasingly fierce market competition, enterprises need to manage customer information more accurately and efficiently to improve customer satisfaction, enhance market competitiveness, and achieve sustainable development. Traditional enterprise customer information management methods have many limitations. Early manual recording and storage methods are not only inefficient but also prone to errors, failing to meet enterprises' needs for efficient, accurate, and secure customer information management. While some subsequent customer management systems have achieved a degree of informatization, most merely record customer information statically. These systems often focus only on basic customer attributes, such as customer name, industry, and company size, ignoring the temporal and spatial relationships of customer behavior. For example, a customer might exhibit a pattern of concentrated purchasing at the end of a quarter and frequently place orders simultaneously with three other companies in the same region. However, existing systems struggle to capture and analyze such temporal and spatial relationships, preventing enterprises from fully exploring the potential value and behavioral patterns of their customers.

[0003] In terms of data processing, existing systems rely on relatively simplistic data cleaning methods when dealing with raw customer data. For static information and dynamic behavioral data contained within the raw customer data, such as historical purchase records, purchase amounts, order times, and related companies, effective handling is difficult when missing or outlier values ​​are present. Traditional methods may simply ignore missing values ​​or crudely truncate outliers, failing to accurately reconstruct the true state of the data and thus affecting the accuracy of subsequent data analysis and decision-making. For example, when purchase amount data is missing, the inability to use appropriate methods to fill it in can lead to biases in customer behavior analysis models built upon this data.

[0004] From the perspective of customer behavior prediction, existing systems lack effective prediction mechanisms. When predicting future customer purchasing behavior, they do not fully consider the time-series characteristics and spatial correlations of customer behavior. Most systems rely solely on simple historical data statistics for prediction, failing to accurately capture the dynamic changes and complex patterns of customer behavior. For example, they cannot accurately predict the time interval of a customer's next purchase or the trend of changes in purchase amount, resulting in a lack of scientific basis for enterprises when formulating sales strategies and allocating resources. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides an information management system and method for enterprise customers, solving the problems mentioned in the background section.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: an information management system for enterprise customers, comprising:

[0009] The customer basic information module is used to collect and clean raw customer data;

[0010] The spatiotemporal sequence analysis module uses the LSTM algorithm to build a customer behavior prediction model.

[0011] The geographic information association module, in conjunction with the GIS system, completes the visualization analysis of regional value.

[0012] The data storage module adopts a distributed architecture to store structured and unstructured data;

[0013] The interactive display module provides users with a visual operation interface;

[0014] Each module communicates bidirectionally through a data interface.

[0015] As a further aspect of the present invention: the original customer data collected by the customer basic information module includes static information and dynamic behavioral data;

[0016] Static information includes customer name C n Industry I n Enterprise size S n ;

[0017] Dynamic behavioral data includes historical procurement records P n，t Purchase amount A n，t Order time T n，t Related companies U n，k .

[0018] Where n is the customer ID, t is the timestamp, and k is the associated company ID.

[0019] As a further aspect of the present invention: the data cleaning of the customer basic information module includes missing value processing and outlier value processing;

[0020] In the missing value handling, when the purchase amount is missing, it is filled with the average of the last 3 purchase amounts of the nth customer, using the following formula:

[0021] Among them, AP n，t For the filled purchase amount, A n，t-iLet i represent the actual purchase amount of the nth customer in the previous i purchases, where i represents the index of the purchase count.

[0022] In the outlier handling, purchases exceeding three standard deviations from the historical average of the nth customer are excluded; the comparison formula is as follows:

[0023] Where μ n，A σ is the average historical purchase amount of customer n. n，A This represents the corresponding standard deviation.

[0024] As a further aspect of the present invention: the spatiotemporal sequence analysis module includes a data sequence construction unit, a feature engineering unit, and an LSTM model training unit;

[0025] As a further aspect of the present invention: the data sequence construction unit organizes the historical purchase times of customer n into a time interval sequence ΔT in sequence. n =[Δt n，1 ,Δt n，2 ,……,Δt n，m ],in This indicates the time interval between two consecutive purchases.

[0026] As a further aspect of the present invention: the features extracted by the feature engineering unit include:

[0027] Time characteristic: Quarter identifier Q t Its value ranges from 1 to 4, and the month identifier is M. t Its value ranges from 1 to 12;

[0028] Behavioral characteristic: The rate of change between two consecutive purchase amounts, calculated using the following formula:

[0029] Among them, R n，z Let be the rate of change between the purchase amount of the z-th and z-1-th purchases.

[0030] As a further aspect of the present invention: the structure of the LSTM model training unit is as follows:

[0031] Input layer: Contains a time interval sequence ΔT n Time characteristics (Q) t M t ), behavioral characteristics R n，z The input dimension is 5;

[0032] Hidden layers: 2 LSTM layers, the first layer contains 64 neurons and the second layer contains 32 neurons, using the tanh activation function.

[0033] Among them, h tW is the hidden state at the current moment. h Let h be the weight matrix. t−1 The hidden state of x in the previous moment t For the current input features, b h For bias terms;

[0034] Output layer: A linear activation function is used to predict the time interval ΔtP of the next purchase. n，m+1 The loss function is the mean squared error:

[0035] Where j represents the sample index, and the value of j ranges from 1 to m, representing the summation of the prediction error for the j-th sample.

[0036] As a further aspect of the present invention: the geographic information association module includes a geographic coordinate mapping unit, a customer value calculation unit, a regional clustering analysis unit, and a GIS visualization unit;

[0037] The geographic coordinate mapping unit maps the latitude and longitude (L and L) of the customer's registered address. n =(Lng n ,Lat n );

[0038] Converted to planar coordinates using Gauss-Kruger projection (X) n ,Y n );

[0039] The formula is:

[0040] Where s is a preset scaling factor, and c x c y This is the offset, set according to the projection strip parameters.

[0041] As a further aspect of the present invention: the customer value calculation unit uses gross profit margin as the core value indicator, and the calculation formula is as follows:

[0042] Among them, G n Let C be the average gross profit margin for customer n. n，z The cost amount for this customer's zth purchase.

[0043] As a further aspect of the present invention: the regional clustering analysis unit uses a density clustering algorithm to divide the customer into regions according to geographical coordinates;

[0044] The radius is set to 5 kilometers, which is 5000 meters in plane coordinates, and the minimum sample size is set to 3.

[0045] For the customer set C within region r r ={n1,n2,……,nk},pass: Calculate the regional average gross profit margin (GP) r .

[0046] As a further aspect of the present invention: the GIS visualization unit displays the GP of each region on the electronic map in the form of a heat map. r Distribution, where darker colors indicate higher gross profit margins in different regions;

[0047] Simultaneously, mark the locations of enterprises that have placed orders with the target customer, and connect the related customers with line segments. The width of the line segment is positively correlated with the number of times they have placed orders together, as shown in the formula:

[0048] Among them W n，k N is the line segment width connecting clients n and k. n，k This refers to the number of times both parties place orders together.

[0049] A method for managing enterprise customer information, implemented through an enterprise customer information management system, includes the following stages:

[0050] Data collection and preprocessing stage: New data is added daily, cleaned, and then stored in a standardized manner;

[0051] Model training and update phase: The LSTM model is updated monthly, and the geographic information clustering results are updated quarterly;

[0052] Customer management application phase: Based on the model output, procurement forecast time and regional value distribution are generated, triggering collaborative procurement early warning;

[0053] Data validation and feedback phase: Calculate prediction error and dynamically optimize model weights.

[0054] As a further aspect of the present invention: the data acquisition and preprocessing stage includes:

[0055] Step S101: Automatically synchronize new data from the customer management system every day at midnight, including the day's purchase records and related company change information;

[0056] Step S102: Call the cleaning function of the customer basic information module to perform missing value filling and outlier removal, and generate a standardized data file;

[0057] Step S103: Store the processed data into the data storage module according to the "customer ID-timestamp" index, and retain the mapping relationship between the original data and the cleaned data.

[0058] As a further aspect of the present invention: the model training and update stage includes:

[0059] Step S201: Trigger the model update program of the spatiotemporal sequence analysis module on the 1st of each month to read customer behavior data from the past 12 months;

[0060] Step S202: Use the sliding window method to generate training samples and reconstruct the time interval sequence and feature matrix, where the window size is 90 days;

[0061] Step S203: Incrementally train the LSTM model with a learning rate of 0.001 and save the updated model parameters after 50 iterations.

[0062] Step S204: The geographic information association module updates customer coordinates and regional clustering results quarterly, and updates the GIS heat map layer synchronously.

[0063] As a further aspect of the present invention: the customer management application stage includes:

[0064] Step S301: When a user queries customer n, the system outputs the predicted time TP for the next 3 purchases. n，m+1 TP n，m+2 TP n，m+3 ,in ;

[0065] Step S302: Display a heat map showing the location distribution of related companies on the customer details page, and mark the average gross profit margin of the region and the number of times they place orders together;

[0066] Step S303: When the customer predicts the procurement time is at the end of the quarter, i.e., Q t When the number of related enterprises is 4 and at least 2 of them are included in the forecast period, the early warning mechanism will be triggered.

[0067] As a further aspect of the present invention: the data verification and feedback stage includes:

[0068] Step S401: After the actual purchase occurs, calculate the error between the predicted time and the actual time: ;

[0069] Step S402: When the error exceeds 7 days for 3 consecutive times, mark the customer as "model to be optimized" and increase the weight of its historical data in the next model update.

[0070] (III) Beneficial Effects

[0071] This invention provides an information management system and method for enterprise customers. Compared with the prior art, it has the following advantages:

[0072] First, data quality and management efficiency are improved. The customer basic information module effectively ensures data integrity and accuracy through targeted missing value handling (such as filling with the average of the last three purchase amounts) and outlier removal (based on a 3x standard deviation rule), laying a reliable foundation for subsequent analysis. Simultaneously, daily automatic synchronization and standardized storage of newly added data, combined with a distributed storage architecture using a "customer ID-timestamp" index, achieves efficient data management and rapid retrieval, significantly reducing manual processing costs and improving the timeliness of data updates.

[0073] Secondly, it enhances the accuracy of customer behavior prediction. The spatiotemporal series analysis module, based on the LSTM algorithm, integrates multi-dimensional information such as time interval sequences, quarterly / monthly time characteristics, and purchase amount change rates. Through a deep learning structure with two layers of LSTM units, it can accurately predict the time interval for a customer's next purchase. The monthly incremental model update mechanism, combined with training samples generated by the sliding window method, enables the model to dynamically adapt to changes in customer behavior, further improving prediction accuracy and providing a scientific basis for enterprises to plan inventory and arrange production in advance.

[0074] Secondly, it enables the visualization and refined analysis of regional value. The geographic information association module converts customer geographic coordinates into planar coordinates using Gauss-Kruger projection, divides regions using a density clustering algorithm (5km radius, minimum of 3 samples), and calculates the average value of each region using gross profit margin as the core indicator. Finally, it visually displays this information as a heat map through a GIS system. This process not only clearly presents the distribution of customer value in different regions but also connects and marks related enterprises and the number of joint orders by line segments, helping enterprises to deeply explore the potential of regional markets and optimize resource allocation and market expansion strategies.

[0075] Furthermore, the system optimizes customer management and collaborative decision-making. In the customer management application phase, the system can output the predicted timing of the next three purchases and trigger collaborative procurement alerts under specific conditions (such as the end of a quarter and at least two related companies entering the forecast period). This helps companies proactively connect with customer needs, strengthen collaboration with related companies, and improve customer satisfaction and loyalty. Simultaneously, the data verification and feedback mechanism calculates prediction errors, marks customers with consecutively exceeding errors as "model needs optimization," and adjusts their historical data weights, forming a closed loop of "prediction-verification-optimization" to continuously improve system performance.

[0076] Finally, it possesses excellent scalability and practicality. Each module achieves bidirectional communication through data interfaces, offering a flexible structure that facilitates subsequent functional expansion and system integration. The interactive display module provides a visual user interface, lowering the barrier to entry and enabling enterprise managers to intuitively obtain customer information, forecast results, and regional value analysis, providing convenient and powerful support for business decision-making.

[0077] In summary, this invention, through the organic combination of data processing, predictive analysis, geographic correlation, and application management, comprehensively enhances enterprises' ability to control customer information, the scientific nature of market forecasting, and the accuracy of business decisions, and has extremely high practical application value. Attached Figure Description

[0078] Figure 1 This is a system block diagram of an enterprise customer information management system according to the present invention.

[0079] Figure 2 This is a flowchart illustrating an information management method for enterprise customers according to the present invention. Detailed Implementation

[0080] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0081] Please see Figure 1 and Figure 2 As shown, the embodiments of the present invention provide the following technical solutions:

[0082] As an embodiment of the present invention:

[0083] This invention is an information management system for enterprise customers, comprising: a customer basic information module, a spatiotemporal sequence analysis module, a data storage module, and an interactive display module;

[0084] Each module achieves bidirectional communication through a data interface. The customer basic information module is responsible for collecting and cleaning the original customer data, the spatiotemporal sequence analysis module builds a customer behavior prediction model based on the LSTM algorithm, the data storage module uses a distributed architecture to store structured and unstructured data, and the interactive display module provides users with a visual operation interface.

[0085] The specific implementation method of the customer basic information module is as follows:

[0086] Step 1.1, Data Acquisition:

[0087] Collect static information and dynamic behavioral data of customers;

[0088] Static information includes customer name C n Industry I n Enterprise size S n ;

[0089] Dynamic behavioral data includes historical procurement records P n，tPurchase amount A n，t Order time T n，t Related companies U n，k ;

[0090] Where n is the customer ID, t is the timestamp, and k is the associated company ID;

[0091] Step 1.2, Data Cleaning:

[0092] Step 1.2.1, Missing Value Handling: For records with missing purchase amounts, fill them with the average of the last three purchase amounts from the nth customer, using the following formula:

[0093] Among them, AP n，t For the filled purchase amount, A n，t-i Let i represent the actual purchase amount of the nth customer in the previous i purchases, where i represents the index of the purchase count.

[0094] Step 1.2.2, Outlier Handling: When the amount of a purchase deviates from the historical average of the nth customer by more than 3 standard deviations, it is judged as an outlier and removed.

[0095] The comparison formula is as follows:

[0096] Where μ n，A σ is the average historical purchase amount of customer n. n，A This represents the corresponding standard deviation.

[0097] The specific implementation of the spatiotemporal sequence analysis module is as follows:

[0098] This module constructs a prediction model for the nth customer purchasing cycle using the LSTM algorithm. The specific steps are as follows:

[0099] Step 2.1, Data Sequence Construction:

[0100] Arrange customer n's historical purchase times into a time interval sequence ΔT in chronological order. n =[Δt n，1 ,Δt n，2 ,……,Δt n，m ],in Indicates the time interval between two consecutive purchases;

[0101] Step 2.2, Feature Engineering:

[0102] Extracting time features includes the quarter identifier Q t Its value ranges from 1 to 4, and the month identifier is M. t Its value ranges from 1 to 12;

[0103] Behavioral characteristics include the rate of change between two consecutive purchase amounts:

[0104] Among them, R n，z This represents the rate of change between the purchase amount of the z-th and z-1-th transactions;

[0105] Step 2.3, LSTM model training:

[0106] Input layer: Contains a time interval sequence ΔT n Time characteristics Q t and M t Behavioral characteristics R n，z The input dimension is set to 5;

[0107] Hidden layers: Two LSTM layers are used, with the first layer containing 64 neurons and the second layer containing 32 neurons, employing the tanh activation function.

[0108] Among them, h t W is the hidden state at the current moment. h Let h be the weight matrix. t−1 The hidden state of x in the previous moment t For the current input features, b h This is a bias term.

[0109] Output layer: A linear activation function is used to predict the time interval ΔtP of the next purchase. n，m+1 The loss function uses mean squared error:

[0110] Where j represents the sample index, and the value of j ranges from 1 to m, representing the summation of the prediction error for the j-th sample;

[0111] Meanwhile, the model parameters are iteratively updated using the Adam optimizer until the loss function converges;

[0112] Step 2.4, Procurement Cycle Forecast:

[0113] For customer n, the time intervals and related features of the three most recent purchases are input into the trained model to obtain the predicted time interval ΔtP for the next purchase. n，m+1 The predicted procurement date is:

[0114] ;

[0115] Where m represents the total number of purchases by customer n, and the number of time intervals is m-1;

[0116] This embodiment systematically collects and precisely cleans customer static information and dynamic behavioral data through a customer basic information module, ensuring data quality reliability. The customer behavior prediction model built based on the LSTM algorithm can effectively uncover spatiotemporal patterns in historical customer purchasing data, accurately predicting the next purchasing cycle. This helps enterprises proactively plan inventory management, resource allocation, and customer maintenance, significantly improving the initiative and efficiency of customer relationship management. Simultaneously, the distributed data storage architecture ensures efficient storage and access to massive amounts of structured and unstructured data, while the interactive display module provides users with an intuitive and convenient interface, further enhancing the system's practicality and ease of use.

[0117] As a second embodiment of the present invention:

[0118] In its specific implementation, compared to Embodiment 1, the technical solution of this embodiment differs only in that it further includes a geographic information association module. This module, in conjunction with a GIS system, completes regional value visualization analysis. The specific implementation of the geographic information association module is as follows:

[0119] This module uses a GIS system to perform correlation analysis between customer geographic information and value. The steps are as follows:

[0120] Step 3.1, Geographic Coordinate Mapping:

[0121] Collect latitude and longitude information of customer registration address L n =(Lng n ,Lat n ), where Lng is longitude and Lat is latitude, converted to plane coordinates (X) using Gauss-Kruger projection. n ,Y n The conversion formula is:

[0122] ;

[0123] Where s is a preset scaling factor, and c x c y This is the offset, set according to the parameters of the projection zone.

[0124] Step 3.2, Customer Value Calculation:

[0125] Gross profit margin is used as the core value indicator, and the calculation formula is as follows:

[0126]

[0127] Among them, G n Let C be the average gross profit margin for customer n. n，z This is the cost amount for the customer's zth purchase.

[0128] Step 3.3, Region Cluster Analysis:

[0129] Density clustering algorithm is used to divide customers into regions based on their geographical coordinates;

[0130] The radius is set to 5 kilometers, which is 5000 meters in plane coordinates, and the minimum sample size is set to 3.

[0131] For the customer set C within region r r ={n1,n2,……,n k},pass: Calculate the regional average gross profit margin (GP) r ;

[0132] Step 3.4, GIS Visualization:

[0133] The GP of each region is displayed on the electronic map in the form of a heat map. r Distribution, where darker colors indicate higher gross profit margins in different regions;

[0134] Simultaneously, mark the locations of enterprises that have placed orders with the target customer, and connect the related customers with line segments. The width of the line segment is positively correlated with the number of times they have placed orders together, as shown in the formula:

[0135]

[0136] Among them W n，k N is the line segment width connecting clients n and k. n，k The number of times both parties place orders together is 1-5. When the number of times both parties place orders together is 1-5, the width of the line segment is 0.6-1.0.

[0137] The newly added geographic information association module in this embodiment enables in-depth correlation analysis between customer geographic information and value through a GIS system. It converts customer latitude and longitude into planar coordinates using Gauss-Kruger projection, divides regions using density clustering algorithms, calculates the average gross profit margin for each region using gross profit margin as the core value indicator, and finally visualizes this in the form of a heat map on an electronic map, allowing businesses to intuitively understand the distribution of customer value in different regions. Furthermore, the location marking and connection display of related enterprises help businesses understand the business relationships between customers within a region, providing strong geographic information support for regional market expansion, precision marketing, and supply chain optimization, thereby enhancing the company's control over the regional market.

[0138] As an embodiment of the present invention:

[0139] In specific implementation, compared with Embodiment 1 and Embodiment 2, the technical solution of this embodiment is to combine the solutions of Embodiment 1 and Embodiment 2.

[0140] This embodiment combines the customer behavior prediction function of Embodiment 1 with the geographic information correlation analysis function of Embodiment 2, achieving a comprehensive upgrade of customer information management. On the one hand, it can accurately predict customer purchasing cycles, helping enterprises to better maintain customers and plan resources; on the other hand, through the visual analysis of geographic information, it can clearly understand the distribution of regional customer value and their relationships, enabling enterprises to make more scientific and comprehensive decisions when formulating marketing strategies, expanding markets, and optimizing service networks, based on both individual customer behavior characteristics and regional market rules, thereby maximizing the enterprise's operational efficiency and market competitiveness.

[0141] A method for managing enterprise customer information, implemented through an enterprise customer information management system, includes the following stages:

[0142] Phase 1: Data Acquisition and Preprocessing Phase

[0143] Step S101: Automatically synchronize new data from the customer management system every day at midnight, including the day's purchase records and related company change information.

[0144] Step S102: Call the cleaning function of the customer basic information module to perform missing value filling and outlier removal on the newly added data and generate a standardized data file.

[0145] Step S103: Store the processed data into the data storage module according to the "customer ID-timestamp" index, and retain the mapping relationship between the original data and the cleaned data.

[0146] Phase Two: Model Training and Update Phase

[0147] Step S201: Trigger the model update program of the spatiotemporal sequence analysis module on the 1st of each month to read customer behavior data from the past 12 months;

[0148] Step S202: Reconstruct the time interval sequence and feature matrix, and generate training samples using the sliding window method (window size = 90 days);

[0149] Step S203: Incrementally train the LSTM model using the new samples, adjust the learning rate to 0.001, and save the updated model parameters after 50 iterations;

[0150] Step S204: The geographic information association module updates customer coordinates and regional clustering results quarterly, and synchronously updates the heat map layer of the GIS system;

[0151] Phase Three: Customer Management Application Phase

[0152] Step S301: When a user queries customer n, the system automatically calls the prediction model and outputs the predicted time TP for the next 3 purchases. n，m+1 TP n，m+2 TP n，m+3 ;

[0153] Step S302: Display a heat map of the location distribution of its affiliated companies on the customer details page, and mark the average gross profit margin and the number of joint orders in each region;

[0154] Step S303: When it is predicted that a customer's purchase time is at the end of the quarter, i.e., Q t =4, and when more than 2 related companies enter the forecast period at the same time, the early warning mechanism is triggered and a collaborative procurement prompt is generated.

[0155] Phase Four: Data Verification and Feedback Phase

[0156] Step S401: After each actual purchase, the system automatically calculates the error between the predicted time and the actual time. ;

[0157] Step S402: When the error exceeds 7 days for 3 consecutive times, mark the customer as "model to be optimized" and increase the weight of its historical data in the next model update.

[0158] Through the above implementation methods, the present invention realizes dynamic management and spatiotemporal dimension analysis of customer information, which can accurately predict the procurement cycle and intuitively display the regional customer value distribution, thus solving the limitations of static recording in existing systems.

[0159] It should be stated that all user data collected in this application was collected with the user's consent and authorization, and the use of user data is legal and compliant, and the use and processing of user data comply with the relevant laws, regulations and standards of the relevant regions.

[0160] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0161] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0162] The above formulas are all dimensionless calculations. Dimensionless calculation involves introducing a reference benchmark, such as the maximum, minimum, standard deviation, or theoretical extreme value of a physical quantity, to transform the original physical quantity into a dimensionless relative value. This value is usually mapped to a specific interval, such as [0,1] or [-1,1], which eliminates the influence of units while preserving the relative size relationship of the physical quantities. The formula is derived from software simulation based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0163] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0164] The various embodiments of this application have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. An information management system for enterprise customers, characterized in that, include: The customer basic information module is used to collect raw customer data, including static information and dynamic behavioral data; Static information includes customer name C n Industry I n Enterprise size S n Dynamic behavioral data includes historical procurement records (P). n，t Purchase amount A n，t Order time T n，t Related companies U n，k Where n is the customer ID, t is the timestamp, and k is the associated company ID; simultaneously, the original customer data undergoes data cleaning, which includes: Missing value handling: When the purchase amount is missing, it is filled with the average of the last 3 purchase amounts of the nth customer; Outlier handling: Purchases exceeding three standard deviations from the historical average of the nth customer are excluded. The spatiotemporal sequence analysis module includes a data sequence construction unit, a feature engineering unit, and an LSTM model training unit; The data sequence construction unit organizes the historical purchase times of customer n into a time interval sequence ΔT in chronological order. n =[Δt n，1 ,Δt n，2 ,……,Δt n，m ],in, This represents the time interval between two consecutive purchases, where m represents the total number of purchases by customer n, and i represents the index of the purchase count. The features extracted by the feature engineering unit include: Time characteristic: Quarter identifier Q t Its value ranges from 1 to 4, and the month identifier is M. t Its value ranges from 1 to 12; Behavioral characteristics: through Calculate the rate of change R between the z-th and z-1th consecutive purchase amounts. n，z ; The structure of the LSTM model training unit is as follows: Input layer: Contains a time interval sequence ΔT n Time characteristics (Q) t M t ), behavioral characteristics R n，z The input dimension is 5; Hidden layers: 2 LSTM layers, the first layer contains 64 neurons and the second layer contains 32 neurons, using the tanh activation function. Among them, h t W is the hidden state at the current moment. h Let h be the weight matrix. t−1 The hidden state of x in the previous moment t For the current input features, b h For bias terms; Output layer: A linear activation function is used to predict the time interval ΔtP of the next purchase. n，m+1 The loss function uses mean squared error; The geographic information association module includes a geographic coordinate mapping unit, a customer value calculation unit, a regional clustering analysis unit, and a GIS visualization unit; The geographic coordinate mapping unit maps the latitude and longitude (L and L) of the customer's registered address. n =(Lng n ,Lat n ); Converted to planar coordinates using Gauss-Kruger projection (X) n ,Y n ); The formula is: Where s is a preset scaling factor, and c x c y This is the offset; The customer value calculation unit through Calculate the average gross profit margin G for customer n. n As a core value indicator, C n，z This is the cost amount for the customer's zth purchase. The regional clustering analysis unit uses a density clustering algorithm to divide the customer into regions according to geographical coordinates, with a radius of 5 kilometers (5000 meters in planar coordinates) and a minimum sample size of 3. For the customer set C within region r r ={n1,n2,……,n k },pass: Calculate the regional average gross profit margin (GP) r ; The GIS visualization unit displays the GP of each region on the electronic map in the form of a heat map. r distributed; Simultaneously, mark the locations of enterprises that have placed orders with the target customer, and connect the related customers with line segments. The width of the line segment is positively correlated with the number of times they have placed orders together, as shown in the formula: Among them, W n，k N is the width of the line segment connecting clients n and k. n，k This refers to the number of times both parties place orders together. The data storage module adopts a distributed architecture to store structured and unstructured data; The interactive display module provides users with a visual operation interface.

2. A method for managing enterprise customer information, wherein the method is implemented through an enterprise customer information management system as described in any one of claims 1, characterized in that, This method Includes the following stages: Data collection and preprocessing stage: New data is added daily, cleaned, and then stored in a standardized manner; Model training and update phase: The LSTM model is updated monthly, and the geographic information clustering results are updated quarterly; Customer management application phase: Based on the model output, procurement forecast time and regional value distribution are generated, triggering collaborative procurement early warning; Data validation and feedback phase: Calculate prediction error and dynamically optimize model weights.

3. The enterprise customer information management method according to claim 2, characterized in that: The data acquisition and preprocessing stage includes: Step S101: Automatically synchronize new data from the customer management system every day at midnight, including the day's purchase records and related company change information; Step S102: Call the cleaning function of the customer basic information module to perform missing value filling and outlier removal, and generate a standardized data file; Step S103: Store the processed data into the data storage module according to the "Customer ID-Timestamp" index.

4. The enterprise customer information management method according to claim 2, characterized in that: The model training and update phase includes: Step S201: Trigger the model update program of the spatiotemporal sequence analysis module on the 1st of each month to read customer behavior data from the past 12 months; Step S202: Use the sliding window method to generate training samples and reconstruct the time interval sequence and feature matrix; Step S203: Incrementally train the LSTM model with a learning rate of 0.001 and save the updated model parameters after 50 iterations. Step S204: The geographic information association module updates customer coordinates and regional clustering results quarterly, and updates the GIS heat map layer synchronously.

5. The enterprise customer information management method according to claim 2, characterized in that: The customer management application phase includes: Step S301: When a user queries customer n, the system outputs the predicted time TP for the next 3 purchases. n，m+1 TP n，m+2 TP n，m+3 ,in ; Step S302: Display a heat map showing the location distribution of related companies on the customer details page, and mark the average gross profit margin of the region and the number of times they place orders together; Step S303: When the customer predicts the procurement time is at the end of the quarter, i.e., Q t When the number of related enterprises is 4 and at least 2 of them are included in the forecast period, the early warning mechanism will be triggered.

6. The enterprise customer information management method according to claim 2, characterized in that: The data verification and feedback phase includes: Step S401: After the actual purchase occurs, calculate the error between the predicted time and the actual time: ; Step S402: When the error exceeds 7 days for 3 consecutive times, mark the customer as "model to be optimized".

Images