Enterprise feature portrait system based on artificial intelligence
Through the enterprise feature portrait system based on artificial intelligence, the problem of data update lag is solved, real-time data update and system flexibility are realized, timeliness and adaptability of decisions are improved, and the enterprise needs for real-time analysis are met.
Patent Information
- Application Number
- CN202410958063.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-17
- Publication Date
- 2024-11-15
AI Technical Summary
Traditional enterprise feature portrait systems have a lag in data processing, which is difficult to reflect the latest changes in the company in real time, affecting the timeliness and accuracy of decisions, and lacking flexibility in design, making it difficult to quickly adapt to changes in market and enterprise needs.
Adopting an enterprise feature portrait system based on artificial intelligence, through data acquisition, preprocessing, splitting, modular storage and dynamic update mechanisms, combined with particle swarm optimization and genetic algorithms, real-time data update and flexible model adjustment, and supporting hot plugging Unplugged and modular design ensures real-time data and rapid system expansion.
Real-time update of data is realized, timeliness and accuracy of decision-making is improved, can quickly adapt to market changes, meet the company's needs for real-time analysis, and improve the flexibility of the system and data processing efficiency.
Smart Images

Figure CN118965059A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of enterprise feature portrait systems, and specifically to an enterprise feature portrait system based on artificial intelligence. Background Art
[0002] The enterprise feature profiling system is a system that uses big data and machine learning technology to conduct in-depth analysis of internal and external data of the enterprise, thereby building a comprehensive and accurate enterprise information archive. This system provides enterprises with a detailed and multi-dimensional operation and management view by integrating and analyzing the production, operation, management, market and other multi-faceted information of the enterprise.
[0003] In this regard, the patent document with publication number CN117522146A discloses a method and system for constructing enterprise risk portraits and risk warnings, wherein the enterprise risk portrait construction and risk warning method includes: obtaining enterprise data from a dynamic data source to generate first aggregated data; screening the first aggregated data to generate second aggregated data; generating risk characteristics and general characteristics based on the second aggregated data to construct risk labels; constructing an enterprise risk portrait model based on the risk characteristics and the general characteristics to obtain an enterprise risk portrait score; constructing an enterprise risk prediction model based on the enterprise risk portrait score and the risk label to generate a prediction result. In this way, the enterprise risk portrait construction and risk warning method can better identify the risks faced by the enterprise and provide real-time risk monitoring and warning.
[0004] In this regard, the patent document with publication number CN117575148A discloses a method, system and terminal for calculating enterprise portraits based on big data and multi-dimensional features. The method for calculating enterprise portraits based on big data and multi-dimensional features includes the following steps: obtaining the corresponding multi-dimensional feature data of the enterprise according to the part that the user is interested in; calculating the enterprise portrait through the obtained multi-dimensional feature data. The enterprise portrait calculation system includes: a multi-dimensional feature data acquisition device and an enterprise portrait calculation device. The terminal includes a memory and a processor. It has the advantages of being able to provide users with personalized enterprise portraits in a targeted manner, effectively improving the efficiency of users making scientific and accurate decisions based on the enterprise's big data, and enabling users to obtain the required enterprise portraits with extremely high efficiency to meet relevant needs.
[0005] Traditional enterprise feature profiling systems often have the problem of delayed updates in data processing, making it difficult to reflect the latest changes in the enterprise in real time, which to some extent affects the timeliness and accuracy of decision-making. Many existing systems lack sufficient flexibility in design and are difficult to quickly adapt to changes in market and enterprise needs. When it is necessary to add new data sources or adjust existing models, the entire system often needs to be restructured on a large scale.
[0006] In view of the above problems, this paper proposes an enterprise feature profiling system based on artificial intelligence. Summary of the invention
[0007] The purpose of the present invention is to provide an enterprise feature profiling system based on artificial intelligence, which solves the problem of low efficiency in data lagging processing in the background technology.
[0008] To achieve the above purpose, the present invention provides the following technical solution: an enterprise feature profiling system based on artificial intelligence, comprising:
[0009] Data collection: Data collection accesses multiple data sources, including internal enterprise databases, external data interfaces, and social media data;
[0010] Data preprocessing, data source cleaning and verification, removal of duplicate data, correction of erroneous data, and filling of missing values;
[0011] Data type splitting: splitting data into data blocks according to data type;
[0012] Data attribute splitting: further splitting data blocks according to their attributes to form data slices;
[0013] Modular data storage, classified data blocks are stored in different modules;
[0014] Dynamic data update, dynamic data update adds new data blocks, deletes obsolete data blocks, and updates data slices in data blocks, data management;
[0015] Model establishment, build the main framework, the main framework is the core structure of the model, logical operations extract new main frameworks, and operate independently. The main framework can be copied unlimitedly. The main framework templates are divided into multiple categories, including data processing, data analysis, and machine learning models;
[0016] The template contains basic structure definition, data interface, processing flow and logical operation steps;
[0017] The structure definition includes the organization of data and storage location;
[0018] The data interface defines the standards for data input and output;
[0019] The processing flow describes the steps for processing data;
[0020] The logical operation steps describe the specific operation;
[0021] Create a template library to store and manage all main framework templates, and create an index for each template in the template library to support fast retrieval and selection;
[0022] Use the template dynamic generation algorithm to select the most suitable template from the template library;
[0023] Parameter optimization, used to optimize template parameters, the main steps include selection, crossover, mutation and fitness evaluation;
[0024] Particle swarm optimization, which finds the optimal parameter combination by simulating the movement and evolution of particle swarms;
[0025] The fitness function of the genetic algorithm is:
[0026] Fit(p)=f(p)
[0027] Among them, P is the parameter set, f(p) is the fitness evaluation function;
[0028] Particle swarm optimization formula:
[0029] v i (t+1)=w×v i (t)+c1×r1×(p best , i -x i (t))+c2×r2×(g best -
[0030] x i (t))x i (t+1)=x i (t)+v i (t+1);
[0031] v i (t) is the velocity of ion i at time t;
[0032] x i (t) is the position of particle i at time t;
[0033] p best , i is the best historical position of particle i;
[0034] g best is the global optimal position;
[0035] w represents the inertia weight, which controls the inertia of the particle moving forward;
[0036] c1 and c2 are learning factors, representing the ion learning optimal position and the global optimal position, respectively;
[0037] r1 and r2 are random numbers in the range [0,1];
[0038] Position update formula:
[0039] x i (t+1)=xi (t)+v i (t+1);
[0040] x i (t+1): the position of particle i at time t+1;
[0041] v i (t+1): velocity of particle i at time t+1;
[0042] In the PSO algorithm, particles move in space by their velocity and position. The velocity update formula determines the particle movement, and the position update formula updates the new position of the particle.
[0043] The inertia weight w controls the forward inertia of the particle. When w is large, the particle maintains the current speed direction and searches globally. When w is small, the particle searches locally.
[0044] The learning factors c1 and c2 determine the degree to which the particle approaches its own historical best position and the global best position when updating the speed;
[0045] Random factors r1 and r2 increase search randomness;
[0046] Model update: build models between data blocks and the main framework through hot plugging;
[0047] During data analysis and logical operations, relevant data blocks are selected for operation, and data pieces that are not relevant to the operation are deleted or disabled;
[0048] System maintenance, establish a system monitoring mechanism, monitor the system's operating status and performance indicators in real time, detect abnormal situations, and trigger alarms.
[0049] Preferably, input a data set D, traverse the data set, identify the data type, and use a classification algorithm to split the data into data blocks B by type:
[0050] B t ={x∈D|type(x)=t}.
[0051] Preferably, the data type is quickly identified through predefined rules, and the classification model is trained to automatically identify the data type. Given an input x and a type t, the mathematical formula for calculating the maximum value of the probability P(t|x) can be expressed as:
[0052] type
[0053] x is a data sample; T is the set of all possible data types; p(t|x) is the type of data x, which is achieved by selecting the type t with the maximum probability;
[0054] Use multi-threaded or distributed computing frameworks to process large-scale data in parallel to increase classification speed. Use caches for common data types and patterns to speed up the classification process.
[0055] Preferably, an index structure is established to accelerate the identification and query of data attributes, and a parallel computing framework is used to process multiple data blocks simultaneously to improve the splitting efficiency and divide the data partitioning strategy to avoid excessive concentration of data on a certain node;
[0056] Input data block B, traverse the data block, identify data attributes, and further split it into data pieces P according to the attributes:
[0057] p ' ={x∈B|attribute(x)=a}.
[0058] Preferably, data integrity check:
[0059] Input the original data D and the split data piece P, and calculate the checksum of the original data:
[0060] Checksum
[0061] Calculate the checksum of the split data piece:
[0062] Checksum
[0063] Compare checksums to ensure data integrity.
[0064] Preferably, in data slice management, irrelevant or outdated data slices are cleaned up regularly to optimize storage and computing resources. The steps are as follows:
[0065] Establish a correlation model and calculate the correlation score of the data slice. The calculation formula is as follows:
[0066]
[0067] where w i is the weight of the i-th feature, f i is the i-th feature function, and n is the number of features.
[0068] Preferably, a threshold of the relevance score is set according to business requirements, and data slices below this threshold will be marked as irrelevant, and data slices are scanned regularly and deleted.
[0069] Preferably, to improve computing efficiency, the following strategies are adopted: sharding the data, establishing an efficient index structure to speed up queries, using a parallel computing framework to process large-scale data slices, distributed computing to increase processing speed, and using cache to store frequently accessed data slices to reduce I / O operations.
[0070] Preferably, by introducing a machine learning algorithm, the intelligent level of data slice management is improved, the data slices are cleaned, standardized and feature engineered, a training data set is generated, a machine learning model is trained, and a clustering model is used to group similar data slices. The calculation formula is:
[0071]
[0072] K is the number of clusters, n i is the number of data pieces in cluster i, n i is the center of the i clusters.
[0073] Preferably, data is collected from different channels, formatted using standardized data interfaces, and stored in a unified database. Mapping rules are used to convert data into a unified data model structure. Consistency verification scripts are run regularly, and hash functions and checksums are used to check data consistency. Model verification functions are used to check the structural integrity of the data model. For inconsistent or incomplete data detected, machine learning algorithms are used to repair them, and verification reports are generated to record inconsistent data and repair status.
[0074] Compared with the prior art, the present invention has the following beneficial effects:
[0075] 1. The enterprise feature portrait system based on artificial intelligence provided by the present invention can ensure that the data in the enterprise feature portrait system is updated in real time and accurately reflects the latest status of the enterprise by adopting advanced data synchronization technology and real-time data processing algorithms. This not only helps to improve the timeliness of decision-making, but also can timely discover and respond to market changes.
[0076] 2. The present invention provides an enterprise feature profiling system based on artificial intelligence. The modularity and hot-swappable capabilities of the system enable the system to be quickly adjusted and expanded according to different business needs. Whether adding new data sources or adjusting analysis models, it can be easily completed without affecting the operation of the existing system. Through intelligent splitting and classification of data, combined with efficient data processing algorithms, my design concept significantly improves the efficiency of data processing. This enables the system to process large amounts of data in a short period of time and meet the needs of enterprises for real-time analysis. BRIEF DESCRIPTION OF THE DRAWINGS
[0077] Figure 1 is a system flow chart of the present invention;
[0078] Figure 2 This is a flow chart of the system model of the present invention;
[0079] Figure 3 This is a schematic diagram of the model flow chart for constructing the main framework of the present invention. DETAILED DESCRIPTION
[0080] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0081] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings.
[0082] Combination Figure 1-Figure 3 , an enterprise feature profiling system based on artificial intelligence of the present invention comprises:
[0083] Data collection: Data collection accesses multiple data sources, including internal enterprise databases, external data interfaces, and social media data;
[0084] Data preprocessing, data source cleaning and verification, removal of duplicate data, correction of erroneous data, and filling of missing values;
[0085] Data type splitting: splitting data into data blocks according to data type;
[0086] Input data set D, traverse the data set, identify the data type, and use the classification algorithm to split the data into data blocks B by type:
[0087] B t ={x∈D|type(x)=t};
[0088] The data type can be quickly identified through predefined rules, and the classification model can be trained to automatically identify the data type. Given the input x and type t, the mathematical formula for calculating the maximum value of the probability P(t|x) can be expressed as:
[0089] type
[0090] x is a data sample; T is the set of all possible data types; p(t|x) is the type of data x, which is achieved by selecting the type t with the maximum probability;
[0091] Use multi-threaded or distributed computing frameworks to process large-scale data in parallel to improve classification speed. Use caches for common data types and patterns to speed up the classification process.
[0092] Data attribute splitting: further splitting data blocks according to their attributes to form data slices;
[0093] Establish an index structure to speed up the identification and query of data attributes, use a parallel computing framework to process multiple data blocks simultaneously, improve splitting efficiency, and divide data partitioning strategies to avoid excessive concentration of data on a certain node;
[0094] Input data block B, traverse the data block, identify data attributes, and further split it into data pieces P according to the attributes:
[0095] p ' ={x∈B|attribute(x)=a};
[0096] Secondly, check the data integrity:
[0097] Input the original data D and the split data piece P, and calculate the checksum of the original data:
[0098] Checksum
[0099] Calculate the checksum of the split data piece:
[0100] Checksum
[0101] Compare checksums to ensure data integrity;
[0102] Modular data storage, classified data blocks are stored in different modules;
[0103] Dynamic data update, dynamic data update adds new data blocks, deletes obsolete data blocks, and updates data slices in data blocks, data management;
[0104] Model establishment, build the main framework, the main framework is the core structure of the model, logical operations extract new main frameworks, and operate independently. The main framework can be copied unlimitedly. The main framework templates are divided into multiple categories, including data processing, data analysis, and machine learning models;
[0105] The template contains basic structure definition, data interface, processing flow and logical operation steps;
[0106] The structure definition includes the organization of data and storage location;
[0107] The data interface defines the standards for data input and output;
[0108] The processing flow describes the steps for processing data;
[0109] The logical operation steps describe the specific operation;
[0110] Create a template library to store and manage all main framework templates, and create an index for each template in the template library to support fast retrieval and selection;
[0111] Use the template dynamic generation algorithm to select the most suitable template from the template library;
[0112] Parameter optimization, used to optimize template parameters, the main steps include selection, crossover, mutation and fitness evaluation;
[0113] Particle swarm optimization, which finds the optimal parameter combination by simulating the movement and evolution of particle swarms;
[0114] The fitness function of the genetic algorithm is:
[0115] Fit(p)=f(p)
[0116] Among them, P is the parameter set, f(p) is the fitness evaluation function;
[0117] Particle swarm optimization formula:
[0118] v i (t+1)=w×v i (t)+c1×r1×(p best , i -x i (t))+c2×r2×(g best -
[0119] x i (t))x i (t+1)=x i (t)+v i (t+1);
[0120] v i (t) is the velocity of ion i at time t;
[0121] x i (t) is the position of particle i at time t;
[0122] p best , i is the best historical position of particle i;
[0123] g best is the global optimal position;
[0124] w represents the inertia weight, which controls the inertia of the particle moving forward;
[0125] c1 and c2 are learning factors, representing the ion learning optimal position and the global optimal position, respectively;
[0126] r1 and r2 are random numbers in the range [0,1];
[0127] First, define the template description language to standardize the description of template structure, interface and processing flow. Then, establish a template library to store all templates and implement version control and indexing mechanisms. The template selection algorithm intelligently selects the most suitable template based on requirements and template features. Implement a template dynamic generation tool, use an optimization algorithm to dynamically generate main framework instances, and introduce an intelligent recommendation system to analyze historical operation records through machine learning to recommend the most suitable template.
[0128] By creating a flexible main frame template library and combining it with intelligent selection and dynamic generation algorithms, the flexibility and adaptability of the system can be greatly improved. This design not only ensures that the most appropriate main frame is used for each logical operation, improves the efficiency of data processing and operation, but also maintains the consistency and integrity of the data model.
[0129] Position update formula:
[0130] x i (t+1)=x i (t)+v i (t+1);
[0131] x i (t+1): the position of particle i at time t+1;
[0132] v i (t+1): velocity of particle i at time t+1;
[0133] Step 1: Initialize the particle velocity and position, as well as the global optimal position g best and the best historical position p of each particle best , i ; Step 2, calculate the fitness value of each particle. If the fitness value of the current particle is better than its historical best position, update p best , i If the fitness value of the current particle is better than the global optimal position, update g best , step 3, according to the speed update formula and position update formula, update the speed and position of each particle, repeat the above steps until the termination condition is met (such as reaching the maximum number of iterations or the fitness value reaches a predetermined threshold); first, you need to find the function: f(x) = x 1 The minimum value of the particle's initial position and velocity is defined, and the particle velocity v is initialized. i (0)x i (0) Random;
[0134] In each iteration, the fitness value f((t)) of the particle is calculated and p is updated best , i and g best; Use the velocity update formula and position update formula to update the particle's velocity and position.
[0135] In the PSO algorithm, particles move in space by their velocity and position. The velocity update formula determines the particle movement, and the position update formula updates the new position of the particle.
[0136] The inertia weight w controls the forward inertia of the particle. When w is large, the particle maintains the current speed direction and searches globally. When w is small, the particle searches locally.
[0137] The learning factors c1 and c2 determine the degree to which the particle approaches its own historical best position and the global best position when updating the speed;
[0138] When particle i, at time t, has a velocity v t , position x i (t), the best position is p best , i , the best position is g best , at the next time t+1, the velocity and position of the particle will be updated:
[0139] Speed update: v i (t+1) is the current speed v i The adjustment result of (t) combines the particle's adjustment to its historical best position p best , i and the global optimal position g best trend and add randomness; the new position x i (t+1) is the current speed v i (t+1) result.
[0140] Random factors r1 and r2 increase search randomness;
[0141] Model update: build models between data blocks and the main framework through hot plugging;
[0142] During data analysis and logical operations, relevant data blocks are selected for operation, and data pieces that are not relevant to the operation are deleted or disabled;
[0143] In data slice management, irrelevant or outdated data slices are cleaned up regularly to optimize storage and computing resources. The steps are as follows:
[0144] Establish a correlation model and calculate the correlation score of the data slice. The calculation formula is as follows:
[0145]
[0146] where w i is the weight of the i-th feature, f iis the i-th feature function, n is the number of features;
[0147] Set a relevance score threshold based on business needs. Data slices below this threshold will be marked as irrelevant. Scan data slices regularly and delete irrelevant data slices.
[0148] To improve computational efficiency, the following strategies can be adopted:
[0149] Shard the data and build an efficient index structure to speed up queries. Use a parallel computing framework to process large-scale data slices. Distributed computing improves processing speed. Use cache to store frequently accessed data slices to reduce I / O operations.
[0150] By introducing machine learning algorithms, the intelligent level of data slice management is improved, data slices are cleaned, standardized and feature engineered, training data sets are generated, machine learning models are trained, and clustering models are used to group similar data slices. The calculation formula is:
[0151]
[0152] K is the number of clusters, n i is the number of data pieces in cluster i, n i is the center of the i clusters;
[0153] Collect data from different channels, format it using standardized data interfaces, store the formatted data in a unified database, use mapping rules to convert data into a unified data model structure, run consistency verification scripts regularly, use hash functions and checksums to check data consistency, use model verification functions to check the structural integrity of the data model, use machine learning algorithms to repair detected inconsistent or incomplete data, generate verification reports, and record inconsistent data and repair situations;
[0154] System maintenance, establish a system monitoring mechanism, monitor the system's operating status and performance indicators in real time, detect abnormal situations, and trigger alarms.
[0155] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0156] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. An enterprise feature profiling system based on artificial intelligence, characterized by: include, Data collection: Data collection accesses multiple data sources, including internal enterprise databases, external data interfaces, and social media data; Data preprocessing, data source cleaning and verification, removal of duplicate data, correction of erroneous data, and filling of missing values; Data type splitting: splitting data into data blocks according to data type; Data attribute splitting: further splitting data blocks according to their attributes to form data slices; Modular data storage, classified data blocks are stored in different modules; Dynamic data update, dynamic data update adds new data blocks, deletes obsolete data blocks, and updates data slices in data blocks, data management; Model establishment, build the main framework, the main framework is the core structure of the model, logical operations extract new main frameworks, and operate independently. The main framework can be copied unlimitedly. The main framework templates are divided into multiple categories, including data processing, data analysis, and machine learning models; The template contains basic structure definition, data interface, processing flow and logical operation steps; The structure definition includes the organization of data and storage location; The data interface defines the standards for data input and output; The processing flow describes the steps for processing data; The logical operation steps describe the specific operation; Create a template library to store and manage all main framework templates, and create an index for each template in the template library to support fast retrieval and selection; Use the template dynamic generation algorithm to select the most suitable template from the template library; Parameter optimization, used to optimize template parameters, the main steps include selection, crossover, mutation and fitness evaluation; Particle swarm optimization, which finds the optimal parameter combination by simulating the movement and evolution of particle swarms; The fitness function of the genetic algorithm is: Fit(p)=f(p) Among them, P is the parameter set, f(p) is the fitness evaluation function; Particle swarm optimization formula: v i (t+1)=w×v i (t)+c1×r1×(p best , i -x i (t))+c2×r2×(g best - x i (t))x i (t+1)=x i (t)+v i (t+1); v i (t) is the velocity of ion i at time t; x i (t) is the position of particle i at time t; p best , i is the best historical position of particle i; g best is the global optimal position; w represents the inertia weight, which controls the inertia of the particle moving forward; c1 and c2 are learning factors, representing the ion learning optimal position and the global optimal position, respectively; r1 and r2 are random numbers in the range [0,1]; Position update formula: x i (t+1)=x i (t)+v i (t+1); x i (t+1): the position of particle i at time t+1; v i (t+1): velocity of particle i at time t+1; In the PSO algorithm, particles move in space by their velocity and position. The velocity update formula determines the particle movement, and the position update formula updates the new position of the particle. The inertia weight w controls the forward inertia of the particle. When w is large, the particle maintains the current speed direction and searches globally. When w is small, the particle searches locally. The learning factors c1 and c2 determine the degree to which the particle approaches its own historical best position and the global best position when updating the speed; Random factors r1 and r2 increase search randomness; Model update: build models between data blocks and the main framework through hot plugging; During data analysis and logical operations, relevant data blocks are selected for operation, and data pieces that are not relevant to the operation are deleted or disabled; System maintenance, establish a system monitoring mechanism, monitor the system's operating status and performance indicators in real time, detect abnormal situations, and trigger alarms.
2. The enterprise feature profiling system based on artificial intelligence according to claim 1 is characterized by: Input data set D, traverse the data set, identify the data type, and use the classification algorithm to split the data into data blocks B by type: B t ={x∈D|type(x)=t}.
3. The enterprise feature profiling system based on artificial intelligence according to claim 2 is characterized by: The data type can be quickly identified through predefined rules, and the classification model can be trained to automatically identify the data type. Given the input x and type t, the mathematical formula for calculating the maximum value of the probability P(t|x) can be expressed as: x is a data sample; T is the set of all possible data types; p(t|x) is the type of data x, which is achieved by selecting the type t with the maximum probability; Use multi-threaded or distributed computing frameworks to process large-scale data in parallel to increase classification speed. Use caches for common data types and patterns to speed up the classification process.
4. The enterprise feature profiling system based on artificial intelligence according to claim 1 is characterized by: Establish an index structure to speed up the identification and query of data attributes, use a parallel computing framework to process multiple data blocks simultaneously, improve splitting efficiency, and divide data partitioning strategies to avoid excessive concentration of data on a certain node; Input data block B, traverse the data block, identify data attributes, and further split it into data pieces P according to the attributes: p ' ={x∈B|attribute(x)=a}.
5. The enterprise feature profiling system based on artificial intelligence according to claim 4 is characterized by: Data integrity check: Input the original data D and the split data piece P, and calculate the checksum of the original data: Calculate the checksum of the split data piece: Compare checksums to ensure data integrity.
6. The enterprise feature profiling system based on artificial intelligence according to claim 1 is characterized by: In data slice management, irrelevant or outdated data slices are cleaned up regularly to optimize storage and computing resources. The steps are as follows: Establish a correlation model and calculate the correlation score of the data slice. The calculation formula is as follows: where w i is the weight of the i-th feature, f i is the i-th feature function, and n is the number of features.
7. The enterprise feature profiling system based on artificial intelligence according to claim 1 is characterized by: Set a threshold for the relevance score based on business needs. Data slices below this threshold will be marked as irrelevant. Scan the data slices regularly and delete irrelevant data slices.
8. The enterprise feature profiling system based on artificial intelligence according to claim 7 is characterized by: To improve computing efficiency, the following strategies are adopted: sharding data, establishing efficient index structures to speed up queries, using parallel computing frameworks to process large-scale data slices, distributed computing to increase processing speed, and using cache to store frequently accessed data slices to reduce I / O operations.
9. The enterprise feature profiling system based on artificial intelligence according to claim 8 is characterized by: By introducing machine learning algorithms, the intelligent level of data slice management is improved, data slices are cleaned, standardized and feature engineered, training data sets are generated, machine learning models are trained, and clustering models are used to group similar data slices. The calculation formula is: K is the number of clusters, n i is the number of data pieces in cluster i, n i is the center of the i clusters.
10. The enterprise feature profiling system based on artificial intelligence according to claim 9, characterized in that: Collect data from different channels, format it using standardized data interfaces, store the formatted data in a unified database, use mapping rules to convert the data into a unified data model structure, run consistency verification scripts regularly, use hash functions and checksums to check data consistency, use model verification functions to check the structural integrity of the data model, use machine learning algorithms to repair detected inconsistent or incomplete data, generate verification reports, and record inconsistent data and repair situations.
Citation Information
Patent Citations
Visualization system based on hot-pluggable dynamic component
CN112256260A
Rule engine-based rule event dynamic loading and updating method and related equipment
CN115129736A
Intelligent enterprise portrait analysis system and method based on unified social credit code
CN115827934A
Enterprise risk level prediction method and system based on deep learning
CN117391247A
Enterprise portrait system and method
CN118134078A
