Digital precise drainage sale management method and system fused with knowledge graph

Abstract

The invention relates to the technical field of digital precise drainage sale management methods, in particular to a digital precise drainage sale management method and system fused with a knowledge graph, and the method comprises the following steps: obtaining and preprocessing multi-source heterogeneous data; multi-modal data fusion and entity identification are realized based on the preprocessed data; performing relation extraction and attribute mapping according to the fused data and the identified entities; constructing dynamic time sequence diagram embedding based on the extracted relationship and the mapped attributes; performing knowledge reasoning and graph completion by utilizing dynamic time sequence graph embedding; performing real-time updating and incremental learning on the completed knowledge graph; generating a precise drainage strategy based on the updated knowledge graph; optimizing the sale-on-sale process according to the accurate drainage strategy; according to the method, effect evaluation and feedback are carried out on the optimized sale-on-sale process, and a feedback result is used for each step of the optimization method, so that dynamic time sequence diagram embedding is realized, user interest and market trend change can be captured in real time, and recommendation timeliness is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital precision traffic diversion and sales agency management methods, and more specifically, to a digital precision traffic diversion and sales agency management method and system integrating knowledge graphs. Background Art

[0002] With the rapid development of e-commerce, digital marketing and precise traffic generation have become one of the core competitiveness of major e-commerce platforms. In recent years, recommendation systems and data analysis technologies have made significant progress in this field, but they still face many challenges.

[0003] Traditional recommendation systems mainly rely on collaborative filtering algorithms, such as user-based collaborative filtering and item-based collaborative filtering. Although these methods are simple and effective, they have limitations such as cold start problems, data sparsity problems, and difficulty in capturing dynamic changes in user interests. For example, for new users or new products, traditional collaborative filtering algorithms often cannot provide effective recommendations due to the lack of historical interaction data.

[0004] To solve these problems, researchers have introduced content-based recommendation methods and hybrid recommendation methods. These methods improve the quality of recommendations by analyzing product features and user profiles. However, these methods can usually only process structured data and are unable to fully utilize the rich information contained in a large amount of unstructured data (such as user reviews, product descriptions, etc.).

[0005] In recent years, the application of deep learning technology has brought new breakthroughs to recommendation systems. For example, collaborative filtering methods based on deep neural networks can automatically learn feature representations and improve the accuracy of recommendations. However, these methods still focus mainly on direct user-item interactions and have difficulty modeling complex multi-entity relationships.

[0006] The introduction of knowledge graph technology provides a new approach to solving the above problems. By representing entities such as users, products, attributes and their relationships as graph structures, knowledge graphs can better capture complex semantic relationships. However, existing recommendation methods based on knowledge graphs still have the following problems:

[0007] 1. Difficulty in integrating multi-source heterogeneous data: Existing methods often find it difficult to effectively integrate heterogeneous data from different sources, such as user behavior data, product attribute data, social network data, etc.

[0008] 2. Insufficient dynamic modeling: Most methods construct static knowledge graphs, which are difficult to reflect the dynamic changes of user interests and market trends in a timely manner.

[0009] 3. Limited reasoning ability: Existing methods mainly rely on simple graph embedding or path mining techniques, lacking the ability to deeply reason about complex semantic relationships.

[0010] 4. Insufficient explainability: Many deep learning-based methods are "black box" models, which make it difficult to provide clear explanations for recommendation results.

[0011] 5. Long-tail problem: Existing methods tend to recommend popular products, but fail to explore and utilize long-tail products.

[0012] 6. Insufficient real-time and scalability: Faced with the massive data and high concurrent requests of large-scale e-commerce platforms, many methods are difficult to meet the requirements of real-time and scalability.

[0013] These problems severely limit the application effect of existing technologies in e-commerce precision marketing and agency sales management. Therefore, a new method that can comprehensively solve the above problems is urgently needed to improve the effect of digital precision traffic diversion and agency sales management. Summary of the invention

[0014] In view of the above-mentioned problems existing in the prior art, the present invention proposes a digital precise traffic diversion and sales management method and system integrating knowledge graph. This method constructs a comprehensive, dynamic and intelligent knowledge graph system by innovatively combining advanced technologies such as quantum computing ideas, multimodal data fusion, dynamic time series graph embedding, and multi-granularity causal reasoning.

[0015] The present invention provides a digital precise traffic diversion and agency sales management method integrating knowledge graphs, comprising the following steps: acquiring and preprocessing multi-source heterogeneous data; realizing multimodal data fusion and entity recognition based on the preprocessed data; performing relationship extraction and attribute mapping based on the fused data and the recognized entities; constructing a dynamic time series graph embedding based on the extracted relationships and the mapped attributes; performing knowledge reasoning and graph completion using the dynamic time series graph embedding; performing real-time updating and incremental learning on the completed knowledge graph; generating a precise traffic diversion strategy based on the updated knowledge graph; optimizing the agency sales process based on the precise traffic diversion strategy; evaluating and providing feedback on the optimized agency sales process, and using the feedback results for each step of the optimization method.

[0016] Specifically, in the step of acquiring and preprocessing multi-source heterogeneous data, an adaptive multi-dimensional anomaly detection algorithm is used, and the algorithm includes the following steps:

[0017] a) Use locality sensitive hashing to map high-dimensional data to low-dimensional space, where the mapping function is:

[0018]

[0019] Where x is the input data, a is a random vector, b is a random offset, and w is the bucket width; b) Dynamic clustering is applied within each hash bucket using the improved DBSCAN algorithm, where the ∈ parameter is adaptively adjusted:

[0020] ∈i =μ i +α·σ i

[0021] Among them, ∈ i is the neighborhood radius of the i-th dimension, μ i and σ i are the mean and standard deviation of the i-th dimension respectively, and α is an adjustable parameter;

[0022] c) Determine the anomaly score:

[0023]

[0024] Among them, AS(x) is the anomaly score of point x, d(x,C) is the distance from point x to the nearest cluster center C, σ C is the standard deviation of cluster C;

[0025] d) Dynamically determine the abnormal threshold based on extreme value theory.

[0026] Specifically, the step of realizing multimodal data fusion and entity recognition includes using a quantum-inspired multimodal fusion network, including the following steps:

[0027] a) For each modality of data x i Encoding quantum states:

[0028]

[0029] Among them, |ψ i > is the quantum state, α ij is the complex amplitude, |j> is the ground state, and satisfies ∑ j |α ij | 2

[0030] =1;

[0031] b) Apply the inter-modal entanglement layer and design the entanglement operator U ent :U ent =exp(-iH ent ),

[0032] in, σ x ,σ y ,σ z is the Pauli matrix, J ij is the learnable entanglement strength;

[0033] c) Using the quantum measurement layer, design a set of learnable measurement operators M k , and get the classic output:

[0034] y k =<ψ|Mk |ψ>

[0035] Among them, y k is the measurement result, |ψ> is the quantum state;

[0036] d) Process the measurements using traditional neural network layers to obtain the final fused representation.

[0037] Specifically, in the step of performing relationship extraction and attribute mapping, a graph-enhanced recursive neural reasoning network is used, including the following steps:

[0038] a) Use the TransE algorithm to embed the existing knowledge graph:

[0039] ||h+rt|| 2 ≈0

[0040] Among them, h, r, t are the embedding vectors of the head entity, relation, and tail entity respectively;

[0041] b) Use BiLSTM to encode the input sentence:

[0042] h t =BiLSTM(x t ,h t-1 )

[0043] Among them, h t is the hidden state at time t, x t is the input at time t;

[0044] c) Determine the attention score of sentence representation and graph embedding:

[0045] α i =softmax(v T tanh(W[h i ; g j ]))

[0046] Among them, α i is the attention score, h i is the hidden state of the sentence, g j is graph embedding, v and W are learnable parameters;

[0047] d) Design a recursive structure to simulate multi-hop reasoning:

[0048] z t =f(W z [h t ; z t-1 ;c t ])

[0049] Among them, z t is the reasoning state at time t, ht is the hidden state of the sentence, c t is the graph-enhanced context vector, W z is a learnable parameter;

[0050] e) Use the softmax layer to adjust the final z T Categorize relationships.

[0051] Specifically, the step of constructing the dynamic time sequence graph embedding includes using a quantum-inspired space-time graph embedding algorithm, which includes the following steps:

[0052] a) Initialize entity e i and the relationship j The quantum bitmap embedding of:

[0053] Q(e i )=|ψ i >=α i |0>+β i |1>

[0054] Q(r j )=|φ j >=γ j |0>+δ j |1>where |α i | 2 +|β i | 2 =1,|γ j | 2 +|δ j | 2 =1;

[0055] b) Define the nonlinear space-time evolution operator U t :

[0056] U t =exp(-iH t )

[0057] Among them, H t is a time-varying Hamiltonian operator:

[0058]

[0059] σ k is the Pauli matrix, w k (t) and J ij (t) is the time-varying coefficient;

[0060] c) Introducing the quantum walk attention mechanism:

[0061] A QW (i,j,t)=<ψ i(t)|U QW (t)|ψ j (t)>

[0062] Among them, U QW (t) is a quantum walk operator based on graph structure;

[0063] d) Multiscale time-dependent modeling using fractional quantum field theory:

[0064] Φ(x,t)=∫K α (xy,ts)ψ(y,s)dyds

[0065] Among them, K α is the fractional diffusion kernel, α∈(0,2) is the fractional parameter;

[0066] e) Introduce the topological entropy regularization term:

[0067]

[0068] Among them, p ij is the transition probability from node i to node j;

[0069] f) Design non-equilibrium thermodynamic optimization objectives:

[0070]

[0071] Among them, L task Loss of main task, is the rate of change of system entropy;

[0072] g) Design of adaptive quantum measurement strategies:

[0073]

[0074] Among them, M is a learnable measurement operator.

[0075] Specifically, the step of performing knowledge reasoning and graph completion includes using a multi-granularity causal reasoning network, including the following steps:

[0076] a) Use spectral clustering algorithm to divide the knowledge graph into multiple granularity levels;

[0077] b) At each granularity level, the PC algorithm is used to discover potential causal relationships;

[0078] c) Estimating causal strength using structural equation modeling (SEM):

[0079] X j =f j (PA j ,U j )

[0080] Among them, X j is the dependent variable, PA j For X j The parent node set, U j is an exogenous variable;

[0081] d) Design an attention mechanism to integrate causal information of different granularities:

[0082] α g =softmax(v T tanh(W g h g ))h final =∑ g α g h g Among them, α g is the attention weight, h g represents causal information of different granularities, v and W g is a learnable parameter;

[0083] e) Based on the final representation h final Perform link prediction.

[0084] Specifically, in the step of generating a precise traffic diversion strategy, a hierarchical spatiotemporal attention strategy network is used, including the following steps:

[0085] a) Define state representation:

[0086] s t =[u t ;p t ;c t ]

[0087] Among them, u t is the user feature, p t is the product feature, c t is a contextual feature;

[0088] b) Define a two-level action space: A l for advanced strategies) and A 2 For specific operations;

[0089] c) Determine spatiotemporal attention:

[0090] α t =softmax(v T tanh(W s s t +W h h t-1 ))

[0091] Among them, h t-1is the historical hidden state, v,W s ,W h is a learnable parameter;

[0092] d) Build a policy network:

[0093] π θ (a|s)=softmax(f θ (s,α))

[0094] Among them, f θ is a parameterized neural network,

[0095] e) Build a value network:

[0096]

[0097] in, is a parameterized neural network;

[0098] f) Training using an asynchronous advantage actor-critic algorithm.

[0099] Specifically, the step of optimizing the agency sales process further includes using a graph neural network to perform agency matching and anomaly detection.

[0100] Specifically, in the step of effect evaluation and feedback, dynamic weight multi-objective Bayesian optimization is used, including the following steps:

[0101] a) Define multiple objective functions:

[0102] F(x)=[f 1 (x),f 2 (x),...,f m (x)]

[0103] Among them, f i (x) is the i-th objective function,

[0104] b) Establish a Gaussian process model for each objective function:

[0105] f i (x)~GP(μ i (x),k i (x,x′))

[0106] Among them, μ i (x) is the mean function, k i (x,x′) is the kernel function;

[0107] c) Determine dynamic weights:

[0108]

[0109] Among them, λ is the temperature parameter, is the current optimal value of target i, σ i is the standard deviation;

[0110] d) Calculation acquisition function:

[0111]

[0112] Among them, EI i (x) is the expected improvement of the ith objective;

[0113] e) Use the DIRECT algorithm to maximize EI(x) and select the next evaluation point;

[0114] f) Evaluate the new point, update the Gaussian process model and dynamic weights, and repeat steps ce.

[0115] A digital precise traffic diversion and agency sales management system integrating knowledge graphs includes: a data preprocessing module for acquiring and preprocessing multi-source heterogeneous data; a data fusion module for realizing multimodal data fusion and entity recognition; a relationship extraction module for performing relationship extraction and attribute mapping; a graph embedding module for constructing dynamic time series graph embedding; a knowledge reasoning module for performing knowledge reasoning and graph completion; an update learning module for performing real-time update and incremental learning; a strategy generation module for generating precise traffic diversion strategies; a process optimization module for optimizing agency sales processes; an evaluation feedback module for performing effect evaluation and feedback; wherein the output of the evaluation feedback module is connected to the input of the data preprocessing module to form a closed-loop optimization.

[0116] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0117] 1. Effective integration of multi-source heterogeneous data: Through the AMODA algorithm and the QIMMFN algorithm, this method can effectively process and fuse heterogeneous data from different sources to build a more comprehensive user and product representation.

[0118] 2. Dynamic time series modeling: The QISGE algorithm implements dynamic time series graph embedding, which can capture changes in user interests and market trends in real time and improve the timeliness of recommendations.

[0119] 3. Deep knowledge reasoning: The MGCIN algorithm greatly enhances the system's reasoning ability through multi-granularity causal reasoning, and can discover deeper semantic relationships.

[0120] 4. Personalized strategy generation: The HSASPN algorithm can generate highly personalized traffic strategies based on user characteristics and contextual information to improve user satisfaction.

[0121] 5. Long-tail product mining: By comprehensively considering multiple factors, this method significantly improves the utilization rate of long-tail products and enriches the diversity of recommendations.

[0122] 6. Real-time and scalability: Using incremental learning and distributed processing technology, this method can cope with large-scale data and high-concurrency requests, meeting the needs of real-time recommendations.

[0123] 7. Optimization of the agency sales process: Through graph neural network technology, this method realizes the intelligent matching and risk management of agents, and improves the overall operational efficiency.

[0124] 8. Adaptive optimization: The DWMOBO algorithm implements dynamic optimization of multiple objectives, enabling the system to strike a balance between different objectives (such as conversion rate, user satisfaction, platform revenue, etc.) and to adjust strategies adaptively.

[0125] 9. Explainability: The knowledge graph-based method provides better explainability, which helps to understand the reasons behind the recommendation results and increases user trust.

[0126] 10. Knowledge accumulation: Through continuous learning and graph completion, this method can continuously accumulate and update domain knowledge and form valuable knowledge assets.

[0127] In summary, the digital precision traffic generation and consignment management method integrated with knowledge graph provided by the present invention has comprehensively improved the accuracy, personalization and intelligence level of digital marketing by innovatively integrating multiple advanced technologies. It not only solves many problems faced by existing technologies, but also provides strong technical support for the precision marketing and user services of e-commerce platforms, and is expected to significantly improve the platform's operational efficiency and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0128] Figure 1 This is a flow chart of the digital precision traffic diversion and sales agency management method integrating knowledge graph of the present invention.

[0129] Figure 2 This is a schematic diagram of the framework of the deep learning-based smart market product demand forecasting system of the present invention. DETAILED DESCRIPTION

[0130] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects thereof are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0131] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0132] The present invention proposes a digital precision traffic diversion and consignment management method and system integrating knowledge graph, which is particularly suitable for precision marketing and user service optimization of e-commerce platforms. The following will describe the implementation of the present invention in detail in combination with an application scenario of a large e-commerce platform.

[0133] The e-commerce platform has millions of users and hundreds of thousands of products, generating a large amount of user behavior data and transaction data every day. The main challenges faced by the platform include: how to accurately recommend suitable products to users, how to optimize the matching and management of agents, and how to improve overall marketing efficiency and user satisfaction. The present invention provides an innovative solution by integrating knowledge graph technology, which specifically includes the following steps:

[0134] Data preprocessing step,In this step, the data preprocessing module first obtains and preprocesses heterogeneous data from multiple sources. These data sources include:

[0135] a) User personal information database: contains basic information of users, such as age, gender, geographic location, etc.

[0136] b) User behavior log: records users’ browsing, searching, clicking and other behaviors on the platform.

[0137] c) Product information database: contains detailed information about products, such as category, price, inventory, etc.

[0138] d) Transaction record database: records all completed transaction information.

[0139] e) Social media data: public information of users on major social platforms obtained through API.

[0140] The data preprocessing module uses the adaptive multi-dimensional anomaly detection algorithm (AMODA) to process these heterogeneous data. The specific steps of the AMODA algorithm are as follows:

[0141] a) Use Locality Sensitive Hashing (LSH) to map high-dimensional data into a low-dimensional space:

[0142]

[0143] Where x is the input data (e.g., a user’s feature vector), a is a random vector (usually sampled from a standard normal distribution), b is a random offset (usually uniformly sampled in the range [0, w)), and w is the bucket width (determined according to the data distribution, usually between 1 and 10).

[0144] For example, suppose we have a user feature vector x = [25, 1, 10000, 50], which represents age, gender, annual income, and number of logins in the past 30 days.

[0145] We might choose a=[0.1,-0.5,0.001,0.2],b=2,w=5.

[0146]

[0147] b) Apply dynamic clustering within each hash bucket using a modified DBSCAN algorithm where the ∈ parameter is adaptively adjusted:

[0148] ∈ i =μ i +α·σ i

[0149] Among them, ∈ i is the neighborhood radius of the i-th dimension, μ i and σ i are the mean and standard deviation of the i-th dimension respectively, and α is an adjustable parameter (usually between 0.5 and 2).

[0150] Assume that in a certain hash bucket, the mean μ of the age dimension is 30, the standard deviation σ is 5, and we choose α=1, then:

[0151] ∈ age =30+1·5=35

[0152] c) Calculate the anomaly score:

[0153]

[0154] Among them, AS(x) is the anomaly score of point x, d(x,C) is the distance from point x to the nearest cluster center C, σ C is the standard deviation of cluster C.

[0155] Assume that the distance between a user point x and the nearest cluster center is d(x,C) = 10, and the standard deviation of the cluster is σ C =3, then:

[0156]

[0157] d) Use extreme value theory (EVT) to dynamically determine the anomaly threshold.

[0158] Through the AMODA algorithm, we can effectively detect and process abnormal data, such as identifying abnormal user behavior or product information, thereby improving the accuracy of subsequent analysis.

[0159] Data fusion step, in this step, the data fusion module uses the quantum-inspired multimodal fusion network (QIMMFN) to achieve multimodal data fusion and entity recognition. This is particularly effective for integrating various user information (such as personal information, browsing history, purchase history, social media data, etc.). The specific steps of the QIMMFN algorithm are as follows:

[0160] a) For each modality of data x i Encoding quantum states:

[0161]

[0162] Among them, |ψ i > is the quantum state, α ij is the complex amplitude, |j> is the ground state, and satisfies ∑ j |α ij | 2 =1.

[0163] Suppose we have text descriptions of users and image data. For text, we may have |ψ text >=0.6|0>+0.8|1>, for the image, we may have b) Apply the inter-modal entanglement layer and design the entanglement operator U ent :

[0164] U ent =exp(-iH ent )

[0165] in, σ y ,σ z is the Pauli matrix, J ij is the learnable entanglement strength. This step allows information from different modalities to influence each other, for example, a user’s browsing history may affect the interpretation of their social media behavior;

[0166] c) Using the quantum measurement layer, design a set of learnable measurement operators M k , and get the classic output:

[0167] y k =<ψ|M k |ψ>

[0168] Among them, y k is the measurement result, |ψ> is the quantum state.

[0169] Suppose we design a measurement operator M = |0><0|-|1><1|, for the previous text quantum state, the measurement result will be:

[0170] y=<ψ text |M|ψtext >=0.6 2 -0.8 2 =-0.28

[0171] d) Use traditional neural network layers to process the measurement results and obtain the final fused representation. Through the QIMMFN algorithm, we can effectively fuse multimodal data, such as integrating users' text comments, shopping history, and social media activities into a unified user representation, which is crucial for subsequent personalized recommendations and precision marketing.

[0172] Relationship extraction step: In this step, the relationship extraction module uses the graph-enhanced recursive neural reasoning network (GERNRN) to perform relationship extraction and attribute mapping. This is particularly important for understanding the complex relationships between users and products, users and users, and products and products. The specific steps of the GERNRN algorithm are as follows:

[0173] a) Use the TransE algorithm to embed the existing knowledge graph

[0174] ||h+rt|| 2 ≈0

[0175] Among them, h, r, t are the embedding vectors of the head entity, relation, and tail entity respectively. Assuming we have a relation "user A likes product B", we may get: h 用户A +r 喜欢 ≈t 商品B .

[0176] b) Use BiLSTM to encode the input sentence: h t =BiLSTM(x t ,h t-1 )

[0177] Among them, h t is the hidden state at time t, x t is the input at time t. This step can process text data such as user comments or product descriptions to capture sequence information.

[0178] c) Calculate the attention score of sentence representation and graph embedding:

[0179] α i =softmax(v T tanh(W[h i ; g j ]))

[0180] Among them, α i is the attention score, h i is the hidden state of the sentence, g jis the graph embedding, v and W are learnable parameters. This step enables the model to focus on the knowledge graph information most relevant to the current input.

[0181] d) Design a recursive structure to simulate multi-hop reasoning:

[0182] z t =f(W z [h t ; z t-1 ;c t ])

[0183] Among them, z t is the reasoning state at time t, h t is the hidden state of the sentence, c t is the graph-enhanced context vector, W z are learnable parameters. This allows the model to perform complex reasoning, such as “user A likes item B, item B belongs to category C, so user A may also like other items in category C.”

[0184] e) Use the softmax layer to adjust the final z T Perform relationship classification. Through the GERNRN algorithm, we can effectively extract the relationship between entities, such as identifying "items that users often buy together" and "similar user groups", which is crucial for building a richer knowledge graph and making more accurate recommendations.

[0185] Graph embedding step, in this step, the graph embedding module uses the quantum inspired spatiotemporal graph embedding (QISGE) algorithm to build dynamic temporal graph embedding. This is particularly important for capturing the dynamic changes of user interests and market trends. The specific steps of the QISGE algorithm are as follows:

[0186] a) Initialize entity e i and the relationship j The quantum bitmap embedding of:

[0187] Q(e i )=|ψ i >=α i |0>+β i |1>

[0188] Q(r j )=|φ j >=γ j |0>+δ j |1>

[0189] Among them, |α i | 2 +|β i | 2 =1,|γ j |2 +|δ j | 2 =1; for a user entity, we may have Q(userA)=0.6|0>+0.8|1>, which represents some characteristics of user A.

[0190] b) Define the nonlinear space-time evolution operator U t :

[0191] U t =exp(-iH t )

[0192] Among them, H t is a time-varying Hamiltonian operator:

[0193]

[0194] σ k is the Pauli matrix, w k (t) and J ij (t) is the time-varying coefficient; this allows the representation of entities and relations to change dynamically over time, for example capturing seasonal changes in user interests.

[0195] c) Introducing the quantum walk attention mechanism:

[0196] A QW (i,j,t)=<ψ i (t)|U QW (t)|ψ j (t)>

[0197] Among them, U QW (t) is a quantum walk operator based on the graph structure; this helps the model focus on the most relevant parts of the graph, such as focusing on other users or products that are most relevant to the current user in a recommendation system.

[0198] d) Multiscale time-dependent modeling using fractional quantum field theory:

[0199] Φ(x,t)=∫K α (xy,ts)ψ(y,s)dyds

[0200] Among them, K α is a fractional diffusion kernel and α∈(0,2) is a fractional parameter; this allows the model to capture dependencies at different time scales, such as users’ short-term interests and long-term preferences.

[0201] e) Introduce the topological entropy regularization term:

[0202]

[0203] Among them, pij is the transition probability from node i to node j; this helps to preserve key features of the graph structure, such as the key structure of a user’s social network.

[0204] f) Design non-equilibrium thermodynamic optimization objectives:

[0205]

[0206] Among them, L task Loss of main task, is the rate of change of the system entropy; this optimization objective allows the model to adapt to new changes while maintaining stability, such as adapting to new market trends while maintaining basic user interests.

[0207] g) Design of adaptive quantum measurement strategies:

[0208]

[0209] Here, M is a learnable measurement operator. This step converts the quantum representation into a classical representation for subsequent processing.

[0210] Knowledge reasoning step, in this step, the knowledge reasoning module uses the multi-granularity causal reasoning network (MGCIN) to perform knowledge reasoning and graph completion. This is particularly important for understanding the reasons behind user behavior and predicting future trends. The specific steps of the MGCIN algorithm are as follows:

[0211] a) Use spectral clustering algorithms to divide the knowledge graph into multiple levels of granularity. For example, we can divide the graph into user level, product category level, and overall market level. This allows reasoning at different levels of abstraction.

[0212] b) At each level of granularity, use the PC algorithm to discover potential causal relationships. The PC algorithm builds a causal graph through conditional independence tests. For example, at the user level, we may find a causal chain such as "user age" → "purchase category" → "purchase frequency".

[0213] c) Estimating causal strength using structural equation modeling (SEM):

[0214] X j =f j (PA j ,U j )

[0215] Among them, X j is the dependent variable, PA j For X j The parent node set, U jis an exogenous variable. For example, we may get the following equation: Purchase frequency = 0.3·age + 0.5·income + 0.2·random factor;

[0216] d) Design an attention mechanism to integrate causal information of different granularities:

[0217] α g =softmax(v T tanh(W g h g )

[0218]

[0219] Among them, α g is the attention weight, h g represents causal information of different granularities, v and W g is a learnable parameter. This allows the model to dynamically adjust the attention paid to information of different granularities according to specific circumstances. For example, when predicting a user’s purchase behavior, it may pay more attention to user-level causal relationships; while when predicting overall market trends, it may pay more attention to market-level causal relationships.

[0220] e) Based on the final representation h final Perform link prediction. This step can be used to predict missing relationships in the knowledge graph, such as predicting products that users may be interested in but have not yet interacted with.

[0221] Through the MGCIN algorithm, we can have a deeper understanding of user behavior and market dynamics, such as identifying complex patterns such as "young users are more likely to buy electronic products during the holidays", thereby providing stronger support for precision marketing.

[0222] Update learning step, in this step, the update learning module is responsible for real-time update and incremental learning of the knowledge graph. This is crucial to maintain the timeliness and adaptability of the system. It mainly includes the following aspects:

[0223] a) Use the ReservoirSampling algorithm to maintain a fixed-size sample pool to ensure the balance between new and old data. For example, we can maintain a sample pool containing the most recent 1 million user behaviors, and when new data enters, the old data is replaced with a certain probability.

[0224] b) Apply Elastic Weight Consolidation (EWC) to prevent catastrophic forgetting. EWC prevents important parameters from changing too much by adding additional loss terms to them. For example, when updating the user interest model, we protect the parameters that represent the user's long-term interests from changing drastically due to short-term behavior.

[0225] c) Implement an adaptive sliding window, where the window size is dynamically adjusted based on the data flow rate. For example, during peak shopping periods, we may reduce the window size to capture user interest changes more quickly; while during quiet periods, we may expand the window to obtain more stable patterns.

[0226] d) Use the Hoeffding tree algorithm for real-time decision tree updates. This allows us to quickly update classification or regression models when new data arrives, such as adjusting user grouping or product classification in real time.

[0227] e) Apache Flink is used for distributed stream processing to ensure the scalability of the system. This enables the system to process large-scale real-time data streams, such as processing click stream data of millions of users at the same time.

[0228] f) Implement a distributed state management mechanism based on the Raft consensus algorithm. This ensures that the state of the knowledge graph remains consistent in a distributed environment and prevents data inconsistency caused by node failures.

[0229] Through these technologies, the system can continuously learn and adapt to new user behaviors and market trends, keeping recommendations and marketing strategies timely.

[0230] Strategy Generation Step,In this step, the strategy generation module uses the Hierarchical Spatiotemporal Attention Strategy Network (HSASPN) to generate accurate traffic diversion strategies. This is crucial for recommending the right products to the right users at the right time. The specific steps of the HSASPN algorithm are as follows:

[0231] a) Define state representation: s t =[u t ;p t ;c t ]

[0232] Among them, u t is the user feature, p t is the product feature, c t The state may include the user's demographic information, recent browsing history, characteristics of the currently browsed item, time, location, etc.

[0233] b) Define a two-level action space: A 1 (Advanced Strategy) and A 2 (Specific operation). For example, A 1 This may include strategies such as "recommending similar products", "recommending complementary products", and "recommending popular products". 2 It is the specific product ID or discount plan.

[0234] c) Calculate spatiotemporal attention: α t=softmax(v T tanh(W s s t +W h h t-1 ))

[0235] Among them, h t-1 is the historical hidden state, v,W s ,W h are learnable parameters. This allows the model to dynamically adjust its focus based on the current state and historical information. For example, when a user is browsing a phone, the model may pay more attention to the user's budget and brand preferences.

[0236] d) Build a policy network:

[0237] π θ (a|s)=softmax(f θ (s,α))

[0238] Among them, f θ is a parameterized neural network. This network outputs the probability of taking each action in a given state.

[0239] e) Build a value network:

[0240]

[0241] in, It is a parameterized neural network. This network estimates the long-term value of a given state, helping the system to strike a balance between short-term benefits and long-term user value.

[0242] f) Use the A3C (Asynchronous Advantage Actor-Critic) algorithm for training. This allows the system to learn efficiently in a large-scale parallel environment and adapt to the high concurrency characteristics of e-commerce platforms.

[0243] Through the HSASPN algorithm, the system is able to generate fine-grained personalized marketing strategies. For example, for a user who often browses sports equipment on weekends, the system may push a limited-time discount on sports shoes on Friday afternoon to maximize the conversion probability.

[0244] Process optimization step: In this step, the process optimization module mainly uses graph neural network (GNN) to optimize the consignment sales process. This is crucial to improving the overall operational efficiency and user satisfaction of the platform. It mainly includes the following aspects:

[0245] a) Use GraphAttentionNetworks (GAT) to build a bipartite graph of vendors and products. This allows the system to capture complex relationships between vendors and products. For example, some vendors may perform better on specific categories of products.

[0246] b) Apply the Hungarian algorithm for optimal matching. Based on the representation learned by GAT, the system can find the most suitable distributor for each product, or assign the most suitable product combination to each distributor.

[0247] c) Implement graph-based anomaly detection algorithms, such as GraphAutoencoder. This can help identify abnormal sales behavior, such as sudden sales surges or abnormal return rates.

[0248] d) Use the LocalOutlierFactor (LOF) algorithm to detect local anomalies. This helps to find abnormal sales behavior in a specific product category or a specific time period.

[0249] e) Design an abnormal pattern recognition method based on time series graphs. This can capture the abnormal patterns of sales behavior over time, such as the difference between normal sales fluctuations and abnormal fluctuations of seasonal products.

[0250] f) Build a dynamic Bayesian network model to evaluate the risks in the consignment sales process in real time. This allows the system to dynamically adjust risk management strategies, such as adjusting credit limits based on the consignment seller’s historical performance and current market conditions.

[0251] g) Implement a risk mitigation strategy generator based on Monte Carlo Tree Search (MCTS). This can help the system generate a series of possible mitigation strategies and evaluate their effectiveness when potential risks are discovered.

[0252] Through these technologies, the system can continuously optimize the agency sales process, improve matching efficiency, reduce risks, and thus enhance the overall platform performance.

[0253] Evaluation and feedback step: In this step, the evaluation and feedback module uses the dynamic weighted multi-objective Bayesian optimization (DWMOBO) algorithm to perform effect evaluation and feedback. This is crucial for continuously optimizing system performance and adapting to the ever-changing market environment. The specific steps of the DWMOBO algorithm are as follows:

[0254] a) Define multiple objective functions: F(x) = [f 1 (x),f 2 (x),...,f m (x)]

[0255] Among them, f i(x) is the i-th objective function. For example, f 1 It may be the conversion rate, f 2 It may be user satisfaction, f 3 It could be platform revenue, etc.

[0256] b) Establish a Gaussian process model for each objective function:

[0257] f i (x)~GP(μ i (x),k i (x,x′))

[0258] Among them, μ i (x) is the mean function, k i (x,x′) is the kernel function. This allows the system to model the uncertainty of each objective function

[0259] c) Calculate dynamic weight:

[0260]

[0261] Where λ is the temperature parameter (usually set between 0.1 and 1), is the current optimal value of target i, σ i is the standard deviation. This allows the system to dynamically adjust the focus on different goals. For example, if the conversion rate is already high, the system may pay more attention to improving user satisfaction.

[0262] d) Calculation acquisition function:

[0263]

[0264] Among them, EI i (x) is the expected improvement of the ith objective. This function guides the system to strike a balance between exploration (trying new strategies) and exploitation (using known good strategies)

[0265] e) Use the DIRECT algorithm to maximize EI(x)\text{EI}(x)EI(x) and select the next evaluation point. The DIRECT algorithm is a deterministic global optimization algorithm that is particularly suitable for dealing with multi-objective problems.

[0266] f) Evaluate the new point, update the Gaussian process model and dynamic weights, and repeat steps ce.

[0267] Through the DWMOBO algorithm, the system is able to find the best balance between multiple potentially conflicting objectives. For example, the system may find that slightly reducing the accuracy of recommendations for certain high-value users, but significantly increasing their exploratory recommendations, can improve overall user satisfaction and platform revenue in the long run.

[0268] Finally, the output of the evaluation feedback module is fed back to the data preprocessing module to form a complete closed-loop optimization system. For example, if it is found that the satisfaction of certain user groups continues to decline, the system may conduct a more detailed analysis of the data of these users in the data preprocessing stage, so as to give them more attention in subsequent steps.

[0269] Through the above nine steps, the digital precision traffic generation and consignment sales management method integrated with knowledge graph of the present invention can comprehensively improve the operational efficiency and user experience of the e-commerce platform. From data processing to strategy generation, to effect evaluation and feedback, each step uses advanced algorithms and technologies to ensure the efficiency, accuracy and adaptability of the system. This method can not only achieve accurate personalized recommendations and marketing, but also optimize the entire consignment sales process, while adapting to the ever-changing e-commerce environment through continuous learning and optimization.

[0270] The digital precision traffic diversion and sales management system integrated with knowledge graph of the present invention is a highly integrated intelligent platform, which includes the following core modules:

[0271] The data preprocessing module 1 is responsible for acquiring and preprocessing multi-source heterogeneous data and using the AMODA algorithm to handle outliers and noise.

[0272] Data fusion module 2 uses the QIMMFN algorithm to achieve multimodal data fusion and entity recognition, integrating data from different sources and types into a unified representation.

[0273] Relation extraction module 3 uses the GERNRN algorithm to perform relationship extraction and attribute mapping to construct a preliminary knowledge graph structure.

[0274] The graph embedding module 4 uses the QISGE algorithm to construct dynamic temporal graph embeddings to capture the dynamic relationships between entities.

[0275] The knowledge reasoning module 5 performs knowledge reasoning and graph completion through the MGCIN algorithm to enrich the content of the knowledge graph.

[0276] The update learning module 6 is responsible for real-time updating and incremental learning to ensure that the system can adapt to the latest data and trends.

[0277] Strategy generation module 7 uses the HSASPN algorithm to generate precise traffic diversion strategies to achieve personalized recommendations and marketing.

[0278] Process optimization module 8 uses graph neural networks to optimize the sales process and improve overall operational efficiency.

[0279] The evaluation and feedback module 9 uses the DWMOBO algorithm to perform effect evaluation and feedback to continuously optimize system performance.

[0280] These modules work closely together to form a closed-loop intelligent system. Data starts with preprocessing, and then goes through steps such as fusion, relationship extraction, and graph embedding to build a rich knowledge graph. Based on this graph, the system generates accurate marketing strategies and optimizes the consignment sales process. Finally, the performance of the entire system is continuously optimized through the evaluation feedback module.

[0281] In order to verify the superiority of the present invention, we conducted a series of comparative experiments. The following is a detailed comparison of Example 1 of the present invention and three comparative examples:

[0282] Example 1: Method and system for fully implementing the present invention

[0283] Comparative Example 1: Traditional Collaborative Filtering Recommendation System

[0284] Comparative Example 2: Recommendation System Based on Simple Graph Neural Network

[0285] Comparative Example 3: System using traditional knowledge graph but without dynamic updates

[0286] Test indicators and their testing standards and methods:

[0287] 1. Recommendation accuracy (Precision@K): The proportion of correct recommendations in Top-K recommendations.

[0288] Detection method: Use the holdout method to divide the user interaction data into a training set and a test set, and calculate the proportion of correct predictions in the test set.

[0289] 2. User satisfaction: the user’s satisfaction with the recommendation results.

[0290] Detection method: Through user questionnaire survey, using Likert scale of 1-5 for scoring.

[0291] 3. Average processing time: the average time it takes for the system to process a request.

[0292] Detection method: Record system response time and calculate the average value.

[0293] 4. Long-tail product coverage: the proportion of long-tail products (lower sales) in the recommended results.

[0294] Detection method: Define long-tail products (such as sales volume 20% lower than the average) and calculate the proportion of long-tail products in the recommendation results.

[0295] 5. Knowledge graph completeness: the richness of entities and relationships in the knowledge graph.

[0296] Detection method: Count the number of entities and relations in the graph and compare with the benchmark dataset.

[0297] The test results are shown in the following table:

[0298] index Example 1 Comparative Example 1 Comparative Example 2 Comparative Example 3 Recommendation accuracy (P@10) 0.82 0.65 0.73 0.7 User satisfaction (1-5 points) 4.5 3.2 3.8 3.6 Average processing time (ms) 150 80 200 180 Long-tail product coverage (%) 35 10 20 25 Knowledge graph completeness (%) 95 N / A 70 85

[0299] According to the test results, Example 1 (i.e., the complete implementation of the present invention) shows significant advantages in most indicators and can therefore be considered as the best embodiment.

[0300] The analysis and explanation are as follows:

[0301] 1. Recommendation accuracy: The method of the present invention achieves a high accuracy of 0.82, far exceeding other comparison methods. This is mainly due to the combination of the QISGE algorithm and the MGCIN algorithm, which enables the system to capture more complex and dynamic user-product relationships.

[0302] 2. User satisfaction: The high score of 4.5 reflects the high recognition of users for the recommendation results. This is not only related to the high accuracy, but also to the fact that the HSASPN algorithm can generate more personalized and diverse recommendations.

[0303] 3. Average processing time: Although the processing time (150ms) of the present invention is higher than that of the simple collaborative filtering method, considering the complexity of the system and the performance improvement, this processing time is acceptable and still meets the requirements of real-time recommendation.

[0304] 4. Long-tail product coverage: The present invention performs well in this indicator (35%), which proves that the system can effectively tap the potential of long-tail products. This is mainly due to the effective fusion of multimodal data by the QIMMFN algorithm, which enables the system to understand the characteristics of products more comprehensively.

[0305] 5. Knowledge graph completeness: The high completeness of 95% reflects the advantages of the present invention in building and maintaining knowledge graphs. This is mainly due to the combination of the GERNRN algorithm and the real-time update mechanism.

[0306] These test results fully illustrate the following advantages of the present invention:

[0307] 1. High-precision recommendations: By integrating multi-source data and adopting advanced graph embedding technology, this system can provide more accurate personalized recommendations.

[0308] 2. User experience optimization: High user satisfaction indicates that the system not only provides accurate recommendations, but also performs well in terms of diversity and timeliness of recommendations.

[0309] 3. Long-tail discovery capability: The system significantly improves the exposure rate of long-tail products, which is of great significance for improving the overall product utilization rate of the platform and the diversity of user choices.

[0310] 4. Knowledge accumulation and utilization: A highly complete knowledge graph enables the system to perform deeper reasoning and prediction, providing a solid foundation for precision marketing.

[0311] 5. Real-time and adaptability: Despite the high complexity of the system, it can still complete processing within an acceptable time and adapt to changing user needs and market environment through continuous learning.

[0312] In general, these results verify the innovations of the present invention in terms of integrating knowledge graphs, multimodal data processing, dynamic time series modeling, etc., as well as the significant performance improvements brought about by these innovations in practical applications. This system not only improves the accuracy of recommendations and user satisfaction, but also performs well in the utilization of long-tail products and knowledge accumulation, providing strong technical support for the precision marketing and user services of e-commerce platforms.

[0313] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A digital precision sales agent management method integrating knowledge graph, characterized in that: The following steps are involved: Acquire and preprocess multi-source heterogeneous data; Based on the preprocessed data, multimodal data fusion and entity recognition are realized; relationship extraction and attribute mapping are performed based on the fused data and the recognized entities; dynamic time sequence graph embedding is constructed based on the extracted relationship and the mapped attributes; knowledge reasoning and graph completion are performed using the dynamic time sequence graph embedding; Real-time update and incremental learning of the completed knowledge graph; Generate accurate traffic diversion strategies based on the updated knowledge graph; Optimize the sales process based on the precise traffic diversion strategy; Evaluate and provide feedback on the optimized sales process, and use the feedback results in each step of the optimization method.

2. The method according to claim 1, characterized in that In the step of acquiring and preprocessing multi-source heterogeneous data, an adaptive multi-dimensional anomaly detection algorithm is used, and the algorithm includes the following steps: a) Use locality sensitive hashing to map high-dimensional data to low-dimensional space, where the mapping function is: Where x is the input data, a is a random vector, b is a random offset, and w is the bucket width; b) Apply dynamic clustering within each hash bucket using a modified DBSCAN algorithm where the ∈ parameter is adaptively adjusted: ∈ i =μ i +a·s i Among them, ∈ i is the neighborhood radius of the i-th dimension, μ i and σ i are the mean and standard deviation of the i-th dimension respectively, and α is an adjustable parameter; c) Determine the anomaly score: Among them, AS(x) is the anomaly score of point x, d(x,C) is the distance from point x to the nearest cluster center C, σ C is the standard deviation of cluster C; d) Dynamically determine the abnormal threshold based on extreme value theory.

3. The method according to claim 1, characterized in that In the step of realizing multimodal data fusion and entity recognition, a quantum-inspired multimodal fusion network is used, which includes the following steps: a) For each modality of data x i Encoding quantum states: Among them, |ψ i > is the quantum state, α ij is the complex amplitude, |j> is the ground state, and satisfies ∑ j |α ij | 2 =1; b) Apply the inter-modal entanglement layer and design the entanglement operator U ent :U ent =exp(-iH ent ), in, σ x ,σ y ,σ z is the Pauli matrix, J ij is the learnable entanglement strength; c) Using the quantum measurement layer, design a set of learnable measurement operators M k , and get the classic output: y k =<ψ|M k |ψ> Among them, y k is the measurement result, |ψ> is the quantum state; d) Process the measurements using traditional neural network layers to obtain the final fused representation.

4. The method according to claim 1, characterized in that: In the step of performing relationship extraction and attribute mapping, a graph-enhanced recursive neural reasoning network is used, which includes the following steps: a) Use the TransE algorithm to embed the existing knowledge graph: ||h+rt||2≈0 Among them, h, r, t are the embedding vectors of the head entity, relation, and tail entity respectively; b) Use BiLSTM to encode the input sentence: h t =BiLSTM(x t ,h t-1 ) Among them, h t is the hidden state at time t, x t is the input at time t; c) Determine the attention score of sentence representation and graph embedding: α i =softmax(v T tanh(W[h i ;g j ])) Among them, α i is the attention score, h i is the hidden state of the sentence, g j is graph embedding, v and W are learnable parameters; d) Design a recursive structure to simulate multi-hop reasoning: z t =f(W z [h t ;z t-1 ;c t ]) Among them, z t is the reasoning state at time t, h t is the hidden state of the sentence, c t is the graph-enhanced context vector, W z is a learnable parameter; e) Use the softmax layer to adjust the final z T Categorize relationships.

5. The method according to claim 1, characterized in that The step of constructing a dynamic time-series graph embedding includes using a quantum-inspired space-time graph embedding algorithm, which includes the following steps: a) Initialize entity e i and the relationship j The quantum bitmap embedding of: Q(e i )=|ψ i >=a i |0>+β i |1> Q(r j )=|φ j >=c j |0>+d j |1> among them,|a i | 2 +|b i | 2 =1,|γ j | 2 +|d j | 2 =1; b) Define the nonlinear space-time evolution operator U t : U t =exp(-iH t ) Among them, H t is a time-varying Hamiltonian operator: σ k is the Pauli matrix, w k (t) and J ij (t) is the time-varying coefficient; c) Introducing the quantum walk attention mechanism: A QW (i,j,t)=<ψ i (t)|U QW (t)|ψ j (t)> Among them, U QW (t) is a quantum walk operator based on graph structure; d) Multiscale time-dependent modeling using fractional quantum field theory: Φ(x,t)=∫K α (xy,ts)ψ(y,s)dyds Among them, K α is the fractional diffusion kernel, α∈(0,2) is the fractional parameter; e) Introduce the topological entropy regularization term: Among them, p ij is the transition probability from node i to node j; f) Design non-equilibrium thermodynamic optimization objectives: Among them, L task Loss of main task, is the rate of change of system entropy; g) Design of adaptive quantum measurement strategies: Among them, M is a learnable measurement operator.

6. The method according to claim 1, characterized in that In the step of performing knowledge reasoning and graph completion, a multi-granularity causal reasoning network is used, including the following steps: a) using a spectral clustering algorithm to divide the knowledge graph into multiple granularity levels; b) using a PC algorithm to discover potential causal relationships at each granularity level; c) using a structural equation model to estimate causal strength: X j =f j (PA j ,U j ) Among them, X j is the dependent variable, PA j For X j The parent node set, U j is an exogenous variable; d) Design an attention mechanism to integrate causal information of different granularities: α g =softmax(v T tanh(W g h g ))h final =∑ g α g h g Among them, α g is the attention weight, h g represents causal information of different granularities, v and W g is a learnable parameter; e) Based on the final representation h final Perform link prediction.

7. The method according to claim 1, characterized in that In the step of generating a precise traffic diversion strategy, a hierarchical spatiotemporal attention strategy network is used, including the following steps: a) Define state representation: s t =[u t ;p t ;c t ] Among them, u t is the user feature, p t is the product feature, c t is a contextual feature; b) Define a two-level action space: A l for advanced strategies) and A2 for specific operations; c) Determine spatiotemporal attention: α t =softmax(v T tanh(W s s t +W h h t-1 )) Among them, h t-1 is the historical hidden state, v,W s ,W h is a learnable parameter; d) Build a policy network: p θ (a|s)=softmax(f θ (s,a)) Among them, f θ is a parameterized neural network, e) Build a value network: in, is a parameterized neural network; f) Training using an asynchronous advantage actor-critic algorithm.

8. The method according to claim 1, characterized in that The step of optimizing the agency sales process includes using graph neural networks for agency matching and anomaly detection.

9. The method according to claim 1, characterized in that: In the step of effect evaluation and feedback, dynamic weight multi-objective Bayesian optimization is used, including the following steps: a) Define multiple objective functions: F(x)=[f1(x),f2(x),...,f m (x)] Among them, f i (x) is the i-th objective function, b) Establish a Gaussian process model for each objective function: F i (x)~GP(μ i (x),k i (x,x′)) Among them, μ i (x) is the mean function, k i (x,x′) is the kernel function; c) Determine dynamic weights: Among them, λ is the temperature parameter, f imax is the current optimal value of target i, σ i is the standard deviation; d) Calculation acquisition function: Among them, EI i (x) is the expected improvement of the ith objective; e) Use the DIRECT algorithm to maximize EI(x) and select the next evaluation point; f) Evaluate the new point, update the Gaussian process model and dynamic weights, and repeat steps ce.

10. A digital precision sales agent management system that integrates knowledge graphs, characterized by: include: Data preprocessing module, used to acquire and preprocess multi-source heterogeneous data; Data fusion module, used to realize multimodal data fusion and entity recognition; Relation extraction module, used for relationship extraction and attribute mapping; Graph embedding module, used to build dynamic time-series graph embeddings; Knowledge reasoning module, used to perform knowledge reasoning and graph completion; Update learning module, used for real-time update and incremental learning; Strategy generation module, used to generate precise traffic diversion strategies; Process optimization module, used to optimize the agency sales process; An evaluation and feedback module is used for effect evaluation and feedback; wherein the output of the evaluation and feedback module is connected to the input of the data preprocessing module to form a closed-loop optimization.

Images