Intelligent Cloud Data Acquisition and Dimensionality Reduction Customer Acquisition Method Based on Compressed Sensing

Abstract

The present invention belongs to the technical field of data processing, and specifically relates to an intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing. The method includes: a data acquisition step, a data analysis and processing step, a data fusion processing step, and a target user customer acquisition processing step; the data acquisition step is used to obtain user behavior data from multi-source data; the data analysis and processing step is used to calculate the similarity between users according to the temporal characteristics, and construct a behavior propagation equation based on the similarity between users according to the similarity between users; the data fusion processing step is used to perform feature fusion on the kernel tensor, temporal characteristics, and behavior propagation tensor to obtain fusion features; the target user customer acquisition processing step is used to predict the customer conversion rate of each user based on the fusion features. The present invention significantly improves the depth and accuracy of user behavior analysis, and provides an efficient and reliable solution for precise customer acquisition and user portrait construction.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and particularly relates to an intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing. Background Art

[0002] Driven by the big data era, user behavior analysis and precise customer acquisition have become core concerns in many industries, especially in the fields of e-commerce, finance, social media, etc. By deeply mining user behavior data, not only can an accurate user portrait be constructed, but also personalized recommendation and advertising placement strategies can be optimized, thereby improving business conversion rates. However, the multi-source heterogeneity, dimensional complexity, and time dynamics of user behavior data pose huge technical challenges to data analysis and target user screening. Currently, there are various technologies attempting to solve this problem, but there are still limitations, and more efficient and accurate analysis methods are urgently needed.

[0003] In the prior art, the core methods of user behavior analysis mainly include feature extraction methods based on matrix factorization, behavior prediction models driven by deep learning, and traditional similarity calculation and recommendation system algorithms. Among them, matrix factorization techniques (such as SVD and PCA) are commonly used for dimensionality reduction processing and main feature extraction of high-dimensional user behavior data. These methods have certain advantages in processing structured data and relatively low-dimensional user behavior data. However, with the diversification of data sources and the explosive growth of data dimensions, a single matrix factorization method faces the problem of "curse of dimensionality", that is, when dealing with multi-dimensional heterogeneous data, it is difficult for the matrix model to capture the complex correlations between data. In addition, traditional matrix methods have poor adaptability to sparse data. When there are a large number of missing values or high sparsity in the data, the performance of the model will decrease significantly, resulting in insufficient accuracy of behavior feature extraction. Deep learning models have been widely used in the field of user behavior prediction in recent years, such as time series analysis methods based on RNN and LSTM, feature extraction models based on CNN, and recommendation systems driven by multi-layer perceptrons (MLP). These models can capture complex non-linear relationships in user behavior data through deep feature learning. However, deep learning methods have extremely high requirements for data volume and computing resources. Especially when dealing with large-scale and multi-modal data, the training cost and real-time performance of model inference are difficult to meet the actual application requirements. In addition, deep learning models are usually regarded as "black box" models, and the interpretability of their results is poor, making it difficult to intuitively explain the relationship between user behavior features and model outputs, which brings troubles to decision-making support in business applications. Summary of the Invention

[0004] In view of this, the main object of the present invention is to provide a method for intelligent cloud data acquisition and dimensionality reduction for customer acquisition based on compressive sensing, so as to achieve efficient acquisition, dimensionality reduction processing and accurate analysis of multi-source heterogeneous data. By constructing a kernel tensor, temporal features and a behavior propagation equation, the present invention captures the global characteristics, temporal dynamics and group propagation laws of user behavior, generates fusion features with high expressive power, and combines the gamma function dynamic weights to achieve the sparsity control of user behavior and the conversion prediction of target customers. This method significantly improves the data processing efficiency, the depth and accuracy of user behavior analysis, and provides an efficient and reliable solution for accurate customer acquisition, personalized recommendation and user portrait construction.

[0005] The technical solution adopted by the present invention is as follows:

[0006] A method for intelligent cloud data acquisition and dimensionality reduction for customer acquisition based on compressive sensing, the method comprising: a data acquisition step, a data analysis and processing step, a data fusion processing step and a target user customer acquisition processing step; the data acquisition step is used to obtain user behavior data from multi-source data and represent the user behavior data as a multi-order tensor; the data analysis and processing step is used to perform tensor decomposition on the multi-order tensor to obtain a kernel tensor, construct temporal features based on the decomposed kernel tensor, calculate the similarity between users according to the temporal features, and construct a behavior propagation equation based on the similarity between users to obtain a behavior propagation tensor; the data fusion processing step is used to perform feature fusion on the kernel tensor, temporal features and behavior propagation tensor to obtain fusion features; the target user customer acquisition processing step is used to predict the customer conversion rate of each user based on the fusion features to obtain the conversion prediction rate of the user that can be converted into a target customer; compare the conversion prediction rate of each user with the set confidence level, and use the users with a conversion prediction rate higher than the confidence level as target potential customers.

[0007] Further, the multi-order tensor is represented by the following formula:

[0008]

[0009] wherein, the multi-order tensor represents the behavior feature value of user n in data source s, dimension d, and time t; N s represents the total number of sampling points in data source s; Ω is the sampling space, which represents the region for calculating the integral and is defined as the spatial range where the user and the sampling points are located; B d,s (t) is the behavior feature function of dimension d in data source s at time t, which is a quantity for describing the behavior feature and includes: click-through rate, purchase volume or number of visits; t i is the sampling time point; ∥x n -x i∥2 represents the user location x n and the Euclidean distance from the sampling point location x i ; v i (t) represents the feature vector of sampling point i at time t, reflecting the specific features of the sampling points of user behavior data, including: purchase records or browsing behaviors; σ t is the standard deviation of the Gaussian distribution.

[0010] Furthermore, in the data analysis and processing step, the multi-order tensor is decomposed using Tucker decomposition to obtain the core tensor, and the formula is as follows:

[0011]

[0012] where the core tensor satisfies the following sparse constraint conditions:

[0013]

[0014] where U N is the user modal basis matrix, U D is the dimension modal basis matrix, U T is the time modal basis matrix, U S is the data source modal basis matrix, all of which are preset matrices; k is the sparsity parameter; ∥·∥0 represents the L0 norm; ∥·∥ F represents the Frobenius norm; rank(·) is the operation of the rank of the matrix; represents the core tensor of user n, and when it is user i, it is expressed as

[0015] Furthermore, in the data analysis and processing step, based on the decomposed core tensor, the time series features are constructed through the following formula

[0016]

[0017] where tra represents the transpose operation of the matrix; τ is the time integration variable; tr(·) represents the operation of the trace of the matrix; W(t - τ) represents the time-weighted matrix at time t - τ; W(t) represents the time-weighted matrix at time t, which is preset; represents the time series features of user n, and when it is user i, it is expressed as represents the core tensor 's time gradient.

[0018] Furthermore, in the data analysis and processing step, through the following formula, the similarity between users is calculated based on the time series features, and based on the similarity between users

[0019]

[0020] Among them, <·,·> represents the tensor inner product operation; U N (:, i) represents the basis vector of user i; U N (:, j) represents the basis vector of user j; σ s is the similarity adjustment parameter and is a preset value; represents the temporal feature of user i; represents the temporal feature of user j; represents the similarity between user i and user j; Taking all the similarities between users as an element, a similarity matrix between users is formed is the element in the i-th row and j-th column of

[0021] Furthermore, in the data analysis and processing step, based on the following formula, a behavior propagation equation based on the similarity between users is constructed:

[0022]

[0023] Among them, is the Laplace operator; represents the behavior propagation tensor of user n. When it is user i, it is expressed as div(·) is the divergence operator.

[0024] Furthermore, in the data fusion processing step, based on the following formula, it is used to fuse the kernel tensor, temporal feature and behavior propagation tensor to obtain the fused feature

[0025]

[0026] Among them, represents the fused feature of user n. When it is user i, it is expressed as represents the time gradient of is the corresponding form under the time integration variable; is the corresponding form under the time integration variable; is the corresponding form under the time integration variable; is the tensor outer product operation.

[0027] Furthermore, in the target user customer acquisition processing step, based on the following formula, based on the fused feature, the customer conversion rate of each user is predicted to obtain the conversion prediction rate r that the user can be converted into a target customer i (t + Δt):

[0028]

[0029] Among them, r i (t + Δt) represents the conversion prediction rate of user i, which means the conversion prediction rate that this user can be converted into a target customer after the window time Δt starting from time t; is the corresponding form under the time integral variable; is the corresponding form under the time integral variable; is the corresponding form under the time integral variable; is the dynamic weight based on the gamma function.

[0030] Furthermore, the dynamic weight based on the gamma function is expressed by the following formula:

[0031]

[0032] Adopting the above technical solutions, the present invention has the following beneficial effects: First, the present invention realizes the efficient acquisition and dimensionality reduction processing of multi-source heterogeneous data through the compressive sensing theory, and solves the "curse of dimensionality" problem in high-dimensional data processing of traditional technologies. The core idea of compressive sensing is based on the sparsity assumption, and high-dimensional data is reconstructed through a small number of sampling points, thereby significantly reducing the data storage and calculation costs. In the present invention, user behavior data is modeled as a high-order tensor, and this multi-dimensional representation can capture the complex correlations of data in multiple dimensions such as time, space, and behavior categories. However, high-dimensional tensor data processing faces challenges such as high computational complexity and large memory occupancy. The present invention reduces the dimensionality of high-dimensional data through tensor decomposition technology, combined with the sparsity assumption of compressive sensing, and decomposes it into a core tensor and modal basis matrices, retaining the core information of the data and significantly reducing the data dimension. This processing method not only improves the integration efficiency of multi-source data, but also lays a high-quality data foundation for subsequent feature extraction and analysis. Second, the present invention demonstrates the ability to accurately capture the dynamic changes of user behavior in temporal modeling. User behavior has significant temporal dynamic characteristics, such as changes in purchase intent and shifts in interest hotspots. The present invention constructs temporal features to extract the changing patterns of user behavior in the time dimension. By combining the modeling of the time-weighted matrix and the time gradient, the present invention can sensitively capture the short-term fluctuations and long-term trends of user behavior, which is applicable to analyzing users' short-term response behaviors (such as purchase decisions in promotional activities) and can also reveal users' long-term interest changes. This deep modeling ability of temporal dynamics significantly improves the comprehensiveness and accuracy of user behavior analysis. Behavioral propagation modeling is another major technical highlight of the present invention, which shows excellent ability in the dynamic modeling of user-interaction relationships. Traditional user analysis methods usually only focus on individual behavioral characteristics and are difficult to capture the mutual influence between users. The present invention simulates the diffusion and propagation process of user behavior in a group by introducing a user similarity matrix and a behavioral propagation equation. The propagation model considers the synergistic effect between user similarity, behavioral dynamics, and global features, and accurately describes the propagation path of user behavior in time and space through the combination of the Laplace operator and the divergence operator. This non-linear propagation modeling method can not only identify the behavioral correlations between users, but also predict the potential direction of behavior diffusion, providing a scientific basis for the mining of target customers and the prediction of group behavior. Brief Description of the Drawings

[0033] Figure 1 It is a schematic flowchart of the method for intelligent cloud data acquisition and dimensionality reduction customer acquisition based on compressive sensing provided by an embodiment of the present invention. Detailed Embodiments

[0034] All features disclosed in this specification, or steps in all methods or processes disclosed, except for mutually exclusive features and / or steps, can be combined in any manner.

[0035] Any feature disclosed in this specification (including any additional claims, abstract) can be replaced by other equivalent or similar-purpose alternative features, unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only an example of a series of equivalent or similar features.

[0036] Example 1: Refer to Figure 1 , an intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing, the method comprising: a data acquisition step, a data analysis and processing step, a data fusion processing step, and a target user customer acquisition processing step; the data acquisition step is used to obtain user behavior data from multi-source data and represent the user behavior data as a multi-order tensor; the data analysis and processing step is used to perform tensor decomposition on the multi-order tensor to obtain a core tensor, construct time-series features based on the decomposed core tensor, calculate the similarity between users according to the time-series features, and construct a behavior propagation equation based on the similarity between users to obtain a behavior propagation tensor; the data fusion processing step is used to perform feature fusion on the core tensor, time-series features, and behavior propagation tensor to obtain fusion features; the target user customer acquisition processing step is used to predict the customer conversion rate of each user based on the fusion features to obtain the conversion prediction rate of the user who can be converted into a target customer; compare the conversion prediction rate of each user with the set confidence level, and use the users with a conversion prediction rate higher than the confidence level as target potential customers.

[0037] Specifically, compressive sensing is a breakthrough data processing theory. Its core idea is to reconstruct complete data through a small number of random samplings on the premise that the data is sparse or compressible. In the present invention, this theory is widely applied to the data acquisition and dimensionality reduction processes. Specifically, in the data acquisition step of the present invention, user behavior information is obtained through the integration of multi-source data. This data may be derived from the multi-platform interaction behaviors of users, including but not limited to e-commerce browsing records, social media operation logs, and sensor data of mobile devices. When dealing with such data, traditional methods often face problems such as large storage pressure, low transmission efficiency, and information loss. However, the present invention optimizes the data acquisition process through compressive sensing technology and achieves accurate reconstruction of the original data using a small number of sampling points. The core advantage of compressive sensing is that it is based on random linear measurements, maps the compression of high-dimensional signals to a low-dimensional space, and restores the data through a sparsity reconstruction algorithm in subsequent stages. This process not only reduces the storage and computing burdens of the cloud computing platform but also lays a data quality foundation for subsequent analysis and processing. After obtaining the user behavior data, the present invention performs a structured representation of the multi-dimensional data through tensor modeling. Different from the traditional matrix or table representation forms, a tensor is a data structure in the form of a multi-dimensional array that can simultaneously express multi-dimensional information such as the temporal characteristics of user behavior, behavior categories, and user attributes. In the present invention, the user behavior data is modeled as a multi-order tensor, and each dimension represents different characteristics of the data. For example, a typical third-order tensor may include a user dimension, a behavior type dimension, and a time series dimension. This modeling method can not only retain the complex relationships between multi-source data but also achieve dimensionality reduction of the data through tensor decomposition technology. Tensor decomposition, as an efficient dimensionality reduction method, is an important technical support for the data analysis part of the present invention. By decomposing the multi-order tensor, the present invention extracts the core tensor, and this core feature representation retains the main information of the original data and significantly reduces the dimension of the data. Compared with traditional dimensionality reduction methods such as principal component analysis (PCA), tensor decomposition has more advantages in dealing with multi-dimensional data, and it can more comprehensively capture the multi-dimensional structural characteristics of the data.

[0038] Based on the nuclear tensor, the present invention further analyzes the temporal characteristics of users. The temporal characteristics of user behavior are an important basis for evaluating user activity and interest changes, which reflect the operation frequency and behavior patterns of users at different time periods. For example, some users may show high activity during specific time periods, which may be related to their shopping habits, work arrangements, or interest hotspots. By extracting these temporal characteristics, the present invention provides key inputs for user similarity calculation and behavior propagation modeling. Based on the temporal characteristics of users, the present invention calculates the similarity between users, which is not limited to the matching of static attributes but also comprehensively considers the dynamic similarity in behavior patterns. The calculation of user similarity lays the foundation for behavior propagation analysis, and the behavior propagation equation is a key tool for simulating the user behavior propagation process in the present invention. The design of the behavior propagation equation is another innovative point of the present invention. The behavior propagation between users is similar to information diffusion or influence propagation in social networks, which describes how the behavior of a certain user spreads to other users through social connections or interest similarities. In the present invention, the behavior propagation equation constructs a mathematical model for describing user behavior propagation by combining user similarity and behavior dynamics. This model not only considers the behavior characteristics of individual users but also integrates the propagation laws between groups. For example, when the behavior characteristics of user A are highly similar to those of user B, user B may be significantly influenced by the behavior of user A. This process is manifested as the construction of a behavior propagation tensor in a multi-user environment, which reflects the diffusion pattern and influence range of user behavior in the group.

[0039] After the data analysis and processing are completed, the present invention enters the data fusion stage. The core task of this stage is to fuse the nuclear tensor, temporal features, and behavior propagation tensor to generate more predictive fusion features. The significance of feature fusion lies in comprehensively utilizing information from multiple sources to avoid information loss or analysis bias that may be brought about by a single feature. To achieve this goal, the present invention adopts a deep learning embedding method to map various features into a unified feature space. The construction of this feature space not only retains the key information of various features but also extracts the implicit relationships between features through the optimization training of the deep learning network. For example, the nuclear tensor may reflect the overall characteristics of users, while the temporal features reflect the dynamic changes in user behavior, and the behavior propagation tensor describes the interaction relationships between users. The fusion of these features can generate a more comprehensive user profile, providing high-quality input for subsequent user prediction. In the final target user prediction stage, the present invention predicts the customer conversion rate of users based on the fusion features. Customer conversion rate prediction is a key step in the customer acquisition process, which directly affects the accuracy of target customer screening and the customer acquisition efficiency. The present invention adopts a prediction model based on deep learning. This model generates the conversion prediction rate of each user through deep learning of the fusion features. This prediction process not only considers the characteristics of individual users but also combines the propagation laws of group behavior to generate more reliable prediction results. To further improve the stability and credibility of the prediction, the present invention introduces a confidence screening mechanism, that is, compares the conversion prediction rate of each user with a preset confidence threshold, and only selects users higher than the threshold as target potential customers. The confidence screening mechanism effectively avoids the problem of error accumulation in prediction and ensures the accuracy of target customer screening.

[0040] Embodiment 2: Multi-order Tensor It is represented by the following formula:

[0041]

[0042] Among them, the multi-order tensor represents the behavior feature value of user n in data source s at dimension d and time t; N s represents the total number of sampling points in data source s; Ω is the sampling space, representing the region for calculating the integral, defined as the spatial range where users and sampling points are located; B d,s (t) is the behavior feature function of dimension d in data source s at time t, which is a quantity describing behavior features, including: click-through rate, purchase volume, or number of visits; t i is the sampling time point; ∥x n -x i ∥2 represents the Euclidean distance between user position x n and sampling point position x i ; vi (t) represents the feature vector of sampling point i at time t, reflecting the specific features of the sampling points of user behavior data, including: purchase records or browsing behaviors; σ t is the standard deviation of the Gaussian distribution.

[0043] Specifically, in Embodiment 2 of the present invention, the formula constructs a multi-order tensor X n,d,t,s to describe in detail the dynamic characteristics of user behavior features in the spatial, temporal, and behavioral dimensions in a multi-source data scenario. This formula embodies the core idea of the compressive sensing theory, that is, under the sparsity assumption, high-dimensional behavior feature representations are reconstructed through finite data sampling, while combining the spatio-temporal correlation of user behavior and the complexity of multi-dimensional feature interactions, thereby providing strong mathematical support for the accurate prediction of target users. In the formula, the tensor X n,d,t,s defines the behavior feature value of user n in data source s under dimension d and time t. Its essence is a structured data representation in the form of a multi-dimensional array. By comprehensively modeling user behavior and fully considering the complex relationships of multi-source data, especially the dynamic features in time and space, the details often overlooked in traditional data analysis methods can be quantified and expressed. The construction of the formula connects user behavior with the spatial environment in which it is located in the form of an integral. The integral region Ω represents the spatial range where the user and the sampling points are located. Through this integral operation, the formula can comprehensively sample the features of the sampling points globally, thereby capturing the relationship between the user and the surrounding environment. For example, a user's shopping behavior may not only be affected by their own characteristics but also by the potential transmission effect of the behaviors of other nearby users. By weighted summarizing the behavior features of the sampling points within the integral region, the formula can construct a globalized user behavior feature representation, and the construction of this feature depends on multiple important variables in the formula.

[0044] The Euclidean distance ∥x n -x i∥2 is used to characterize the spatial relationship between the user's location and the sampling points, reflecting the spatial dependence of user behavior. This distance factor introduces a natural weighting mechanism, that is, sampling points closer to the user contribute more to their behavioral characteristics, while the influence of sampling points farther away gradually weakens. This weight assignment conforms to the propagation law of user behavior in reality. For example, a user may be more easily influenced by promotional activities at nearby shopping points, while rarely participating in activities at a farther distance. Through the weighted calculation of Euclidean distance, the formula reflects spatial sparsity in behavior modeling, avoiding over-reliance on the behavioral characteristics of irrelevant sampling points, while reducing the computational complexity and providing a guarantee of computational efficiency for subsequent data processing. The dynamic change in time is reflected through the Gaussian decay function in the formula. This function describes the correlation of user behavior in the time dimension, that is, the influence of the behavior of sampling points gradually weakens over time. This Gaussian decay mechanism controls the diffusion range of time through the parameter σ t and can accurately capture the time law of user behavior. For example, in the e-commerce scenario, the purchase behavior of users may be strongly influenced by short-term promotional activities, and this influence gradually fades after the activity ends. By introducing the Gaussian distribution, the formula takes into account the temporal dynamics and sparsity characteristics of user behavior in time modeling, enabling the behavior propagation to more realistically reflect the actual behavior pattern of users. This time dynamic modeling not only enhances the expressive ability of multi-order tensors but also lays a foundation for the subsequent calculation of user similarity and behavior propagation modeling.

[0045] In addition, the formula also captures the characteristics of sampling points and data sources through the eigenvector v i (t) and the eigenfunction B d,s (t), further enriching the semantic information of the tensor representation. The eigenvector v i (t) describes the behavioral characteristics of sampling points at a specific time, such as purchase records or browsing behaviors, and its role is to assign a unique behavioral identifier to each sampling point. The eigenfunction B d,s(t) describes the global characteristics of behavior from the perspectives of dimension and data source, such as click-through rate, traffic volume, etc. The introduction of this function realizes the unified modeling of different data sources and behavior dimensions, facilitating the fusion of multi-source data. In the formula, these two features are fused with user behavior features through the interaction of time and space, thus constructing a more comprehensive user behavior expression. This user behavior modeling method based on multi-order tensors has significant innovation and advantages compared with traditional behavior analysis techniques. Traditional methods usually only consider user behavior in a single dimension or at a single time point, ignoring the dynamic characteristics of user behavior in multi-dimensional space. However, through tensor modeling and formula construction, the present invention realizes the comprehensive quantification of user behavior in the dimensions of time, space, and data source. Especially under the guidance of the compressed sensing theory, the present invention utilizes the sparsity characteristics of data to reconstruct a high-dimensional tensor representation through the behavior of finite sampling points, greatly reducing the complexity of data acquisition and storage. In addition, through the distance factor and Gaussian decay function in the formula, the present invention effectively combines the spatial and temporal dependencies of user behavior, making the behavior modeling more in line with the actual scenario and providing more accurate and reliable input data for subsequent user behavior prediction.

[0046] Example 3: In the data analysis and processing step, the multi-order tensor is decomposed using Tucker decomposition to obtain a core tensor. The formula is as follows:

[0047]

[0048] Among them, the core tensor satisfies the following sparse constraint conditions:

[0049]

[0050] Among them, U N is the user modality basis matrix, U D is the dimension modality basis matrix, U T is the time modality basis matrix, U S is the data source modality basis matrix, all of which are preset matrices; k is the sparsity parameter; ∥·∥0 represents the L0 norm; ∥·∥ F represents the Frobenius norm; rank(·) is the operation of the rank of the matrix; represents the core tensor of user n, and when it is user i, it is represented as

[0051] Specifically, Tucker decomposition is a classical tensor decomposition method. By decomposing high-dimensional data into the product of a low-dimensional core tensor and several modal basis matrices, it effectively extracts the core feature information of the data while significantly reducing the data dimension, thus providing a more concise and efficient data representation form for subsequent user behavior analysis and target user screening. The core idea of this formula is to represent the original multi-order tensor as a linear mapping of each modal basis matrix on its corresponding mode, thereby constructing a sparse low-rank core tensor. Through the mapping of the basis matrices representing the user mode, dimension mode, time mode, and data source mode respectively, the noise, redundant information, and low-correlation data in the original tensor are filtered out, and what remains are the core features of the data. In actual operation, the core tensor is designed to satisfy the sparse constraint condition, and its sparsity is controlled by the parameter k. This design of the sparse constraint is based on the core assumption of the compressed sensing theory, that is, user behavior data is sparse in the multi-dimensional space, and only a few features have a significant impact on user behavior. Through this constraint condition, the core tensor can not only compress high-dimensional data but also retain important information during the dimension reduction process, eliminating irrelevant or noisy data to ensure the accuracy and effectiveness of the data. Especially when user behavior involves multi-source heterogeneous data, this constraint plays an important role in reducing the curse of dimensionality and improving data processing efficiency. Further, the ∥x n,d,t,s ∥ and rank(x n,d,t,s ) in the sparse constraint formula measure the data intensity and the rank size of the tensor respectively. The data intensity describes the activity level of user behavior in a specific scenario, such as the number of purchases, browsing depth, etc.; while the rank of the tensor represents the degree of linear independence of the data. By combining these two factors, the sparse constraint dynamically adjusts the sparsity of the core tensor, enabling it to adapt to the complexity of the data and effectively control the dimension reduction effect of the data. This dynamic adjustment mechanism enables the present invention to be flexibly adapted in different data scenarios. Whether it is the cold start scenario with strong behavior sparsity or the large-scale recommendation system environment with complex data dimensions, excellent performance can be achieved. The intelligent cloud data acquisition and dimension reduction customer acquisition method based on compressed sensing realizes the efficient compression and sparse representation of data through Tucker decomposition, thereby greatly reducing the computational complexity of multi-source data processing. At the same time, as the low-dimensional core representation of behavior features, the core tensor provides more reliable and accurate data support for subsequent behavior propagation modeling and target user conversion prediction. Compared with traditional tensor decomposition methods, the present invention combines the sparsity advantage of the compressed sensing theory with the dimension reduction ability of Tucker decomposition by introducing a sparse constraint, solving the problems of noise interference and redundancy in the analysis of high-dimensional data, and fundamentally improving the utilization efficiency and feature expression ability of multi-source heterogeneous data.

[0052] Example 4: In the data analysis and processing step, based on the decomposed core tensor, the time series features are constructed through the following formula

[0053]

[0054] Among them, tra represents the transpose operation of the matrix; τ is the time integration variable; tr(·) represents the operation of the trace of the matrix; W(t - τ) represents the time-weighted matrix at time t - τ; W(t) represents the time-weighted matrix at time t, which is preset. represents the temporal feature of user n. When it is user i, it is represented as represents the kernel tensor of the time gradient.

[0055] Specifically, the core of the formula is to use the kernel tensor to describe the main features of user behavior in the multi-dimensional space, and at the same time, adjust the influence of behavior over time through the time-weighted matrix W(t - τ). The time-weighted matrix is essentially a preset function used to assign weights to behaviors at different time points, and this weight changes dynamically with the time difference. For example, when a certain behavior of a user occurs at a time point closer to the current time t, the weighted matrix will assign it a greater weight, reflecting its stronger influence on the current moment; while for a time point farther away, its influence gradually decays, thus conforming to the time-dependent characteristics of actual user behavior. Such a time-weighted design not only enables the formula to have a flexible time adjustment ability but also reduces the interference of long-term behavior noise on feature modeling to a certain extent, thereby enhancing the ability to capture short-term behavior trends. This is of great significance in the intelligent customer acquisition scenario. For example, by analyzing the recent access frequency and behavior intensity of users, their purchase intention can be predicted more accurately. To further refine the analysis of user behavior changes, the formula introduces the time gradient of the kernel tensor and measures the change intensity in the time dimension through its Frobenius norm. The time gradient is a key indicator describing the change rate of the kernel tensor in the time dimension and can sensitively reflect the sudden change or stability of user behavior over time. For example, when a user suddenly increases the frequency of a certain specific behavior (such as browsing a certain type of product or participating in a certain type of activity) within a period of time, the time gradient will show a large value, indicating a significant change in behavior; on the contrary, when the behavior remains stable, the value of the time gradient will be small. By introducing this gradient into the formula, the temporal feature can capture the change trend of user behavior and dynamically adjust the influence of behavior according to the change amplitude. This design effectively solves the problem that traditional temporal analysis methods cannot take into account both the smoothness and sudden changes of behavior, enabling the modeling results to not only describe the long-term stable trend of users but also focus on short-term significant behavior changes.

[0056] On this basis, the formula further controls the time gradient through an exponential decay function to ensure the robustness and sparsity of the temporal features. The design of the exponential decay function makes full use of the sparsity assumption in the compressed sensing theory, that is, the significant changes in user behavior often concentrate on a few key time points, while the behavior in most time periods is relatively stable. Specifically, when the time gradient is large, the exponential decay factor will significantly decrease, weakening the impact of mutation behavior on the overall temporal features to avoid the interference of abnormal fluctuations; while when the time gradient is small, the exponential decay factor approaches 1, allowing stable behavior to make a greater contribution to the temporal features. Through this mechanism, the formula can capture behavior changes while suppressing excessive behavior fluctuations, making the constructed temporal features more stable and reliable. In addition, the trace operation of the time modal basis matrix is introduced into the denominator of the exponential decay to measure the feature complexity in the time dimension, thereby further optimizing the adjustment range of the time gradient. As an important property of the matrix, the trace operation reflects the overall characteristics of the kernel tensor in the time dimension, providing a mathematical basis for the optimization of the formula in time dynamic control.

[0057] Through the decomposition and sparsity control of the kernel tensor, the construction of temporal features realizes data compression and dimensionality reduction, providing lightweight inputs for subsequent behavior propagation modeling and conversion rate prediction. At the same time, the combination of the time weighting matrix and the time gradient enables the temporal features to more finely depict the evolution law of user behavior over time, providing support for the dynamic update of user portraits. This dynamic modeling method is particularly suitable for real-time analysis requirements in large-scale data environments, such as real-time capturing of user purchase intentions on e-commerce platforms or monitoring of user interaction trends on social media platforms. In addition, through the sparsity constraints of exponential decay and trace operation, the formula effectively reduces the interference of noise on the modeling results, making the calculation of temporal features not only more efficient but also more robust, and capable of adapting to complex analysis requirements in multi-source heterogeneous data scenarios. Compared with traditional static feature analysis methods, temporal features can comprehensively reflect the dynamic changes of user behavior over time, providing richer information for the accurate prediction of target users. For example, in the customer acquisition scenario, by analyzing the behavior change trends of potential customers in the short term and combining the modeling results of temporal features, it is possible to more accurately judge the purchase intentions and conversion possibilities of customers. Especially in an environment with a large amount of data and complex behavior patterns, this formula significantly improves data processing efficiency and analysis accuracy through the sparsity assumption of compressed sensing, providing theoretical support and technical guarantee for realizing intelligent user analysis and accurate customer acquisition.

[0058] Example 5: In the data analysis and processing step, the similarity between users is calculated according to the following formula based on the temporal features, and according to the similarity between users

[0059]

[0060] Among them, <·,·> represents the tensor inner product operation; U N (:, i) represents the basis vector of user i; U N (:, j) represents the basis vector of user j; σ s is the similarity adjustment parameter, which is a preset value; represents the time series feature of user i; represents the time series feature of user j; represents the similarity between user i and user j; taking all the similarities between users as an element, a similarity matrix between users is formed is the element in the i-th row and j-th column of

[0061] Specifically, in the formula, the similarity of user time series features is obtained through the normalization calculation of the tensor inner product and the Frobenius norm. The time series features and respectively describe the behavior dynamics of user i and user j at time t. Their inner product reflects the consistency in the direction of behavior changes of the two users, while the norm normalization eliminates the influence of feature magnitudes, making the calculation result focus on the similarity of behavior patterns. This similarity calculation method reveals the degree of coincidence of the behavior trends of two users in a certain time period by comparing the dynamic features of user behaviors. For example, when the purchase behaviors, browsing frequencies, or interaction patterns of users i and j have the same change trend at time t, the similarity of the time series features will tend to be a higher value, indicating that they may have similar interests or needs. On the other hand, the formula further considers the relative position relationship of users in the basis vector space by introducing the Euclidean distance of user basis vectors. The basis vector describes the global features of the user, such as long-term behavior preferences or basic attribute differences, and the squared value of the distance quantifies the feature differences between the two users. By combining this distance information in the exponential decay function, the formula can dynamically adjust the influence of the basis vector difference on the similarity. Specifically, when the basis vectors of two users are close, the similarity of their behavior features will be amplified; while when the basis vector differences of users are large, the similarity will rapidly decrease as the distance increases. This design theoretically integrates the dual information of time series dynamic features and global static features, providing a more comprehensive perspective for modeling the multi-dimensional similarity of user behaviors.

[0062] In the formula, σ S as the similarity adjustment parameter is used to control the sensitivity of the basis vector difference to the similarity. A larger σ S value will make the influence of the basis vector difference weaker, that is, it pays more attention to the time series dynamic similarity of users; while a smaller σ SThe value will emphasize the global differences of the basis vectors, thus being more inclined to screen out users who have a high degree of similarity in long-term behavior patterns. The setting of this parameter has strong flexibility and can be optimized and adjusted according to specific application scenarios. For example, in short-term marketing activities, more attention may be paid to the recent behavior characteristics of users, and at this time, a larger σ S value needs to be set; while in the analysis of users' long-term loyalty, a smaller σ S value may be required to highlight the global characteristics. Finally, the similarity between all users is integrated into a user similarity matrix where each element represents the behavioral similarity between user i and user j at time t. This matrix not only intuitively shows the mutual relationships between all users, but also provides key inputs for behavioral diffusion modeling. For example, a group of users with high similarity in the matrix may represent potential customer groups with similar interests or needs, while users with low similarity may reflect personalized preferences and behavioral characteristics. This matrix representation based on the similarity between users can well adapt to subsequent behavioral diffusion modeling and user clustering analysis, providing an efficient data structure for accurate customer acquisition. Through the calculation formula of the similarity between users, the dynamic characteristics, static characteristics and time change trends of multi-dimensional data are effectively integrated, breaking through the limitation of traditional similarity analysis that only focuses on a single dimension. This formula not only combines temporal features and global features theoretically, but also significantly improves the accuracy and robustness of similarity analysis in practice. For example, compared with simple cosine similarity or Euclidean distance, this formula captures the short-term behavioral trends of users through temporal features, and at the same time combines the differences of basis vectors to reflect the long-term characteristics of users, making the calculation results of user similarity closer to the real needs. Especially in a large-scale data environment, kernel tensor dimensionality reduction based on compressed sensing theory greatly reduces the computational complexity, making the generation of the user similarity matrix both efficient and accurate.

[0063] Example 6: In the data analysis and processing step, the following formula is used to construct a behavioral diffusion equation based on the similarity between users:

[0064]

[0065] where is the Laplace operator; represents the behavioral diffusion tensor of user n, and when it is user i, it is expressed as div(·) is the divergence operator.

[0066] Specifically, the first part of the formula describes the diffusion characteristics of the behavior between users through the Laplace operator . The diffusion process here is determined by the user similarity matrix and the temporal features Collective decision-making. The user similarity matrix reflects the correlation of behavioral patterns among users, while the temporal features describe the dynamic characteristics of user behavior over time. The role of the Laplacian operator here is to measure the local rate of change in behavior between users and simulate the spread of behavior among similar users. For example, when users i and j have a high similarity in the similarity matrix, the diffusion rate of behavior propagation will increase significantly, thereby accelerating the spread of this behavior among users with high similarity. In contrast, for user groups with low similarity, the diffusion rate will be naturally inhibited. This diffusion mechanism can well simulate the law of user behavior propagation in social networks, such as the rapid diffusion of popular product recommendations or promotional activity information among similar user groups. In addition, the diffusion term also adjusts the intensity of diffusion through the Frobenius norm of the modal basis matrix, enabling it to adapt to the propagation requirements of different time scales and behavioral characteristics. The second part of the formula is through the kernel tensor and the time gradient of the temporal features to establish the interaction term of behavior propagation. This interaction term physically describes how the global characteristics of user behavior affect its local temporal dynamics. The kernel tensor is the core representation of multi-modal features. It extracts the main features of user behavior through compressive sensing and tensor decomposition, representing the global characteristics of users in multiple dimensions. The time gradient reflects the rate of change of user behavior over time. For example, when a user's recent purchase behavior suddenly increases or decreases, its time gradient will change significantly. Through the combination of these two, the interaction term can describe how global characteristics regulate local temporal changes. For example, a user's long-term interest preference may determine the direction of change in their short-term behavior, and this global-local correlation is mathematically expressed in the interaction term. At the same time, this term adjusts the interaction intensity through the ratio of the basis matrices to ensure the adaptability of behavior propagation under different user characteristic dimensions. The third part of the formula uses the divergence operator to describe the change in the gradient distribution of the kernel tensor under similarity regulation. The role of the divergence operator here is to characterize the fluidity of behavior propagation in the spatial dimension, that is, how user similarity affects the path and scope of behavior propagation. For example, in a user group with high similarity, behavioral features will spread among users with a higher frequency, while in a user group with low similarity, this spread will be significantly inhibited. Through this divergence term, the formula can simulate the direction and intensity of user behavior flow in space, making the modeling of behavior propagation not only limited to the time dimension but also enabling a more accurate description in space. This term also regulates the divergence process through the ratio of the basis matrices to adapt to the propagation requirements in different data source scenarios.

[0067] Embodiment 7: The data fusion processing step is used to fuse the kernel tensor, temporal features, and behavior propagation tensor through the following formula to obtain fused features

[0068]

[0069] where represents the fused feature of user n, and when it is user i, it is represented as represents the time gradient of is the corresponding form of under the time integration variable; is the corresponding form of under the time integration variable; is the corresponding form of

[0070] Specifically, the core idea of the formula is to use the tensor outer product to jointly represent multiple features, thereby constructing a high-dimensional feature tensor The introduction of the outer product operation enables the multi-modal global characteristics represented by the kernel tensor to be integrated simultaneously with the dynamic changes of the temporal features and the behavior diffusion information of the propagation tensor Through this joint representation, the formula can capture the interaction relationships of user behavior in multiple dimensions. For example, the kernel tensor can reflect the long-term preferences of users, the temporal features describe the behavior changes of users in a specific time period, while the propagation tensor reveals the diffusion law of user behavior in the group. The outer product combination of the three unifies this information in a high-dimensional feature space, enabling the fused features to not only contain the independent characteristics of each modality but also reflect the complex interactions between modalities. The time integration operation in the formula further enhances the dynamics of the fused features. By integrating over the time variable τ, the formula can globally summarize the change trends of user behavior in the time dimension, thereby generating a fused representation that comprehensively reflects the full-cycle characteristics of user behavior. For example, for a user who has significant behavior changes in a certain time period, time integration can accurately capture the cumulative impact of these changes, while for a user with relatively stable behavior, it can more prominently highlight the stability of their long-term characteristics. In addition, the integration process adjusts the influence of behavior over time through the introduction of dynamic weights, making the contribution of recent behavior to the fused features greater, while the influence of behavior at more distant time points gradually decays. This dynamic weighting mechanism ensures the timeliness and pertinence of the fused features.

[0071] Time gradient plays the role of sparsity constraint in the formula. By calculating the time gradient of the behavior propagation tensor , the formula can measure the change rate of user behavior propagation in the time dimension. When the time gradient is large, it indicates that there are rapid changes in the behavior propagation process. At this time, the exponential decay factor will be significantly reduced to suppress the interference of violent fluctuations on the fused features; while when the time gradient is small, the exponential decay factor tends to 1, allowing the stable propagation process to make a greater contribution to the fused features. This design is based on the sparsity assumption of the compressive sensing theory, that is, the significant changes in user behavior usually concentrate on a few key time points, and the propagation process is relatively stable in most time periods. Through the sparsity control of the time gradient, the formula effectively avoids the negative impact of abnormal data or violent fluctuations on the fused features, making the generated features more robust. In addition, the formula adjusts the amplitude of exponential decay through the normalization factor to balance the weight relationship between the behavior propagation gradient and the dynamics of temporal features. The normalization factor reflects the overall change intensity of the temporal features, making the adjustment of exponential decay more adaptive. For example, when the temporal features of users change greatly, the normalization factor can amplify the inhibitory effect of the time gradient, so as to highlight the contribution of important behavior changes to the fused features; while when the temporal features change little, the normalization factor weakens the influence of the time gradient, so as to maintain the stability of the fused features.

[0072] Fused features The construction of provides a comprehensive and efficient user behavior representation for accurate customer acquisition. By dynamically fusing the kernel tensor, temporal features, and behavior propagation tensor in the time dimension, the formula can not only capture the behavior characteristics of users in different modalities but also reflect the evolution law of behavior in time and space. This multi-dimensional and multi-modal feature representation provides an important basis for the construction of user portraits and the prediction of customer conversion rates. For example, in the e-commerce scenario, the fused features can be used to analyze the long-term preferences and recent behavior changes of users to help predict their response to specific promotional activities; on social media platforms, the fused features can be used to identify user groups with similar propagation laws, so as to optimize information push and advertising strategies. Compared with traditional feature fusion methods, the formula of the present invention realizes the organic unity of global features, temporal features, and behavior propagation information through the combination of tensor outer product and time dynamics. Especially through the sparsity constraint and dynamic weighting mechanism of the time gradient, the formula avoids the interference of noise data while capturing significant changes in user behavior, making the fused features perform more stably and accurately in complex data environments. This design has significant technical advantages in intelligent cloud data collection and analysis, not only can it adapt to the processing requirements of multi-source heterogeneous data, but also provides strong technical support for target user screening and behavior prediction in accurate customer acquisition.

[0073] Example 8: In the target user customer acquisition processing step, the following formula is used to predict the customer conversion rate of each user based on the fusion features, and obtain the conversion prediction rate r at which the user can be converted into a target customer i (t + Δt):

[0074]

[0075] where r i (t + Δt) represents the conversion prediction rate of user i, indicating the conversion prediction rate at which the user can be converted into a target customer after a window time Δt starting from time t; is the corresponding form under the time integral variable; is the corresponding form under the time integral variable; is the corresponding form under the time integral variable; is the dynamic weight based on the gamma function.

[0076] Specifically, in the formula is the core probability control part, used to calculate the ratio of the fusion feature F i (τ) to the behavior propagation tensor P i (τ), and map it to the probability space of [0, 1]. The fusion feature F i (τ) synthesizes the global features, temporal dynamics, and propagation laws of users, while the behavior propagation tensor P i (τ) describes the propagation range and intensity of user behavior in the group. The ratio of the two reflects the balance between the independence and propagation of user behavior characteristics. For example, when the propagation range of user behavior is large (high P i (τ)) but the fusion feature is small (low F i (τ)), the conversion rate may be low, indicating that user behavior may be more driven by group effects; conversely, a higher fusion feature value indicates stronger independence of user individual behavior and may also result in a higher conversion rate. By using the sigmoid function to perform a non-linear mapping on the ratio, the formula can smoothly handle the influence of different feature combinations on the conversion rate. Secondly, the introduction of the gamma function weight endows the formula with dynamic regulation ability. The form of the gamma function combines the temporal feature gradient and the proportional relationship of the fusion feature, reflecting the influence of the user behavior change rate on the conversion rate. When the user's behavior changes rapidly (high ) When this happens, the gamma function weights are adjusted accordingly to highlight the importance of these changes in conversion rate prediction; while when the behavior changes are relatively stable, the weights tend to be smoothed to reduce the interference of unnecessary fluctuations on the prediction results. In addition, the gamma function parameter k, as a control factor, can be flexibly adjusted according to the application scenario. For example, in a high-dynamic environment, increasing the value of k can enhance the influence of behavior changes, or in a stable scenario, decreasing the value of k can highlight the long-term behavior characteristics. The exponential decay part in the formula further smooths the influence of the time gradient on the conversion rate. Through 's design, when the gradient of the temporal feature is large, the decay factor decreases, thus limiting the amplification effect of abnormal behavior changes on the prediction results; while when the change of the temporal feature is small, the decay factor approaches 1, making the contribution of the behavior stability to the prediction results greater. This mechanism adjusts the balance between the behavior dynamics and the characteristics of the data source through the matrix trace . For example, for the user behavior in a complex data source, the matrix trace can dynamically adjust the amplitude of the decay factor to ensure the adaptability of the formula to different data sources. The design of this conversion rate prediction formula comprehensively integrates the individual characteristics of users and the group behavior patterns, providing a scientific basis for accurate customer acquisition. By capturing the dynamic changes of user behavior within the window period through time integration, the formula can predict the probability that users will convert into target customers in the future. For example, on an e-commerce platform, it can analyze the recent browsing, purchasing behaviors of users and their dissemination effects to accurately evaluate their likelihood of participating in promotional activities; in a social network, it can predict the response degree of users to information dissemination through behavior propagation and group effects. Compared with traditional static prediction methods, this formula has significant advantages in dynamic behavior modeling and multi-modal feature fusion.

[0077] Example 9: Dynamic Weight Based on Gamma Function It is represented by the following formula:

[0078]

[0079] Specifically, the formula for the dynamic weight of the gamma function depends on the lower integration limit, which is determined by the ratio of the time gradient of the user behavior to the intensity of the fusion feature, that is This quantitative index comprehensively reflects the significance of the change in user behavior and the stability of the overall characteristics. For users with large behavioral changes, their time gradient values are high, which makes the decay rate of the gamma function weights relatively fast, thereby highlighting the contribution of these behavioral changes in the weight distribution; while for users with relatively stable behavior, the decay of the gamma function weights is slow, so that the weight distribution more evenly reflects the long-term stability of their behavior. The core idea of this dynamic adjustment of weight distribution is highly consistent with the sparsity assumption of the compressed sensing theory in the present invention, that is, by highlighting a few significant behavioral change points, reducing the influence of unimportant features, and ultimately achieving accurate modeling of the target behavior. The role of the sparsity parameter k in the gamma function further enhances the flexibility and adaptability of sparsity regulation. When k>1, the gamma function pays more attention to the sparsity of the weight distribution, that is, by emphasizing the contribution of significant behavioral change points, suppressing the influence of minor behavioral characteristics. This design is particularly applicable in fast-response business scenarios, such as user conversion prediction during marketing activities. In this scenario, it is necessary to quickly identify the target users who respond most actively to the activity, and a larger k value can ensure that the model preferentially captures these user characteristics with significant behavior. When k≤1, the weight distribution of the gamma function tends to be smoother, focusing more on the overall trend of behavior, which is more advantageous when analyzing users' long-term behavior patterns or loyalty. For example, in the member analysis of an e-commerce platform, a smaller k value can highlight the user's continuous behavioral characteristics and provide a more reliable basis for user stratification and long-term value prediction. The integral form of the gamma function further reflects the important role of time gradient in dynamic weight distribution. Time gradient It is a key indicator for measuring the rate of change of user behavior. Its value directly affects the setting of the lower limit of the gamma function integral, which has a significant impact on the dynamic weight distribution. When the time gradient of user behavior is large, it indicates that their behavior has changed rapidly in a short period of time. At this time, the gamma function weight will automatically concentrate on these time points, highlighting the importance of these rapid changes in conversion prediction; for time periods with slower behavior changes, the gamma function will assign lower weights to avoid the interference of stable behavior on the overall prediction. This nonlinear weight distribution mechanism enables the model to more accurately capture key changes in user behavior, thereby showing higher accuracy and flexibility in target user screening.

[0080] The introduction of the gamma function's dynamic weights not only enhances the ability to model sparse features but also plays a core role in the temporal evolution analysis of dynamic features. Compared with traditional static weight methods, the gamma function achieves the dynamic adaptation ability of the feature weight distribution through the joint regulation of the sparsity parameter k and the temporal gradient. This ability is particularly suitable for the analysis scenarios of multi-source heterogeneous data, where user behavior often exhibits high temporal dynamics and feature complexity. For example, in e-commerce promotion analysis, the gamma function can dynamically adjust weights according to the degree of user behavior's response to promotional activities, thus providing more accurate support for the identification of high-value customers and the formulation of conversion strategies. The gamma function's dynamic weights also demonstrate important robustness in dealing with abnormal behaviors and noisy data. Through the combined effect of the lower integration limit and the sparsity parameter, the gamma function can effectively suppress the negative impact of abnormal behaviors on the model's prediction results. For example, when a user exhibits abnormal short-term high-frequency behaviors during a certain period, the non-linear attenuation mechanism of the gamma function's weights can quickly reduce the weights of these abnormal points, thus avoiding the distortion of the model's prediction results. This mechanism is particularly important in multi-source heterogeneous data scenarios because data noise and abnormal behaviors are inevitable in such scenarios. Through the dynamic weight distribution of the gamma function, the model can enhance its ability to handle abnormal data without affecting the overall prediction accuracy.

[0081] Although the specific embodiments of the present invention have been described above, those skilled in the art should understand that these specific embodiments are merely illustrative. Without departing from the principles and essence of the present invention, those skilled in the art can make various omissions, substitutions, and changes to the details of the above methods and systems. For example, combining the above method steps so as to perform substantially the same functions in a substantially the same way to achieve substantially the same results falls within the scope of the present invention. Therefore, the scope of the present invention is only defined by the appended claims.

Claims

1. An intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing, characterized in that The method includes: a data collection step, a data analysis and processing step, a data fusion processing step, and a target user customer acquisition processing step; the data collection step is used to obtain user behavior data from multi-source data and represent the user behavior data as a multi-order tensor; the data analysis and processing step is used to perform tensor decomposition on the multi-order tensor to obtain a core tensor, construct time series features based on the decomposed core tensor, calculate the similarity between users according to the time series features, and construct a behavior propagation equation based on the similarity between users to obtain a behavior propagation tensor; the data fusion processing step is used to perform feature fusion on the core tensor, time series features, and behavior propagation tensor to obtain fusion features; the target user customer acquisition processing step is used to predict the customer conversion rate of each user based on the fusion features to obtain the conversion prediction rate at which the user can be converted into a target customer; compare the conversion prediction rate of each user with the set confidence level, and use the users with a conversion prediction rate higher than the confidence level as target potential customers.

2. The intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing according to claim 1, characterized in that, Multi - order tensor It is represented by the following formula: ； Among them, the multi-order tensor represents the user in the data source the dimension time the behavioral eigenvalue below; represents the total number of sampling points in the data source ; is the sampling space, representing the region for calculating the integral, defined as the spatial range where the user and the sampling points are located; is the data source in the dimension time the behavioral characteristic function, which is a quantity for describing behavioral characteristics, including: click-through rate, purchase volume or number of visits; is the sampling time point; represents the Euclidean distance between the user location and the sampling point location ; represents the eigenvector of the sampling point in time reflecting the behavioral characteristics of the sampling points of the user behavior data in time , including: purchase records or browsing behaviors; is the standard deviation of the Gaussian distribution.

3. The intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing according to claim 2, wherein, In the data analysis and processing step, the multi-order tensor is decomposed by Tucker decomposition to obtain a core tensor, and the formula is as follows: ； Among them, the nuclear tensor satisfies the following sparse constraint conditions: ； Among them, is the user modal basis matrix, is the dimension modal basis matrix, is the time modal basis matrix, is the data source modal basis matrix, and all are preset matrices; is the sparsity parameter; represents the L0 norm; represents the Frobenius norm; is the operation of the rank of the matrix; represents the user 's kernel tensor. When it is the user , it is represented as .

4. The intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing according to claim 3, characterized in that, The data analysis and processing steps construct temporal features based on the decomposed core tensor through the following formula : ； Among them, represents the transpose operation of a matrix; is the time integration variable; represents the operation of the trace of a matrix; represents that the time is when the time-weighted matrix; represents that the time is when the time-weighted matrix, which is preset; represents the temporal characteristics of the user. When it is the user , it is represented as ; represents the time gradient of the kernel tensor.

5. The intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing according to claim 4, wherein The data analysis and processing steps calculate the similarity between users based on the time series characteristics through the following formula, and based on the similarity between users : ； Among them, represents the tensor inner product operation; represents the user 's basis vector; represents the user 's basis vector; is a similarity adjustment parameter and is a preset value; represents the temporal feature of the user ; represents the temporal feature of the user ; represents the user and the user 's inter - user similarity; taking all inter - user similarities as an element to form an inter - user similarity matrix , is the th row and th column element in 6. The intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing according to claim 5, wherein, In the data analysis and processing step, a behavior propagation equation based on the similarity between users is constructed through the following formula: ； Among them, is the Laplace operator; represents the behavior propagation tensor of the user . When it is for the user , it is expressed as ; is the divergence operator.

7. The intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing according to claim 6, wherein The data fusion processing step is used to fuse the kernel tensor, the temporal feature, and the behavior propagation tensor through the following formula to obtain the fusion feature : ； Among them, represents the fusion feature of the user . When it is for the user , it is expressed as ; represents 's time gradient; is 's corresponding form under the time integration variable; is 's corresponding form under the time integration variable; is 's corresponding form under the time integration variable; is the tensor outer product operation.

8. The intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing according to claim 7, characterized in that The target user customer acquisition processing step uses the following formula to predict the customer conversion rate of each user based on the fusion features, and obtains the conversion prediction rate at which the user can be converted into a target customer : ； Among them, represents the conversion prediction rate of the user , indicating that starting from time , after the window time , the conversion prediction rate that this user can be converted into a target customer; is the corresponding form under the time integral variable; is the corresponding form under the time integral variable; is the corresponding form under the time integral variable; is the dynamic weight based on the gamma function.

9. The intelligent cloud data acquisition and dimensionality reduction customer acquisition method based on compressive sensing according to claim 8, wherein, Dynamic weight based on gamma function It is represented by the following formula: 。

Images