Feature space-based user demand accurate extraction system and method

Abstract

The application provides a feature space-based user demand point accurate extraction system and method. The application introduces data assignment and integral operation, adopts a correlation coefficient calculation method, constructs a historical demand feature space vector method and an accurate demand point output formula, deeply mines user demand features, and constructs an accurate demand point feature extraction algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computer, in particular to a user demand accurate extraction method based on feature space. BACKGROUND

[0002] The rapid development of mobile internet technology enables people to access information anytime and anywhere, and people's life has undergone tremendous changes. Various internet technologies that rely on obtaining users' daily information to obtain user demand and push demand-related information and products to users have emerged.

[0003] However, the existing technology based on obtaining user information to extract demand points is not deep enough in mining user demand points, and lacks precision. This results in a large range of information pushed to users and too much information, which to some extent wastes users' time, makes users have a poor experience, and loses their interest in information browsing or product purchase. SUMMARY

[0004] The technical problem to be solved by the present application is to deeply mine user demand features and construct a precise demand point feature extraction algorithm. Therefore, a user demand point accurate extraction system and method based on feature space are provided.

[0005] The technical scheme of the present application is as follows:

[0006] A user demand point accurate extraction system based on feature space comprises the following parts:

[0007] A historical user information storage module, a historical user information acquisition module, a real-time user information acquisition module, a first preprocessing module, a second preprocessing module, a precise demand point feature extraction module, and a user real-time precise demand generation module.

[0008] The first preprocessing module comprises a communication unit, a time alignment unit, a data preprocessing unit, and a historical demand feature space vector construction unit.

[0009] The historical user information storage module is configured to store historical user information and transmit the historical user information to the historical user information acquisition module, and the historical user information acquisition module is configured to acquire original user demand information and demand attribute information and transmit the original user demand information and the demand attribute information to the communication unit; the communication unit is configured to transmit the received original user demand information and the demand attribute information to the time alignment unit for time alignment processing, obtain aligned user demand information and aligned demand attribute information, and transmit the aligned user demand information and the aligned demand attribute information to the data preprocessing unit for data preprocessing, and transmit the data preprocessed aligned user demand information and the aligned demand attribute information to the historical demand feature space vector construction unit to construct demand feature space vectors and demand attribute feature space vectors, and transmit the demand feature space vectors and the demand attribute feature space vectors to the communication unit and transmit the demand feature space vectors and the demand attribute feature space vectors to the precise demand point feature extraction module through the communication unit;

[0010] The real-time user information acquisition module is configured to acquire user data information at the current time, and transmit the user information at the current time to the second preprocessing module for data preprocessing, and transmit the data preprocessed user information at the current time to the precise demand point feature extraction module.

[0011] The precise demand point feature extraction module is configured to construct a demand point precise extraction algorithm framework, extract precise demand point features, and input the precise demand point features to the user real-time precise demand generation module for demand generation and transmission to the historical user information storage module for storage.

[0012] A user demand precise extraction method based on a feature space, comprising the following steps:

[0013] Step S1: performing time domain alignment processing on original user demand information and demand attribute information acquired from historical user information, and performing integral operation data preprocessing on all user information;

[0014] Step S2: constructing demand feature space vectors;

[0015] Step S3: introducing user demand feature space vectors as weight connections, interacting and fusing historical data and real-time data, constructing a demand point pushing algorithm framework, and obtaining a user demand precise target.

[0016] Preferably, the step S1 comprises:

[0017] The original user information after time alignment is referred to as aligned user information, which includes aligned user demand information and aligned demand attribute information, and is subjected to digital processing, and the digitalized aligned user information is referred to as digitalized user information, denoted as x (0S) wherein s represents seconds;

[0018] According to the actual application scene setting requirement database C, the requirement database is set in advance according to the actual application scene, which is used to filter the digital user information;

[0019] Let C={c1, c2, c3,...,c G}, g={1, 2, 3,...,G}, c G G represents the Gth demand point of the actual application scene, and c g represents any one demand point;

[0020] x (0s) is assigned by the following formula:

[0021] x (0s) assigned to 0 is deleted, and x (0s) assigned to 1 is retained;

[0022] The digital user information x (0s) retained is standardized, and the specific processing method is realized by the following formula:

[0023]

[0024] Wherein represents the user information after data preprocessing, and θ∈[0, π].

[0025] Preferably, the step S2 comprises:

[0026] Let c∈{1, 2, 3,...,M}, d∈{1, 2, 3,...,M}, and c is less than d,

[0027] Then The correlation coefficient θ between is calculated by the following formula:

[0028]

[0029] By comparing the size of θ(c, d), the user demand information space dimension characteristic value whose correlation coefficient with any other space dimension is the maximum value is found, and it is taken as the precise user information feature vector.

[0030] Preferably, the step S2 comprises:

[0031] The correlation factor θ(m, n) between the precise user demand information feature vector and any one space dimension characteristic value is calculated, and the specific calculation method is as follows:

[0032]

[0033] wherein a is a parameter, and a = [1,∞].

[0034] Preferably, the step S3 comprises:

[0035] The real-time data input layer has R neurons, and any neuron can be represented by r.

[0036] The user information at the current time after data preprocessing as the input of the demand point accurate extraction algorithm framework,

[0037] The input of the real-time data input layer is , and the output is also

[0038] The feature insertion layer has R+1 neurons:

[0039] The input of the R neurons is , and the output is O r , and

[0040] The inserted feature neuron is the accurate demand attribute feature vector obtained from the historical demand point attribute information feature vector The output is

[0041] The information fusion layer has Q neurons, and any neuron is represented by q.

[0042] wherein w R+1→q is the weight connection between the neurons of the feature insertion layer and the corresponding neurons of the information fusion layer, and b q represents the bias between the neurons of the feature insertion layer and the corresponding neurons of the information fusion layer.

[0043] The input of any neuron of the information fusion layer is u q =(O R+1 , O r ), and the output of any neuron of the information fusion layer is v q :

[0044]

[0045] wherein represents the interaction operation on the data at both ends of the arrow, f is an interaction function, and

[0046]

[0047]

[0048] Preferably, the step S3 comprises:

[0049] The input of the accurate demand point output layer is v q , and the output is

[0050] where w q→p and b p are the weight connection and bias between any neuron of the information fusion layer and the neuron of the accurate demand point output layer, respectively, and w q→p is valued from , is a feature extraction function, and where β and ε are a set of adjustment parameters, then

[0051]

[0052] The present application has at least the following beneficial effects:

[0053] (1) The data preprocessing method used in the present application introduces data assignment and integral operation, reduces the limitations of data processing, reduces the possibility of data distortion, and makes the data retain the characteristic properties while being preprocessed, and the information is screened and processed, improving the speed of subsequent data operation.

[0054] (2) The correlation coefficient calculation method used in the present application has a faster calculation speed compared with the prior art, has a simpler requirement for the source and characteristics of data, and the correlation factor coefficient obtained is more convenient for selecting the spatial dimension characteristic value.

[0055] (3) The method for constructing the historical demand feature space vector used in the present application greatly analyzes the historical user information and demand attribute data, deeply mines the user demand characteristics, and provides accurate parameters for real-time characteristic demand extraction of users.

[0056] (4) The introduction of β and ε as a set of adjustment parameters in the accurate demand point output formula of the present application makes the function have stronger convergence, prevents data distortion, and uses the historical demand feature space vector as the selection space of the weight, so that the demand point has strong accuracy, and the user has better use feeling. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 is a structure block diagram of the first preprocessing module 40 of the present application;

[0058] Figure 2 is a structure block diagram of the first preprocessing module 40 of the present application;

[0059] Figure 3 The demand point accurate extraction algorithm block diagram of the present application. DETAILED DESCRIPTION

[0060] The embodiments of the present application will be described in detail below with reference to the accompanying drawings. The technical problems are solved and the technical effects are achieved by applying technical means in the present application. It should be noted that the features of each embodiment in the present application can be combined with each other without conflict, and the technical solutions formed thereby are within the protection scope of the present application.

[0061] Reference Figure 1 The present application proposes a feature space-based user demand point accurate extraction system, which comprises the following parts:

[0062] The historical user information storage module 10, the historical user information acquisition module 20, the real-time user information acquisition module 30, the first preprocessing module 40, the second preprocessing module 50, the accurate demand point feature extraction module 60, and the real-time accurate demand generation module 70.

[0063] The historical user information storage module 10 is configured to store historical user information and transmit the historical user information to the historical user information acquisition module 20 through data transmission.

[0064] The historical user information acquisition module 20 is configured to acquire historical user information, including original user demand information and demand attribute information, and send the original user demand information and demand attribute information to the first preprocessing module 40 through data transmission.

[0065] The real-time user information acquisition module 30 is configured to acquire real-time user information. The user information according to the present application includes user demand information and demand attribute information, and the user information at the current time is transmitted to the second preprocessing module 50 through data transmission.

[0066] Reference Figure 2The first preprocessing module 40 comprises a communication unit 401, a time alignment unit 402, a data preprocessing unit 403, and a historical demand feature space vector construction unit 404. The time alignment unit 402 is configured to perform time alignment processing on historical user information and demand attribute information to obtain aligned user information at time 0 in the time domain. The data preprocessing unit 403 is configured to perform data preprocessing on the aligned user information. The historical demand feature space vector construction unit 404 is configured to construct demand feature space vectors and demand attribute feature space vectors. The communication unit 401 is configured to be communicatively connected to the historical user information acquisition module 20 and the precise demand point feature extraction module 60, and to transmit the constructed feature space vectors to the precise demand point feature extraction module 60 in a data transmission manner.

[0067] The second preprocessing module 50 is configured to perform data preprocessing on real-time user data information and transmit the data-preprocessed real-time user data information to the precise demand point feature extraction module 60 in a data transmission manner.

[0068] The precise demand point feature extraction module 60 is configured to precisely extract user demand point features and transmit the precise user demand point features to the user real-time precise demand generation module 70 in a data transmission manner.

[0069] The user demand point precise extraction system based on the feature space stores historical user information in the historical user information storage module 10, and transmits the historical user information to the historical user information acquisition module 20. The historical user information acquisition module 20 acquires original user demand information and demand attribute information and transmits them to the communication unit 401 in the first preprocessing module 40. The communication unit 401 transmits the received original user demand information and demand attribute information to the time alignment unit 402 for time alignment processing, obtains aligned user demand information and aligned demand attribute information, and transmits them to the data preprocessing unit 403 for data preprocessing. The data-preprocessed aligned user demand information and aligned demand attribute information are sent to the historical demand feature space vector construction unit 404 to construct demand feature space vectors and demand attribute feature space vectors. The demand feature space vectors and demand attribute feature space vectors are sent to the communication unit 401 and transmitted to the precise demand point feature extraction module 60 through the communication unit 401.

[0070] The real-time user information acquisition module 30 acquires user data information at the current time, and transmits the current user information to the second preprocessing module 50 for data preprocessing. The data-preprocessed current user information is sent to the precise demand point feature extraction module 60.

[0071] The accurate demand point feature extraction module 60 constructs an accurate demand point extraction algorithm framework, extracts accurate demand point features, and inputs the accurate demand point features into the user real-time accurate demand generation module 70 for demand generation and transmission to the historical user information storage module 10 for storage.

[0072] The user demand accurate extraction method based on the feature space comprises the following steps:

[0073] S1 aligns the original user demand information and demand attribute information obtained from the historical user information in the time domain to 0, and performs integral operation formula data preprocessing on all user information;

[0074] S11 obtains historical user information from the historical user information storage module 10, and obtains original user demand information and demand attribute information through the historical user information acquisition module 20, wherein the generated original user demand information and demand attribute information are collectively referred to as original user information, which is a prior art and will not be described in detail here.

[0075] In the time alignment unit 402 in the first preprocessing module 40, the original user information is aligned in the time domain:

[0076] Let the original user information be x(T0), and T0 represent the time scale of the original user information. Subtract the corresponding time scale T0 from the time scale record value T0 in the original user information, so that the time scale of the original user information is 0 seconds.

[0077] The original user information after time alignment is referred to as aligned user information, which contains aligned user demand information and aligned demand attribute information, and is subjected to digital processing. The digital processing method described in the present application is a prior art and will not be described in detail here. The digitized aligned user information is referred to as digitized user information, and is denoted as x (0s) , wherein s represents seconds;

[0078] According to the actual application scenario, a demand database C is set, which is set in advance according to the actual application scenario and is used to filter the digitized user information described in the present application;

[0079] Let C={c1, c2, c3,...,c G}, g={1, 2, 3,..., G}, c G represent the Gth demand point of the actual application scenario of the method described in the present application, and c g represents any one demand point;

[0080] x (0s) is assigned a value by the following formula:

[0081] x is assigned to 0 (0s) deletion, x is assigned to 1 (0s) retained

[0082] digitized user information x retained (0s) standardized, and the specific processing is realized by the following formula:

[0083]

[0084] wherein represents the user information after data preprocessing, and θ∈[0, π].

[0085] The data preprocessing method used in the application introduces data assignment and integral operation, reduces the limitations of data processing, reduces the possibility of data distortion, and makes the data retain the characteristic properties of the data while the data is preprocessed, and the information is screened and processed, thereby improving the speed of subsequent data operation.

[0086] S2 constructs a demand feature space vector

[0087] S21 constructs a feature space vector for the homogenized user information after data preprocessing in the historical demand feature space vector construction unit 404 in the first preprocessing module 40:

[0088] Let the homogenized user demand information after data preprocessing be denoted as Let the homogenized demand attribute information after data preprocessing be denoted as

[0089] Let the user demand feature space vector be denoted as:

[0090]

[0091] wherein M represents the spatial dimension of the user demand information, and let m={1, 2, 3,..., M}.

[0092] then

[0093] Similarly, let the demand attribute feature space vector be denoted as

[0094]

[0095] wherein N represents the spatial dimension of the demand attribute information, and let n={1, 2, 3,..., N}.

[0096] then

[0097] The correlation coefficient θ between the user demand feature space vector and

[0098] Let c∈{1,2,3,...,M}, d∈{1,2,3,...,M}, and c is less than d.

[0099] The correlation coefficient θ between the user demand feature space vector and And The correlation coefficient θ between the user demand feature space vector and

[0100]

[0101] By comparing the size of θ(c,d), the user demand information space dimension characteristic value with the maximum correlation coefficient with any other space dimension is found And it is used as the precise user demand information feature vector.

[0102] The correlation coefficient calculation method adopted by the present application has a faster calculation speed compared with the prior art, has a simpler requirement for the source and data characteristics of data, and the correlation factor coefficient obtained is more convenient for the selection of space dimension characteristic values.

[0103] S22 calculates the correlation factor θ(m,n) of any one space dimension characteristic value in the precise user demand information feature vector and The specific calculation method is as follows:

[0104]

[0105] Wherein, α is a parameter, and α=[1,∞].

[0106] The correlation factor calculation method used by the present application breaks the limitation of data operation in the prior art, introduces the parameter α=[1,∞] to participate in the calculation of the correlation factor, so that the correlation degree between data information can be deeply mined, and the precise demand point can be mined.

[0107] The size of θ(m,n) is compared, the space dimension characteristic value with the maximum θ(m,n) value is obtained, and is recorded as And it is used as the precise demand attribute feature vector.

[0108] The method for constructing the demand feature space vector adopted by the present application greatly analyzes the user information and demand attribute data in combination, deeply mines the user demand characteristics, and provides a precise parameter for real-time characteristic demand extraction of users.

[0109] S3 refers to Figure 3 The user demand feature space vector is introduced as a weight connection to interact and fuse the historical data and real-time data, a demand point pushing algorithm framework is constructed, and a user demand precise target is obtained.

[0110] The user information at the current time is collected from the real-time user information collection module 30, and this technology is not described again because it is prior art. The user information at the current time is preprocessed by the step S12 of the application, and the data preprocessing process is performed in the second preprocessing module 50. The data preprocessing method in the second preprocessing module 50 is the same as the step S12 of the application, and is not described again. R represents the spatial dimension of the user information at the current time, and r = {1, 2, 3,..., R}, and the user information at the current time after data preprocessing is denoted as

[0111] A demand point accurate extraction algorithm framework is constructed in the accurate demand point feature extraction module 60.

[0112] The demand point accurate extraction algorithm framework has four layers. The first layer is a real-time data input layer I, which inputs the user information at the current time after data preprocessing. The second layer is a feature insertion layer O, which inserts a demand attribute feature vector. The third layer is an information fusion layer Q, which performs information interaction and fusion between the user information features obtained in real time and the demand attribute feature vector. The fourth layer is a demand point output layer P, which outputs an accurate demand point feature vector.

[0113] S31, the real-time data input layer has R neurons, and any neuron can be represented by r.

[0114] The user information at the current time after data preprocessing is input into the demand point accurate extraction algorithm framework.

[0115] The input of the real-time data input layer is and the output is also

[0116] S32, the feature insertion layer has R+1 neurons.

[0117] The input of the R neurons is and the output is O r .

[0118] The inserted feature neuron is the accurate demand attribute feature vector obtained from the historical demand point attribute information feature vector and the output is

[0119] S33, the information fusion layer has Q neurons, and any neuron is represented by q.

[0120] where w R+1→q is the weight connection between the neurons of the feature insertion layer and the corresponding neurons of the information fusion layer, and bq This represents the bias between neurons in the feature insertion layer and corresponding neurons in the information fusion layer.

[0121] The input to any neuron in the information fusion layer is u q =(O R+1 O r The output of any neuron in the information fusion layer is v q :

[0122]

[0123] in This indicates that interactive operations are performed on the data at both ends of the arrow, where f is the interactive function, and

[0124]

[0125]

[0126] The information fusion method adopted in this invention uses interactive functions for information fusion, which enables deep integration of historical feature vectors and real-time data, and more comprehensive mining of data information, so that the obtained demand points are closer to the user's real and comprehensive needs.

[0127] The S34 precise demand point output layer P has 1 neuron.

[0128] The input to the precise demand point output layer is v q The output is

[0129] Where w q→p and b p These represent the weight connections and biases between any neuron in the information fusion layer and the neuron in the output layer at the precise demand point, respectively, and w q→p from The value is obtained from the middle. Let be the feature extraction function, and Where β and ε are a set of adjustment parameters. but

[0130]

[0131] The precise demand point output formula adopted in this invention introduces β and ε as a set of adjustment parameters, which makes the function more convergent, prevents data distortion, and makes the demand points more accurate by using the historical demand feature space vector as the selection space of weights, thus giving users a better user experience.

[0132] The total output of the output layer is Y, which is the feature vector of the precise required point.

[0133] The precise demand point feature vector Y is sent into the user real-time precise demand generation module 70 for demand generation, and the generation technology is prior art, which is not described in detail here.

[0134] In summary, the user demand point precise extraction system and method based on feature space are realized.

[0135] The present application is described in reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus generate one or more functions specified in the flowcharts and / or block diagrams. Figure 1 one or more functions specified in the flowcharts and / or block diagrams. Figure 1 means for performing the functions specified in the flowchart(s) and / or block diagram block(s).

[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable data processing apparatus to produce a computer-implemented process such that the instructions executed by the computer or other programmable data processing apparatus provide the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more functions specified in the flowcharts and / or block diagrams. Figure 1 steps for performing the functions specified in the flowchart(s) and / or block diagram block(s).

[0137] Although preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to the embodiments once they have the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including all the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0138] Obviously, various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and scope of the present application. Thus, it is intended that the present application embrace all such modifications and changes and include them within the scope of the appended claims and their equivalents.

Claims

1.A method for accurately extracting user needs based on a feature space, characterized by, Comprise the following steps: Step S1 performs time domain alignment processing on the original user demand information and demand attribute information obtained in the historical user information, and the original user information after time alignment is referred to as aligned user information, which contains aligned user demand information and aligned demand attribute information, and is subjected to digital processing, and the digital aligned user information is referred to as digital user information, denoted as where s represents seconds. According to the actual application scene setting demand database C, the demand database is set in advance according to the actual application scene, which is used to filter the digital user information; Let , g={1,2,3,...,G}, represent the Gth demand point of the actual application scenario, then represent any one demand point; The following equation is assigned a value by: ​ to be assigned the value 0 deletion, assigned the value 1 retention; Digitized user information that is retained Data preprocessing for integral operation is performed, and the specific processing method is realized by the following formula: wherein represents the user information after data preprocessing, ; Step S2, the homogenized user demand information after data preprocessing is recorded as , the homogenized demand attribute information after data preprocessing is recorded as , a demand feature space vector is constructed based on the homogenized user demand information after data preprocessing , a demand attribute feature space vector is constructed based on the homogenized demand attribute information after data preprocessing ; let , , wherein M represents the space dimension of the user demand information, and c is less than d; Then correlation between correlation between is calculated from the following equation: By comparing the size of the correlation coefficient with any other spatial dimension, find the user demand information space dimension characteristic value when the correlation coefficient with any other spatial dimension is maximum , and take it as the accurate user information feature vector; The calculation of the precise user demand information feature vector and The correlation factor of any one spatial dimension characteristic value in the middle , m={1,2,3,...,M}, n={1,2,3,...,N}, N represents the spatial dimension of demand attribute information, and the specific calculation method is as follows: wherein is a parameter, and ; Step S3 introduces user demand feature space vector as weight connection, interacts and fuses historical data and real-time data, constructs demand point pushing algorithm framework, and obtains user demand accurate target; The specific framework has four layers: Real-time data input layer, a total of R neurons, any neuron can be represented by r; The user information of the current moment after data preprocessing As the input of the demand point accurate extraction algorithm framework, the input of the real-time data input layer is , and the output is also ; Feature insertion layer, a total of R+1 neurons: where the inputs to the R neurons are , the outputs are , and , The inserted one characteristic neuron is the accurate demand attribute characteristic vector obtained by the historical demand point attribute information characteristic vector , the output is ; Information fusion layer, with Q neurons, any neuron is represented by q: wherein is a weight connection between a neuron of the feature insertion layer and a corresponding neuron of the information fusion layer, denotes a bias between a neuron of the feature insertion layer and a corresponding neuron of the information fusion layer; The input of any neuron of the information fusion layer is The output of any neuron of the information fusion layer is : wherein represents an interaction operation on data at both ends of the arrow, is an interaction function, and The input of the precise demand point output layer is , and the output is wherein and are the weight connections and biases between the information fusion layer arbitrary neurons to the precision demand point output layer neurons, respectively, and are valued from , is a feature extraction function, and wherein and are a set of tuning parameters, j= then 。 2. A feature space-based user demand point accurate extraction system applied to the feature space-based user demand accurate extraction method of claim 1, characterized in that, Comprise the following parts: Historical user information storage module, historical user information acquisition module, real-time user information acquisition module, first preprocessing module, second preprocessing module, accurate demand point feature extraction module, user real-time accurate demand generation module; The first preprocessing module comprises a communication unit, a time alignment unit, a data preprocessing unit, and a historical demand feature space vector construction unit; The historical user information storage module is used for storing historical user information, and transmitting the historical user information to the historical user information acquisition module, acquiring original user demand information and demand attribute information through the historical user information acquisition module, and transmitting them to the communication unit; The communication unit transmits the received original user demand information and demand attribute information to the time alignment unit for time alignment processing, obtains aligned user demand information and aligned demand attribute information, and transmits them to the data preprocessing unit for data preprocessing, and sends the data preprocessed aligned user demand information and aligned demand attribute information to the historical demand feature space vector construction unit to construct demand feature space vector and demand attribute feature space vector, and sends the demand feature space vector and demand attribute feature space vector to the communication unit, and transmits them to the accurate demand point feature extraction module through the communication unit; Through the real-time user information acquisition module, the user data information at the current time is acquired, and the user information at the current time is transmitted to the second preprocessing module for data preprocessing, and the data preprocessed user information at the current time is sent to the accurate demand point feature extraction module; Through the accurate demand point feature extraction module, the demand point accurate extraction algorithm framework is constructed, the accurate demand point feature is extracted, and the accurate demand point feature is input to the user real-time accurate demand generation module for demand generation and transmission to the historical user information storage module for storage.

Images