CN116595255A - Big data analysis method and system for cloud service pushing - Google Patents
Big data analysis method and system for cloud service pushing Download PDFInfo
- Publication number
- CN116595255A CN116595255A CN202310574565.XA CN202310574565A CN116595255A CN 116595255 A CN116595255 A CN 116595255A CN 202310574565 A CN202310574565 A CN 202310574565A CN 116595255 A CN116595255 A CN 116595255A
- Authority
- CN
- China
- Prior art keywords
- data
- user
- intention
- browsing
- live broadcast
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 67
- 238000007405 data analysis Methods 0.000 title claims abstract description 46
- 238000004364 calculation method Methods 0.000 claims abstract description 72
- 238000011156 evaluation Methods 0.000 claims abstract description 33
- 239000011159 matrix material Substances 0.000 claims description 47
- 238000009826 distribution Methods 0.000 claims description 41
- 238000004458 analytical method Methods 0.000 claims description 31
- 238000012545 processing Methods 0.000 claims description 30
- 238000004422 calculation algorithm Methods 0.000 claims description 19
- 238000001514 detection method Methods 0.000 claims description 19
- 238000000605 extraction Methods 0.000 claims description 17
- 238000012216 screening Methods 0.000 claims description 15
- 238000010276 construction Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000000611 regression analysis Methods 0.000 claims description 8
- 230000008878 coupling Effects 0.000 claims description 7
- 238000010168 coupling process Methods 0.000 claims description 7
- 238000005859 coupling reaction Methods 0.000 claims description 7
- KJONHKAYOJNZEC-UHFFFAOYSA-N nitrazepam Chemical compound C12=CC([N+](=O)[O-])=CC=C2NC(=O)CN=C1C1=CC=CC=C1 KJONHKAYOJNZEC-UHFFFAOYSA-N 0.000 claims description 7
- 238000007619 statistical method Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 3
- 239000000523 sample Substances 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 20
- 230000006399 behavior Effects 0.000 description 12
- 238000013075 data extraction Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 8
- 238000005457 optimization Methods 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000013145 classification model Methods 0.000 description 4
- 238000003860 storage Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 239000012141 concentrate Substances 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 238000013079 data visualisation Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000013486 operation strategy Methods 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 230000008054 signal transmission Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of data analysis, in particular to a big data analysis method and system for cloud service pushing. The method comprises the following steps: acquiring historical user browsing data, and extracting intent from the historical user data so as to generate user historical intent data; scaling calculation is carried out on the historical intent data of the user, and a user pushing model is constructed; acquiring user historical purchase data, and calculating the user historical purchase data and the user historical intention data so as to acquire the user historical purchase intention data; acquiring current purchase data of a user, and performing classification calculation on the current purchase data of the user based on historical purchase intention data of the user so as to generate first intention evaluation data; the first intent assessment data is calculated to obtain developable user data. The cloud service accurate pushing method and device are used for achieving cloud service accurate pushing by adopting a data analysis technology.
Description
Technical Field
The invention relates to the technical field of data analysis, in particular to a big data analysis method and system for cloud service pushing.
Background
With the rapid development of technologies such as cloud computing, big data and artificial intelligence, more and more enterprises and individuals begin to adopt cloud services, such as cloud storage, cloud computing, cloud databases and the like, so as to provide more diversified and efficient services for users. However, how to push service content accurately and quickly has become a concern for cloud service providers.
Disclosure of Invention
The application provides a big data analysis method for cloud service pushing to solve at least one technical problem.
The application provides a big data analysis method for cloud service pushing, which comprises the following steps:
step S1: acquiring historical user browsing data, and extracting intent from the historical user data so as to generate user historical intent data;
step S2: scaling calculation is carried out on the historical intent data of the user, and a user pushing model is constructed;
step S3: acquiring user historical purchase data, and calculating the user historical purchase data and the user historical intention data so as to acquire the user historical purchase intention data;
step S4: acquiring current purchase data of a user, and performing classification calculation on the current purchase data of the user based on historical purchase intention data of the user so as to generate first intention evaluation data;
Step S5: calculating the first intent assessment data to obtain developable user data;
step S6: extracting features of historical user browsing data to obtain user browsing preference features and user browsing cost features, and correcting the user pushing model by using the user browsing preference features and the user browsing cost features to obtain an optimized user pushing model;
step S7: and carrying out intent assessment on the developable user data by using the optimized user pushing model so as to obtain second intent assessment data, and sending the pushing user data to a cloud service pushing system for content screening pushing processing.
According to the method and the device for achieving the user history intention data, the historical user browsing data are obtained through the cloud platform, the intention extraction is carried out on the historical user data, unnecessary pushing of the cloud platform can be reduced, the pushing effect is improved, and therefore the user history intention data are generated. Scaling calculation is carried out on the historical intent data of the user, and a user push model is built, the scaling calculation can unify the historical intent data with different dimensionalities, the variability among the data is reduced, the feature dimensionality of a data set can be reduced, the calculation amount in data processing is reduced, and therefore the user push model is built more accurately, and the accuracy of personalized recommendation is improved. The user historical purchase data is obtained, the user historical purchase data and the user historical intention data are calculated, so that the user historical purchase intention data is obtained, the user preference can be pushed more accurately by combining the user historical purchase data and the historical intention data, and the pushing accuracy and precision are improved. The method comprises the steps of obtaining current purchase data of a user, classifying and calculating the current purchase data of the user based on historical purchase intention data of the user, generating first intention evaluation data, classifying and calculating the current purchase data of the user based on the historical purchase intention data, and rapidly judging whether commodities purchased by the user at present meet the historical intention of the user or not, so that corresponding pushing is carried out, and shopping experience of the user is optimized. The first intention evaluation data is calculated to obtain the developable user data, and by calculating the first intention evaluation data, the developable user can be identified, which helps to concentrate on the user who is most likely to purchase or use the product or service, and improves the pushing accuracy. The method comprises the steps of extracting characteristics of historical user browsing data, obtaining user browsing preference characteristics and user browsing cost characteristics, correcting a user pushing model by utilizing the user browsing preference characteristics and the user browsing cost characteristics, obtaining an optimized user pushing model, better reflecting interests and preferences of a user by extracting the characteristics of the historical user browsing data, correcting the user pushing model by utilizing the characteristics, realizing personalized pushing more accurately, knowing interests and consumption requirements of the user more fully, adjusting and optimizing advertisement putting strategies, improving pushing effect and conversion rate, and realizing more accurate recommendation by optimizing and correcting the characteristics of the historical user browsing data and the user pushing model, thereby improving user liveness and user viscosity. The user data can be subjected to intention evaluation by utilizing the optimized user pushing model, so that second intention evaluation data are obtained, the pushed user data are sent to the cloud service pushing system for content screening pushing processing, the intention evaluation is performed on the user data can be performed by utilizing the optimized user pushing model, the preference and the intention of a user on pushing content can be judged more accurately, the pushing efficiency and the pushing accuracy are improved, the requirements and the interest points of the user can be better known, personalized pushing is realized, the user satisfaction degree and the loyalty degree are improved, the second intention evaluation data can help the platform to adjust pushing strategies, the content screening and pushing effect optimization is realized, the pushing effect and the user feedback rate are improved, and the user dislike rate is reduced.
Optionally, step S1 specifically includes:
step S11: acquiring historical user browsing data;
step S12: extracting website browsing data and live broadcasting browsing data from historical user browsing data, so as to generate the website browsing data and the live broadcasting browsing data;
step S13: performing website intention exploration processing on website browsing data so as to generate website intention data;
step S14: live broadcast intention exploration processing is carried out on live broadcast browsing data, so that live broadcast intention data are generated;
step S15: and carrying out time sequence combination according to the website intention data and the live intention data, thereby obtaining user intention combination data.
According to the method and the device, the historical user browsing data are obtained through the cloud platform, the website browsing data extraction and the live broadcast browsing data extraction are carried out on the historical user browsing data, so that the website browsing data and the live broadcast browsing data are generated, the website browsing data extraction and the live broadcast browsing data extraction are carried out on the historical user browsing data, the browsing behaviors and the preferences of the user can be more comprehensively known, the accuracy and the precision of pushing and personalized services are improved, and the user experience is optimized. The website intention exploration processing is carried out on the website browsing data, so that the website intention data is generated, and commodities or services related to the interests of the user can be more accurately recommended by exploration of the website intention of the user, and the recommendation accuracy and the satisfaction degree of the user are improved. Live intention exploration processing is conducted on live browsing data, so that live intention data are generated, the live intention data can help the cloud platform to know potential demands and interest points of users, live topics and contents are pushed more accurately, and live marketing effects and commodity sales are improved. And carrying out time sequence combination according to the website intention data and the live intention data so as to obtain user intention combination data, and comprehensively analyzing the interests and the demands of the user by carrying out time sequence combination according to the website and the live intention data of the user so as to clearly know the intention and the tendency of the user and improve the accuracy of pushing and personalized services.
Optionally, the website consumption intention exploration process in step S13 includes the following steps:
carrying out statistical analysis on the website browsing data so as to obtain high-frequency intention data of the website;
performing variance analysis on website browsing data so as to obtain website low-frequency intention data;
and carrying out Pelson intention detection on the website browsing data so as to obtain the potential intention data of the website.
According to the invention, the website browsing data is statistically analyzed, so that the high-frequency intention data of the website is obtained, and the behavior and habit of the user can be better known and the data analysis capability and accuracy are improved through the statistical analysis of the website browsing data. And performing variance analysis on the website browsing data to obtain low-frequency intention data of the website, and performing variance analysis on the website browsing data to better know user behaviors and preferences and improve data analysis capability and accuracy. The Pelson intention detection is carried out on the website browsing data, so that the potential intention data of the website is obtained, the Pelson intention detection can find the unaware demands and interests of the user, the original pushing is supplemented and innovated, the user experience and satisfaction are improved, the preference and demands of the user can be better known, and therefore more effective and accurate personalized pushing service is realized.
Optionally, the pearson intent detection is specifically:
step S1331: extracting features of website browsing data to obtain website browsing time data, website page access amount data and user browsing similar page frequency data, and merging the data of the user browsing time data, the user page access amount data and the user browsing similar page frequency data to obtain a website comprehensive data set;
step S1332: calculating the comprehensive data set through the pearson correlation coefficient so as to obtain pearson data;
step S1333: carrying out correlation screening on the pearson data so as to obtain potential characteristic data;
step S1334: carrying out intention degree calculation on the potential characteristic data through association rule analysis so as to obtain intention degree data;
step S1335: carrying out maximum association clustering calculation on the intention data so as to obtain potential intention data of the website;
the intent maximum association clustering calculation specifically comprises the following steps:
model construction is carried out based on the intention data, so that an intention clustering model is obtained;
normalizing the intention data to obtain normalized intention data;
performing similarity matrix calculation on the normalized intent data through the pearson correlation coefficient so as to obtain a similarity matrix;
Normalizing the similarity matrix to obtain a probability distribution matrix;
estimating and calculating the intent data according to the probability distribution matrix based on an EM algorithm, so as to obtain the cluster number;
performing punishment item calculation on the cluster number so as to obtain maximum entropy probability distribution;
optimizing the intent clustering model based on the maximum entropy probability distribution so as to obtain an intent maximum association clustering model;
clustering calculation is carried out on the intent data through the intent maximum association clustering model, so that potential intent data of the website are obtained.
According to the method and the device for processing the web site browsing data, the feature extraction is carried out on the web site browsing data, so that the web site browsing time data, the web site page access amount data and the user browsing similar page frequency data are obtained, the data merging is carried out on the user browsing time data, the user page access amount data and the user browsing similar page frequency data, so that a web site comprehensive data set is obtained, various data features are merged, the user requirements and behaviors can be known more accurately, and therefore the accuracy and the accuracy of data analysis are improved. The comprehensive data set is calculated through the pearson correlation coefficient, so that pearson data is obtained, the unaware demands and interests of the user can be found through analysis of the pearson correlation coefficient, so that new services and products are developed, the evaluation and preference of the user to different features can be known through analysis of the pearson correlation coefficient, the platform design and interaction experience are optimized, the user experience and satisfaction are improved, the pearson correlation coefficient can help enterprises and merchants to better know the demands and preferences of the user, and the data analysis capability and accuracy are improved. And carrying out relevance screening on the pearson data to obtain potential characteristic data, and removing useless or low-relevance characteristic data through pearson relevance screening to reduce errors and interference, so that the relation and weight among different characteristic data are known more accurately. The intent degree is calculated on the potential characteristic data through the association rule analysis, so that the intent degree data is obtained, and the interest and the requirements of the user on different characteristic data can be more accurately known through the association rule analysis and the intent degree calculation. The intent degree data is subjected to maximum association clustering calculation so as to obtain potential intent degree data of websites, users can be more accurately and finely divided according to interests and requirements through the maximum association clustering calculation of the intent degree, so that requirements and characteristics of different user groups are better known, interests and requirements of different user groups can be better known through analyzing the potential intent degree data, and accordingly more effective and accurate personalized recommendation service is realized. Model construction is carried out based on the intention data, so that an intention clustering model is obtained; the intent data is normalized, so that normalized intent data is obtained, the data can be more comparable, deviation is reduced, algorithm precision is improved, data processing is facilitated, and data visualization effect is enhanced. The similarity matrix calculation is carried out on the normalized intention degree data through the pearson correlation coefficient, so that a similarity matrix is obtained, the correlation among different features can be judged more accurately through the pearson correlation coefficient calculation, and the similarity matrix can cluster and classify the different feature data according to the similarity, thereby being beneficial to more carefully analyzing and understanding the data and being used for correlation application. The similarity matrix is normalized, so that a probability distribution matrix is obtained, the normalization can limit the similarity matrix value between 0 and 1, so that huge difference of the similarity matrix value is avoided, and the stability of data is improved; the probability distribution matrix ensures that the sum of all values is 1, i.e. the probability values corresponding to each feature add together to be equal to 1. The method is beneficial to better calculating and analyzing the data and improving the accuracy of data analysis; the probability distribution matrix can better accord with the probability distribution rule, so that the data is more regular and easy to analyze. The method comprises the steps of carrying out estimation operation on intention data according to a probability distribution matrix based on an EM algorithm, so as to obtain a cluster number, carrying out estimation operation on the intention data according to the probability distribution matrix by the EM algorithm, removing noise data, retaining meaningful data, further accurately clustering characteristic data, and carrying out quick and efficient clustering operation based on the probability distribution matrix by the EM algorithm, so that the characteristic data clustering efficiency is improved. And carrying out punishment item calculation on the cluster number so as to obtain maximum entropy probability distribution, and carrying out punishment item calculation on the cluster number when calculating the maximum entropy probability distribution so as to ensure that the cluster result contains the least cluster number, thereby improving the quality of the cluster result. The intent clustering model is optimized based on the maximum entropy probability distribution, so that the intent maximum association clustering model is obtained, the maximum entropy probability distribution is adopted for model optimization, the risk of model overfitting can be effectively reduced, the clustering effect and reliability are improved, the accuracy of the model can be prevented from being influenced by excessive cluster numbers through the maximum entropy principle, and the selection of the cluster numbers is further optimized. The intent data is clustered through the intent maximum association clustering model, so that potential intent data of websites is obtained, users are divided into different clusters according to the intent through the intent maximum association clustering model, potential user groups and corresponding requirements of the potential user groups can be better mined, and potential clients are accurately positioned.
Optionally, the live consumption intention exploration process in step S14 includes the steps of:
s141: performing regression analysis on the live broadcast browsing data so as to obtain live broadcast high-frequency intention data;
s142: performing variance analysis on the live broadcast browsing data so as to obtain live broadcast low-frequency intention data;
s143: and carrying out intention differential privacy detection on the live broadcast browsing data so as to obtain live broadcast potential intention data.
According to the method, regression analysis is carried out on the live broadcast browsing data, so that live broadcast high-frequency intention data is obtained, and regression analysis is carried out on the live broadcast browsing data, so that the live broadcast high-frequency intention data can be accurately predicted, preference and popular trend of live broadcast content can be found, operation decision is refined, and basis is provided for personalized pushing of users. The variance analysis is carried out on the live broadcast browsing data so as to obtain live broadcast low-frequency intention data, the variance analysis is carried out on the live broadcast browsing data, the variability among different audience groups can be distinguished, the live broadcast low-frequency intention data of the audience can be further subdivided, the diversity and the distribution condition of the audience needs can be better known through analyzing the variance of the live broadcast browsing data, so that the content coverage of the live broadcast can be better enlarged, the demands of different audiences can be met, the variance analysis can enable us to find out the instability factors existing in the live broadcast browsing data, and reference and basis are provided for us to adjust live broadcast content and pushing strategies, so that the quality and reliability of live broadcast are improved. The intention difference privacy detection is carried out on live broadcast browsing data, so that live broadcast potential intention data is obtained, the intention difference privacy detection is carried out on live broadcast browsing data, the privacy disclosure risk of a user in the data collection and analysis process can be reduced, the personal privacy rights of the user can be protected, the quality and the credibility of the live broadcast browsing data can be improved, and errors and deviations of the data due to privacy disclosure and other problems are avoided.
Optionally, the step of intentionally detecting differential privacy in step S143 is specifically:
feature extraction is carried out on live broadcast browsing data, so that live broadcast clicking frequency data, live broadcast access time data and user browsing similar live broadcast frequency data are obtained, and data coupling association is carried out on the live broadcast clicking frequency data, the live broadcast access time data and the user browsing similar live broadcast frequency data, so that a live broadcast comprehensive data set is obtained;
carrying out random disturbance processing on the live broadcast comprehensive data set by a Laplace noise method so as to obtain live broadcast noise data;
performing differential calculation on the live noise data so as to obtain live differential data;
calculating live broadcast differential data and live broadcast noise data, so as to obtain live broadcast intention score data of a user;
and classifying and calculating the live broadcast intention score data of the user so as to obtain live broadcast potential intention data.
According to the method, characteristics of live broadcast browsing data are extracted, so that live broadcast clicking frequency data, live broadcast access time data and user browsing similar live broadcast frequency data are obtained, and data coupling association is performed on the live broadcast clicking frequency data, the live broadcast access time data and the user browsing similar live broadcast frequency data, so that a live broadcast comprehensive data set is obtained, the most critical few characteristics are extracted to effectively compress the data set, limited computing and storage resources are utilized to effectively manage the data so as to save cost, key information can be quickly identified from massive live broadcast browsing data through the characteristics extraction, and the data can be clearly displayed, so that understanding and analysis of the data become more efficient; and carrying out data coupling association on live click frequency data, live access time data and data of similar live frequency browsed by a user, comprehensively considering various information of the user, and acquiring more comprehensive user demand information. After the live broadcast comprehensive data set is obtained, deeper data analysis can be performed, live broadcast data is analyzed in multiple dimensions, user requirements and behavior modes are better known, push strategies and live broadcast contents are optimized, and user experience and liveness are improved. The live comprehensive data set is subjected to random disturbance processing by the Laplace noise method, so that live noise data is obtained, the risk of user privacy disclosure can be reduced to a certain extent by adding noise disturbance, the personal privacy rights of users are protected, errors and deviations possibly exist in the live comprehensive data set, and the complexity degree of the data can be increased by adding proper noise disturbance, so that the data is more accurate and real. The live noise data is subjected to differential calculation, so that live differential data is obtained, the differential calculation can highlight the data difference, and the difference and characteristics among different users can be displayed more intuitively; the live differential data can further improve the utility of the data, and can more accurately find potential rules and modes in the data. Calculating live broadcast differential data and live broadcast noise data so as to obtain live broadcast intention score data of a user, wherein the live broadcast intention score data of the user can further evaluate interests and demands of the user more accurately according to live broadcast click times of the user, live broadcast access time, similar live broadcast frequency browsed by the user and other data, and realize more accurate grasp of the demands of the user; the pushing efficiency and the pushing accuracy can be improved, so that the user requirements are better met, and better service is provided. The live intention score data of the user is classified and calculated, so that live potential intention data is obtained, the interest, hobbies and demands of users of different categories can be found through the classified and calculated, the platform is helped to find out trends and hot spots of the demands of the users, the platform is guided to optimize an operation strategy, the data can be further divided, the complexity and difficulty of data analysis are effectively reduced, and the data analysis efficiency is improved.
Optionally, step S3 includes the steps of:
acquiring historical purchase data of a user;
constructing a user purchase intention formula according to the user historical purchase data and the user historical intention data, wherein the user purchase intention formula specifically comprises the following steps:
wherein Q is the purchase intention coefficient, N is the historical purchase times of the user, M is the historical purchase total amount of the user, V is the purchase probability of the user, T is the total browsing time of the user, Y is the total clicking times of the user, and alpha is the clicking similarity coefficient of the user;
the user purchase intention formula fully considers the user historical purchase times N, the user historical purchase total amount M, the user purchase probability V, the user total browsing time length T and the user total clicking times Y which influence the purchase intention coefficient, thereby forming The function relation of the user click intention coefficient is utilized, the user historical purchase times, the user historical purchase total amount, the user purchase probability, the user total browsing time length and the user total click times are utilized to calculate the purchase intention coefficient, the influence of similar browsing pages on the purchase intention coefficient is considered, the user click similarity coefficient alpha is added, the function relation is applicable to the calculation of the purchase intention coefficients of different crowds, and the accuracy and the authenticity of the purchase intention coefficient can be effectively improved.
And calculating the historical purchase data of the user and the historical purchase intention data of the user through the user purchase intention formula so as to obtain the historical purchase intention data of the user.
According to the method, the historical purchase data of the user are obtained through the cloud platform. The user purchase intention formula is constructed according to the user historical purchase data and the user historical intention data, and the purchase mode and the preference of the user can be further found and determined by constructing the user purchase intention formula, so that the efficiency and the accuracy of data analysis are improved, and better data support is provided for the platform. The historical purchase data and the historical purchase intention data of the user are calculated through the user purchase intention formula, so that the historical purchase intention data of the user is obtained, and the user purchase behaviors can be analyzed more comprehensively and carefully through calculating the historical purchase intention data of the user.
Optionally, step S5 specifically includes:
calculating the first intention evaluation data through a user purchase intention formula so as to obtain current intention data of the user;
clustering calculation is carried out on the historical purchase intention data of the user, so that user intention threshold value data are obtained;
and classifying and calculating the current intention data of the user based on the user intention threshold data, so as to obtain the data of the user which can develop.
According to the invention, the first intention evaluation data is calculated through the user purchase intention formula, so that the current intention data of the user is obtained, and the purchase intention and the requirement of the user can be evaluated more accurately through the calculated current intention data of the user, so that the quick grasp of the requirement of the user and more accurate push service are realized. Clustering calculation is carried out on historical purchase intention data of the users, so that user intention threshold data are obtained, the platform can find out which users have strong purchase intention through the clustering calculation, and more business opportunities are provided for the platform; the clustering calculation can group users according to purchase intention, so that a target user group with higher focal power is provided for the platform, and pushing precision is improved. The user current intention data is classified and calculated based on the user intention threshold data, so that the user data capable of developing is obtained, the user can be classified according to the user current intention data, the user with potential purchase intention can be found, the user intention is classified and calculated based on the user intention threshold data, personalized pushing and service to the user intention can be better realized, and the pushing effect is improved.
Optionally, the intent assessment in step S7 is specifically:
Acquiring recent browsing data of a user, and extracting features of the recent browsing data of the user so as to generate recent feature data of the user;
constructing a user portrait according to the user recent feature data so as to obtain user portrait data;
and classifying and calculating the developable user data by optimizing a second pushing model of the user based on the user portrait data so as to obtain second intention evaluation data.
According to the invention, the cloud platform is used for acquiring the recent browsing data of the user, and extracting the characteristics of the recent browsing data of the user, so that the recent characteristic data of the user is generated, and extracting the characteristics of the recent browsing data of the user, so that the interests and the demands of the user can be known, and the platform is helped to establish an accurate user portrait. User portraits are constructed according to the user recent feature data, so that user portrait data is obtained, interests and demands of users can be accurately analyzed based on the user portrait data, more accurate pushing is formulated for different user groups, and pushing effect and accuracy are improved. Based on the user portrait data, the user data can be classified and calculated through optimizing the second pushing model of the user, so that second intention evaluation data is obtained, the user data can be classified and calculated through optimizing the second pushing model of the user, relevant content can be pushed to the user more accurately, the pushing effect is improved, the user can be grouped based on the second intention evaluation data, corresponding preferential and marketing strategies are formulated according to the purchasing power, purchasing habit, personalized requirements and the like of the user, and the purchasing conversion rate is improved.
Optionally, the present specification further provides a big data analysis system for cloud service push, including:
at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor; wherein,,
the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the cloud service push big data analysis method of any of the above.
The cloud service pushing big data analysis system can realize any cloud service pushing big data analysis method, is used for combining operation among all devices and media of signal transmission to complete the cloud service pushing big data analysis method, and the internal structures of the system cooperate with each other, so that collection of user portraits is perfected, and accurate and effective content pushing for users is realized.
Drawings
Other features, objects and advantages of the application will become more apparent upon reading of the detailed description of a non-limiting implementation, made with reference to the accompanying drawings in which:
FIG. 1 is a flow chart of steps of a big data analysis method of cloud service push of the present application;
FIG. 2 is a detailed flowchart of the step S1 of the present invention;
the achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
The following is a clear and complete description of the technical method of the present patent in conjunction with the accompanying drawings, and it is evident that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, are intended to fall within the scope of the present invention.
Furthermore, the drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor methods and/or microcontroller methods.
It will be understood that, although the terms "first," "second," etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.
Referring to fig. 1, the present invention provides a big data analysis method for cloud service push, in this embodiment, the big data analysis method for cloud service push includes the following steps:
step S1: acquiring historical user browsing data, and extracting intent from the historical user data so as to generate user historical intent data;
according to the embodiment of the invention, the historical user browsing data is obtained through the cloud platform, the intention extraction is carried out on the historical user data, and the accuracy of the data is improved.
Step S2: scaling calculation is carried out on the historical intent data of the user, and a user pushing model is constructed;
In the embodiment of the invention, scaling calculation is carried out on the user history intention data by a standardized method, so that the calculated amount in data processing is reduced, scaling history intention data is obtained, model construction and model training are carried out on the scaling history intention data based on a random forest algorithm, and a user pushing model is obtained.
Step S3: acquiring user historical purchase data, and calculating the user historical purchase data and the user historical intention data so as to acquire the user historical purchase intention data;
according to the embodiment of the invention, the user historical purchase data is obtained through the cloud platform, and the user historical purchase data and the user historical intent data are calculated through the user purchase intent formula, so that the user historical purchase intent data is obtained.
Step S4: acquiring current purchase data of a user, and performing classification calculation on the current purchase data of the user based on historical purchase intention data of the user so as to generate first intention evaluation data;
according to the embodiment of the invention, the current purchase data of the user is obtained through the cloud platform, the classification model is constructed on the basis of the logistic regression algorithm, and the classification model is used for classifying and calculating the current purchase data of the user on the basis of the historical purchase intention data of the user, so that the first intention evaluation data is generated.
Step S5: calculating the first intent assessment data to obtain developable user data;
step S6: extracting features of historical user browsing data to obtain user browsing preference features and user browsing cost features, and correcting the user pushing model by using the user browsing preference features and the user browsing cost features to obtain an optimized user pushing model;
according to the embodiment of the invention, the historical user browsing data is subjected to feature extraction, so that the user browsing preference feature and the user browsing cost feature are obtained, the influence of redundant data on the model can be reduced, the user browsing preference feature and the user browsing cost feature are utilized to correct the user pushing model, so that an optimized user pushing model is obtained, the user data can be developed to perform intention assessment, the preference and intention of the user on the pushing content can be accurately judged, and the pushing efficiency and accuracy are improved.
Step S7: and carrying out intent assessment on the developable user data by using the optimized user pushing model so as to obtain second intent assessment data, and sending the pushing user data to a cloud service pushing system for content screening pushing processing.
According to the method and the device for achieving the user history intention data, the historical user browsing data are obtained through the cloud platform, the intention extraction is carried out on the historical user data, unnecessary pushing of the cloud platform can be reduced, the pushing effect is improved, and therefore the user history intention data are generated. Scaling calculation is carried out on the historical intent data of the user, and a user push model is built, the scaling calculation can unify the historical intent data with different dimensionalities, the variability among the data is reduced, the feature dimensionality of a data set can be reduced, the calculation amount in data processing is reduced, and therefore the user push model is built more accurately, and the accuracy of personalized recommendation is improved. The user historical purchase data is obtained, the user historical purchase data and the user historical intention data are calculated, so that the user historical purchase intention data is obtained, the user preference can be pushed more accurately by combining the user historical purchase data and the historical intention data, and the pushing accuracy and precision are improved. The method comprises the steps of obtaining current purchase data of a user, classifying and calculating the current purchase data of the user based on historical purchase intention data of the user, generating first intention evaluation data, classifying and calculating the current purchase data of the user based on the historical purchase intention data, and rapidly judging whether commodities purchased by the user at present meet the historical intention of the user or not, so that corresponding pushing is carried out, and shopping experience of the user is optimized. The first intention evaluation data is calculated to obtain the developable user data, and by calculating the first intention evaluation data, the developable user can be identified, which helps to concentrate on the user who is most likely to purchase or use the product or service, and improves the pushing accuracy. The method comprises the steps of extracting characteristics of historical user browsing data, obtaining user browsing preference characteristics and user browsing cost characteristics, correcting a user pushing model by utilizing the user browsing preference characteristics and the user browsing cost characteristics, obtaining an optimized user pushing model, better reflecting interests and preferences of a user by extracting the characteristics of the historical user browsing data, correcting the user pushing model by utilizing the characteristics, realizing personalized pushing more accurately, knowing interests and consumption requirements of the user more fully, adjusting and optimizing advertisement putting strategies, improving pushing effect and conversion rate, and realizing more accurate recommendation by optimizing and correcting the characteristics of the historical user browsing data and the user pushing model, thereby improving user liveness and user viscosity. The user data can be subjected to intention evaluation by utilizing the optimized user pushing model, so that second intention evaluation data are obtained, the pushed user data are sent to the cloud service pushing system for content screening pushing processing, the intention evaluation is performed on the user data can be performed by utilizing the optimized user pushing model, the preference and the intention of a user on pushing content can be judged more accurately, the pushing efficiency and the pushing accuracy are improved, the requirements and the interest points of the user can be better known, personalized pushing is realized, the user satisfaction degree and the loyalty degree are improved, the second intention evaluation data can help the platform to adjust pushing strategies, the content screening and pushing effect optimization is realized, the pushing effect and the user feedback rate are improved, and the user dislike rate is reduced.
Optionally, step S1 specifically includes:
step S11: acquiring historical user browsing data;
step S12: extracting website browsing data and live broadcasting browsing data from historical user browsing data, so as to generate the website browsing data and the live broadcasting browsing data;
step S13: performing website intention exploration processing on website browsing data so as to generate website intention data;
step S14: live broadcast intention exploration processing is carried out on live broadcast browsing data, so that live broadcast intention data are generated;
step S15: and carrying out time sequence combination according to the website intention data and the live intention data, thereby obtaining user intention combination data.
As an embodiment of the present invention, referring to fig. 2, a detailed implementation step flow diagram of step S1 in fig. 1 is shown, where in this embodiment, step S1 includes the following steps:
step S11: acquiring historical user browsing data;
in the embodiment of the invention, the historical user browsing data is acquired through the cloud platform.
Step S12: extracting website browsing data and live broadcasting browsing data from historical user browsing data, so as to generate the website browsing data and the live broadcasting browsing data;
in the embodiment of the invention, the characteristic extraction is carried out on the historical banking data, so that the website browsing data and the live broadcast browsing data are generated.
Step S13: performing website intention exploration processing on website browsing data so as to generate website intention data;
step S14: live broadcast intention exploration processing is carried out on live broadcast browsing data, so that live broadcast intention data are generated;
step S15: and carrying out time sequence combination according to the website intention data and the live intention data, thereby obtaining user intention combination data.
In the embodiment of the invention, the website intention data and the live broadcast intention data are combined according to the time sequence, so that the intention time sequence combined data is obtained, and the intention time sequence combined data is preprocessed, so that the user intention combined data is obtained.
According to the method and the device, the historical user browsing data are obtained through the cloud platform, the website browsing data extraction and the live broadcast browsing data extraction are carried out on the historical user browsing data, so that the website browsing data and the live broadcast browsing data are generated, the website browsing data extraction and the live broadcast browsing data extraction are carried out on the historical user browsing data, the browsing behaviors and the preferences of the user can be more comprehensively known, the accuracy and the precision of pushing and personalized services are improved, and the user experience is optimized. The website intention exploration processing is carried out on the website browsing data, so that the website intention data is generated, and commodities or services related to the interests of the user can be more accurately recommended by exploration of the website intention of the user, and the recommendation accuracy and the satisfaction degree of the user are improved. Live intention exploration processing is conducted on live browsing data, so that live intention data are generated, the live intention data can help the cloud platform to know potential demands and interest points of users, live topics and contents are pushed more accurately, and live marketing effects and commodity sales are improved. And carrying out time sequence combination according to the website intention data and the live intention data so as to obtain user intention combination data, and comprehensively analyzing the interests and the demands of the user by carrying out time sequence combination according to the website and the live intention data of the user so as to clearly know the intention and the tendency of the user and improve the accuracy of pushing and personalized services.
Optionally, the website consumption intention exploration process in step S13 includes the following steps:
carrying out statistical analysis on the website browsing data so as to obtain high-frequency intention data of the website;
in the embodiment of the invention, the website browsing data is statistically analyzed to obtain the website browsing statistical data, the website browsing statistical data is modeled and trained by utilizing a random forest algorithm to obtain a website high-frequency detection model, and the website browsing data is classified and calculated by the website high-frequency detection model to obtain the website high-frequency intention data.
Performing variance analysis on website browsing data so as to obtain website low-frequency intention data;
in the embodiment of the invention, the characteristic extraction is carried out on the website browsing data so as to obtain page access frequency data and stay time data, the average value and variance calculation are carried out on the page access frequency data so as to obtain page average access frequency data and page access frequency variance data, and the average value and variance calculation are carried out on the stay time data so as to obtain average stay time data and stay time variance data; and performing variance analysis on the page average access frequency data, the page access frequency variance data, the average residence time data and the residence time variance data, thereby obtaining the website low-frequency intention data.
And carrying out Pelson intention detection on the website browsing data so as to obtain the potential intention data of the website.
According to the invention, the website browsing data is statistically analyzed, so that the high-frequency intention data of the website is obtained, and the behavior and habit of the user can be better known and the data analysis capability and accuracy are improved through the statistical analysis of the website browsing data. And performing variance analysis on the website browsing data to obtain low-frequency intention data of the website, and performing variance analysis on the website browsing data to better know user behaviors and preferences and improve data analysis capability and accuracy. The Pelson intention detection is carried out on the website browsing data, so that the potential intention data of the website is obtained, the Pelson intention detection can find the unaware demands and interests of the user, the original pushing is supplemented and innovated, the user experience and satisfaction are improved, the preference and demands of the user can be better known, and therefore more effective and accurate personalized pushing service is realized.
Optionally, the pearson intent detection is specifically:
step S1331: extracting features of website browsing data to obtain website browsing time data, website page access amount data and user browsing similar page frequency data, and merging the data of the user browsing time data, the user page access amount data and the user browsing similar page frequency data to obtain a website comprehensive data set;
In the embodiment of the invention, the characteristic extraction is carried out on the website browsing data, so as to obtain the website browsing time data, the website page access amount data and the user browsing similar page frequency data, and the data merging is carried out on the user browsing time data, the user page access amount data and the user browsing similar page frequency data, so that a website comprehensive data set is obtained, and various data characteristics are subjected to the data merging, so that the accuracy and the accuracy of the data analysis are improved.
Step S1332: calculating the comprehensive data set through the pearson correlation coefficient so as to obtain pearson data;
in the embodiment of the invention, the pearson correlation coefficient is used for calculating the comprehensive data set, and the correlation between different data is calculated. Wherein the pearson correlation coefficient is an index for measuring the linear correlation between two continuous variables, and the value range of the pearson correlation coefficient is-1 to 1; the correlation coefficient is 1, the complete positive correlation is represented, the complete negative correlation is represented when the correlation coefficient is-1, and the no correlation is represented when the correlation coefficient is 0; thus, pearson data was obtained.
Step S1333: carrying out correlation screening on the pearson data so as to obtain potential characteristic data;
in the embodiment of the invention, the threshold value is set according to the pearson data, so that pearson threshold value data is obtained, the pearson data is subjected to correlation screening according to the pearson threshold value data, the variable with low correlation is eliminated, and only high-correlation characteristic data meeting the condition is left, so that potential characteristic data is obtained.
Step S1334: carrying out intention degree calculation on the potential characteristic data through association rule analysis so as to obtain intention degree data;
in the embodiment of the invention, the Apriori algorithm is used for extracting and analyzing the association relation of the potential feature data so as to obtain the potential association rule, and the association rule is subjected to support and confidence calculation so as to obtain the support and the confidence; and calculating the support and the confidence, so as to obtain the intention degree data.
Step S1335: carrying out maximum association clustering calculation on the intention data so as to obtain potential intention data of the website;
the intent maximum association clustering calculation specifically comprises the following steps:
model construction is carried out based on the intention data, so that an intention clustering model is obtained;
in the embodiment of the invention, the intent degree data is divided into a training set and a testing set according to the proportion of 7:3, and the training set and the testing set are modeled by using a Catboost algorithm, so that an intent degree clustering model is obtained.
Normalizing the intention data to obtain normalized intention data;
performing similarity matrix calculation on the normalized intent data through the pearson correlation coefficient so as to obtain a similarity matrix;
In the embodiment of the invention, the normalized intention degree data set is constructed into a matrix, the row vectors are users, the column vectors are characteristic, and the pearson correlation coefficient is used for calculating the similarity between different row vectors to obtain a similarity matrix.
Normalizing the similarity matrix to obtain a probability distribution matrix;
in the embodiment of the invention, the similarity matrix is normalized, so that a normalized similarity matrix is obtained, and the normalized similarity matrix is traversed through circulation, so that a probability distribution matrix is obtained.
Estimating and calculating the intent data according to the probability distribution matrix based on an EM algorithm, so as to obtain the cluster number;
in the embodiment of the invention, the probability distribution matrix is analyzed, so that the potential cluster number is obtained. Initializing the potential cluster number, and determining an initial mean value and a variance. Calculating the intention data to obtain the expected value of the hidden variable; the maximum likelihood function value is obtained by calculating the intent data. Calculating according to the initial mean and variance and the maximized likelihood function, so as to obtain a clustering mean and a clustering variance; and continuously iterating the EM algorithm by using the maximized likelihood function value, the clustering mean value and the clustering variance until convergence, so as to obtain the clustering cluster number.
Performing punishment item calculation on the cluster number so as to obtain maximum entropy probability distribution;
according to the embodiment of the invention, the intent degree data is analyzed, so that a penalty term weight is obtained, a probability distribution matrix is constrained by constructing a maximum entropy model according to the penalty term weight, the model parameters are updated by carrying out iterative optimization on the maximum entropy model so as to maximize likelihood function values, after the model converges, the probability that each sample point belongs to each cluster is calculated according to the parameters of the maximum entropy model, and the maximum entropy probability distribution is calculated according to the result.
Optimizing the intent clustering model based on the maximum entropy probability distribution so as to obtain an intent maximum association clustering model;
clustering calculation is carried out on the intent data through the intent maximum association clustering model, so that potential intent data of the website are obtained.
According to the method and the device for processing the web site browsing data, the feature extraction is carried out on the web site browsing data, so that the web site browsing time data, the web site page access amount data and the user browsing similar page frequency data are obtained, the data merging is carried out on the user browsing time data, the user page access amount data and the user browsing similar page frequency data, so that a web site comprehensive data set is obtained, various data features are merged, the user requirements and behaviors can be known more accurately, and therefore the accuracy and the accuracy of data analysis are improved. The comprehensive data set is calculated through the pearson correlation coefficient, so that pearson data is obtained, the unaware demands and interests of the user can be found through analysis of the pearson correlation coefficient, so that new services and products are developed, the evaluation and preference of the user to different features can be known through analysis of the pearson correlation coefficient, the platform design and interaction experience are optimized, the user experience and satisfaction are improved, the pearson correlation coefficient can help enterprises and merchants to better know the demands and preferences of the user, and the data analysis capability and accuracy are improved. And carrying out relevance screening on the pearson data to obtain potential characteristic data, and removing useless or low-relevance characteristic data through pearson relevance screening to reduce errors and interference, so that the relation and weight among different characteristic data are known more accurately. The intent degree is calculated on the potential characteristic data through the association rule analysis, so that the intent degree data is obtained, and the interest and the requirements of the user on different characteristic data can be more accurately known through the association rule analysis and the intent degree calculation. The intent degree data is subjected to maximum association clustering calculation so as to obtain potential intent degree data of websites, users can be more accurately and finely divided according to interests and requirements through the maximum association clustering calculation of the intent degree, so that requirements and characteristics of different user groups are better known, interests and requirements of different user groups can be better known through analyzing the potential intent degree data, and accordingly more effective and accurate personalized recommendation service is realized. Model construction is carried out based on the intention data, so that an intention clustering model is obtained; the intent data is normalized, so that normalized intent data is obtained, the data can be more comparable, deviation is reduced, algorithm precision is improved, data processing is facilitated, and data visualization effect is enhanced. The similarity matrix calculation is carried out on the normalized intention degree data through the pearson correlation coefficient, so that a similarity matrix is obtained, the correlation among different features can be judged more accurately through the pearson correlation coefficient calculation, and the similarity matrix can cluster and classify the different feature data according to the similarity, thereby being beneficial to more carefully analyzing and understanding the data and being used for correlation application. The similarity matrix is normalized, so that a probability distribution matrix is obtained, the normalization can limit the similarity matrix value between 0 and 1, so that huge difference of the similarity matrix value is avoided, and the stability of data is improved; the probability distribution matrix ensures that the sum of all values is 1, i.e. the probability values corresponding to each feature add together to be equal to 1. The method is beneficial to better calculating and analyzing the data and improving the accuracy of data analysis; the probability distribution matrix can better accord with the probability distribution rule, so that the data is more regular and easy to analyze. The method comprises the steps of carrying out estimation operation on intention data according to a probability distribution matrix based on an EM algorithm, so as to obtain a cluster number, carrying out estimation operation on the intention data according to the probability distribution matrix by the EM algorithm, removing noise data, retaining meaningful data, further accurately clustering characteristic data, and carrying out quick and efficient clustering operation based on the probability distribution matrix by the EM algorithm, so that the characteristic data clustering efficiency is improved. And carrying out punishment item calculation on the cluster number so as to obtain maximum entropy probability distribution, and carrying out punishment item calculation on the cluster number when calculating the maximum entropy probability distribution so as to ensure that the cluster result contains the least cluster number, thereby improving the quality of the cluster result. The intent clustering model is optimized based on the maximum entropy probability distribution, so that the intent maximum association clustering model is obtained, the maximum entropy probability distribution is adopted for model optimization, the risk of model overfitting can be effectively reduced, the clustering effect and reliability are improved, the accuracy of the model can be prevented from being influenced by excessive cluster numbers through the maximum entropy principle, and the selection of the cluster numbers is further optimized. The intent data is clustered through the intent maximum association clustering model, so that potential intent data of websites is obtained, users are divided into different clusters according to the intent through the intent maximum association clustering model, potential user groups and corresponding requirements of the potential user groups can be better mined, and potential clients are accurately positioned.
Optionally, the live consumption intention exploration process in step S14 includes the steps of:
s141: performing regression analysis on the live broadcast browsing data so as to obtain live broadcast high-frequency intention data;
in the embodiment of the invention, the live broadcast browsing data is subjected to characteristic extraction, so that live broadcast browsing characteristic data is obtained, and a linear regression model is established according to the live broadcast browsing characteristic data; and carrying out regression analysis on the live broadcast browsing data by using a linear regression model, thereby obtaining live broadcast high-frequency intention data.
S142: performing variance analysis on the live broadcast browsing data so as to obtain live broadcast low-frequency intention data;
in the embodiment of the invention, the live broadcast browsing data is subjected to characteristic extraction so as to obtain live broadcast access frequency data and live broadcast residence time data, the live broadcast access frequency data is subjected to average value and variance calculation so as to obtain live broadcast average access frequency data and live broadcast access frequency variance data, and the live broadcast residence time data is subjected to average value and variance calculation so as to obtain live broadcast average residence time data and live broadcast residence time variance data; and performing variance analysis on the live broadcast average access frequency data, the live broadcast access frequency variance data, the live broadcast average residence time data and the live broadcast residence time variance data, thereby obtaining live broadcast low-frequency intention data.
S143: and carrying out intention differential privacy detection on the live broadcast browsing data so as to obtain live broadcast potential intention data.
According to the method, regression analysis is carried out on the live broadcast browsing data, so that live broadcast high-frequency intention data is obtained, and regression analysis is carried out on the live broadcast browsing data, so that the live broadcast high-frequency intention data can be accurately predicted, preference and popular trend of live broadcast content can be found, operation decision is refined, and basis is provided for personalized pushing of users. The variance analysis is carried out on the live broadcast browsing data so as to obtain live broadcast low-frequency intention data, the variance analysis is carried out on the live broadcast browsing data, the variability among different audience groups can be distinguished, the live broadcast low-frequency intention data of the audience can be further subdivided, the diversity and the distribution condition of the audience needs can be better known through analyzing the variance of the live broadcast browsing data, so that the content coverage of the live broadcast can be better enlarged, the demands of different audiences can be met, the variance analysis can enable us to find out the instability factors existing in the live broadcast browsing data, and reference and basis are provided for us to adjust live broadcast content and pushing strategies, so that the quality and reliability of live broadcast are improved. The intention difference privacy detection is carried out on live broadcast browsing data, so that live broadcast potential intention data is obtained, the intention difference privacy detection is carried out on live broadcast browsing data, the privacy disclosure risk of a user in the data collection and analysis process can be reduced, the personal privacy rights of the user can be protected, the quality and the credibility of the live broadcast browsing data can be improved, and errors and deviations of the data due to privacy disclosure and other problems are avoided.
Optionally, the step of intentionally detecting differential privacy in step S143 is specifically:
feature extraction is carried out on live broadcast browsing data, so that live broadcast clicking frequency data, live broadcast access time data and user browsing similar live broadcast frequency data are obtained, and data coupling association is carried out on the live broadcast clicking frequency data, the live broadcast access time data and the user browsing similar live broadcast frequency data, so that a live broadcast comprehensive data set is obtained;
carrying out random disturbance processing on the live broadcast comprehensive data set by a Laplace noise method so as to obtain live broadcast noise data;
in the embodiment of the invention, the Laplace distribution is utilized to generate the noise data, and the noise data is added into the live broadcast comprehensive data set to obtain the live broadcast noise data set.
Performing differential calculation on the live noise data so as to obtain live differential data;
calculating live broadcast differential data and live broadcast noise data, so as to obtain live broadcast intention score data of a user;
and classifying and calculating the live broadcast intention score data of the user so as to obtain live broadcast potential intention data.
According to the method, characteristics of live broadcast browsing data are extracted, so that live broadcast clicking frequency data, live broadcast access time data and user browsing similar live broadcast frequency data are obtained, and data coupling association is performed on the live broadcast clicking frequency data, the live broadcast access time data and the user browsing similar live broadcast frequency data, so that a live broadcast comprehensive data set is obtained, the most critical few characteristics are extracted to effectively compress the data set, limited computing and storage resources are utilized to effectively manage the data so as to save cost, key information can be quickly identified from massive live broadcast browsing data through the characteristics extraction, and the data can be clearly displayed, so that understanding and analysis of the data become more efficient; and carrying out data coupling association on live click frequency data, live access time data and data of similar live frequency browsed by a user, comprehensively considering various information of the user, and acquiring more comprehensive user demand information. After the live broadcast comprehensive data set is obtained, deeper data analysis can be performed, live broadcast data is analyzed in multiple dimensions, user requirements and behavior modes are better known, push strategies and live broadcast contents are optimized, and user experience and liveness are improved. The live comprehensive data set is subjected to random disturbance processing by the Laplace noise method, so that live noise data is obtained, the risk of user privacy disclosure can be reduced to a certain extent by adding noise disturbance, the personal privacy rights of users are protected, errors and deviations possibly exist in the live comprehensive data set, and the complexity degree of the data can be increased by adding proper noise disturbance, so that the data is more accurate and real. The live noise data is subjected to differential calculation, so that live differential data is obtained, the differential calculation can highlight the data difference, and the difference and characteristics among different users can be displayed more intuitively; the live differential data can further improve the utility of the data, and can more accurately find potential rules and modes in the data. Calculating live broadcast differential data and live broadcast noise data so as to obtain live broadcast intention score data of a user, wherein the live broadcast intention score data of the user can further evaluate interests and demands of the user more accurately according to live broadcast click times of the user, live broadcast access time, similar live broadcast frequency browsed by the user and other data, and realize more accurate grasp of the demands of the user; the pushing efficiency and the pushing accuracy can be improved, so that the user requirements are better met, and better service is provided. The live intention score data of the user is classified and calculated, so that live potential intention data is obtained, the interest, hobbies and demands of users of different categories can be found through the classified and calculated, the platform is helped to find out trends and hot spots of the demands of the users, the platform is guided to optimize an operation strategy, the data can be further divided, the complexity and difficulty of data analysis are effectively reduced, and the data analysis efficiency is improved.
Optionally, step S3 includes the steps of:
acquiring historical purchase data of a user;
according to the embodiment of the invention, the historical purchase data of the user is obtained through the cloud platform.
Constructing a user purchase intention formula according to the user historical purchase data and the user historical intention data, wherein the user purchase intention formula specifically comprises the following steps:
wherein Q is the purchase intention coefficient, N is the historical purchase times of the user, M is the historical purchase total amount of the user, V is the purchase probability of the user, T is the total browsing time of the user, Y is the total clicking times of the user, and alpha is the clicking similarity coefficient of the user;
the user purchase intention formula fully considers the user historical purchase times N, the user historical purchase total amount M, the user purchase probability V, the user total browsing time length T and the user total clicking times Y which influence the purchase intention coefficient, thereby forming The function relation of the user click intention coefficient is utilized, the user historical purchase times, the user historical purchase total amount, the user purchase probability, the user total browsing time length and the user total click times are utilized to calculate the purchase intention coefficient, the influence of similar browsing pages on the purchase intention coefficient is considered, the user click similarity coefficient alpha is added, the function relation is applicable to the calculation of the purchase intention coefficients of different crowds, and the accuracy and the authenticity of the purchase intention coefficient can be effectively improved.
And calculating the historical purchase data of the user and the historical purchase intention data of the user through the user purchase intention formula so as to obtain the historical purchase intention data of the user.
According to the method, the historical purchase data of the user are obtained through the cloud platform. The user purchase intention formula is constructed according to the user historical purchase data and the user historical intention data, and the purchase mode and the preference of the user can be further found and determined by constructing the user purchase intention formula, so that the efficiency and the accuracy of data analysis are improved, and better data support is provided for the platform. The historical purchase data and the historical purchase intention data of the user are calculated through the user purchase intention formula, so that the historical purchase intention data of the user is obtained, and the user purchase behaviors can be analyzed more comprehensively and carefully through calculating the historical purchase intention data of the user.
Optionally, step S5 specifically includes:
calculating the first intention evaluation data through a user purchase intention formula so as to obtain current intention data of the user;
clustering calculation is carried out on the historical purchase intention data of the user, so that user intention threshold value data are obtained;
according to the embodiment of the invention, model construction is carried out according to the historical purchase intention data of the user based on a K-means algorithm, so that an intention clustering model is obtained, and clustering calculation is carried out on the historical purchase intention data of the user based on the intention clustering model, so that user intention threshold data is obtained.
And classifying and calculating the current intention data of the user based on the user intention threshold data, so as to obtain the data of the user which can develop.
In the embodiment of the invention, the model construction is performed according to the current intention data of the user based on the logistic regression algorithm, so that an intention classification model is obtained, and the intention classification model is utilized to perform classification calculation on the current intention data of the user based on the user intention threshold data, so that the user data capable of developing is obtained.
According to the invention, the first intention evaluation data is calculated through the user purchase intention formula, so that the current intention data of the user is obtained, and the purchase intention and the requirement of the user can be evaluated more accurately through the calculated current intention data of the user, so that the quick grasp of the requirement of the user and more accurate push service are realized. Clustering calculation is carried out on historical purchase intention data of the users, so that user intention threshold data are obtained, the platform can find out which users have strong purchase intention through the clustering calculation, and more business opportunities are provided for the platform; the clustering calculation can group users according to purchase intention, so that a target user group with higher focal power is provided for the platform, and pushing precision is improved. The user current intention data is classified and calculated based on the user intention threshold data, so that the user data capable of developing is obtained, the user can be classified according to the user current intention data, the user with potential purchase intention can be found, the user intention is classified and calculated based on the user intention threshold data, personalized pushing and service to the user intention can be better realized, and the pushing effect is improved.
Optionally, the intent assessment in step S7 is specifically:
acquiring recent browsing data of a user, and extracting features of the recent browsing data of the user so as to generate recent feature data of the user;
according to the embodiment of the invention, the recent browsing data of the user is obtained through the cloud platform, and the characteristic extraction is carried out on the recent browsing data of the user, so that the recent characteristic data of the user is generated.
Constructing a user portrait according to the user recent feature data so as to obtain user portrait data;
according to the embodiment of the invention, the user characteristic vector is obtained by processing the recent characteristic data of the user, and the user characteristic vector is classified and calculated, so that the user portrait data is obtained.
And classifying and calculating the developable user data by optimizing a second pushing model of the user based on the user portrait data so as to obtain second intention evaluation data.
According to the invention, the cloud platform is used for acquiring the recent browsing data of the user, and extracting the characteristics of the recent browsing data of the user, so that the recent characteristic data of the user is generated, and extracting the characteristics of the recent browsing data of the user, so that the interests and the demands of the user can be known, and the platform is helped to establish an accurate user portrait. User portraits are constructed according to the user recent feature data, so that user portrait data is obtained, interests and demands of users can be accurately analyzed based on the user portrait data, more accurate pushing is formulated for different user groups, and pushing effect and accuracy are improved. Based on the user portrait data, the user data can be classified and calculated through optimizing the second pushing model of the user, so that second intention evaluation data is obtained, the user data can be classified and calculated through optimizing the second pushing model of the user, relevant content can be pushed to the user more accurately, the pushing effect is improved, the user can be grouped based on the second intention evaluation data, corresponding preferential and marketing strategies are formulated according to the purchasing power, purchasing habit, personalized requirements and the like of the user, and the purchasing conversion rate is improved.
Optionally, the present specification further provides a big data analysis system for cloud service push, including:
at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor; wherein,,
the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the cloud service push big data analysis method of any of the above.
The cloud service pushing big data analysis system can realize any cloud service pushing big data analysis method, is used for combining operation among all devices and media of signal transmission to complete the cloud service pushing big data analysis method, and the internal structures of the system cooperate with each other, so that collection of user portraits is perfected, and accurate and effective content pushing for users is realized.
The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.
The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. The big data analysis method for cloud service pushing is characterized by comprising the following steps of:
step S1: acquiring historical user browsing data, and extracting intent from the historical user data so as to generate user historical intent data;
step S2: scaling calculation is carried out on the historical intent data of the user, and a user pushing model is constructed;
step S3: acquiring user historical purchase data, and calculating the user historical purchase data and the user historical intention data so as to acquire the user historical purchase intention data;
step S4: acquiring current purchase data of a user, and performing classification calculation on the current purchase data of the user based on historical purchase intention data of the user so as to generate first intention evaluation data;
Step S5: calculating the first intent assessment data to obtain developable user data;
step S6: extracting features of historical user browsing data to obtain user browsing preference features and user browsing cost features, and correcting the user pushing model by using the user browsing preference features and the user browsing cost features to obtain an optimized user pushing model;
step S7: and carrying out intent assessment on the developable user data by using the optimized user pushing model so as to obtain second intent assessment data, and sending the pushing user data to a cloud service pushing system for content screening pushing processing.
2. The method of claim 1, wherein the user history intent data includes website intent data, live intent data, and user intent merge data, and step S1 includes the steps of:
step S11: acquiring historical user browsing data;
step S12: extracting website browsing data and live broadcasting browsing data from historical user browsing data, so as to generate the website browsing data and the live broadcasting browsing data;
step S13: performing website intention exploration processing on website browsing data so as to generate website intention data;
Step S14: live broadcast intention exploration processing is carried out on live broadcast browsing data, so that live broadcast intention data are generated;
step S15: and carrying out time sequence combination according to the website intention data and the live intention data, thereby obtaining user intention combination data.
3. The method according to claim 2, wherein the website consumption intention exploration process in step S13 includes the steps of:
carrying out statistical analysis on the website browsing data so as to obtain high-frequency intention data of the website;
performing variance analysis on website browsing data so as to obtain website low-frequency intention data;
and carrying out Pelson intention detection on the website browsing data so as to obtain the potential intention data of the website.
4. A method according to claim 3, wherein the step of pilson intent detection is specifically:
step S1331: extracting features of website browsing data to obtain website browsing time data, website page access amount data and user browsing similar page frequency data, and merging the data of the user browsing time data, the user page access amount data and the user browsing similar page frequency data to obtain a website comprehensive data set;
step S1332: calculating the comprehensive data set through the pearson correlation coefficient so as to obtain pearson data;
Step S1333: carrying out correlation screening on the pearson data so as to obtain potential characteristic data;
step S1334: carrying out intention degree calculation on the potential characteristic data through association rule analysis so as to obtain intention degree data;
step S1335: carrying out maximum association clustering calculation on the intention data so as to obtain potential intention data of the website;
the intent maximum association clustering calculation specifically comprises the following steps:
model construction is carried out based on the intention data, so that an intention clustering model is obtained;
normalizing the intention data to obtain normalized intention data;
performing similarity matrix calculation on the normalized intent data through the pearson correlation coefficient so as to obtain a similarity matrix;
normalizing the similarity matrix to obtain a probability distribution matrix;
estimating and calculating the intent data according to the probability distribution matrix based on an EM algorithm, so as to obtain the cluster number;
performing punishment item calculation on the cluster number so as to obtain maximum entropy probability distribution;
optimizing the intent clustering model based on the maximum entropy probability distribution so as to obtain an intent maximum association clustering model;
clustering calculation is carried out on the intent data through the intent maximum association clustering model, so that potential intent data of the website are obtained.
5. The method according to claim 4, wherein the live consumption intention probe process in step S14 includes the steps of:
s141: performing regression analysis on the live broadcast browsing data so as to obtain live broadcast high-frequency intention data;
s142: performing variance analysis on the live broadcast browsing data so as to obtain live broadcast low-frequency intention data;
s143: and carrying out intention differential privacy detection on the live broadcast browsing data so as to obtain live broadcast potential intention data.
6. The method according to claim 5, wherein the step of intentionally detecting differential privacy in step S143 is specifically:
feature extraction is carried out on live broadcast browsing data, so that live broadcast clicking frequency data, live broadcast access time data and user browsing similar live broadcast frequency data are obtained, and data coupling association is carried out on the live broadcast clicking frequency data, the live broadcast access time data and the user browsing similar live broadcast frequency data, so that a live broadcast comprehensive data set is obtained;
carrying out random disturbance processing on the live broadcast comprehensive data set by a Laplace noise method so as to obtain live broadcast noise data;
performing differential calculation on the live noise data so as to obtain live differential data;
calculating live broadcast differential data and live broadcast noise data, so as to obtain live broadcast intention score data of a user;
And classifying and calculating the live broadcast intention score data of the user so as to obtain live broadcast potential intention data.
7. The method according to claim 6, wherein step S3 comprises the steps of:
acquiring historical purchase data of a user;
constructing a user purchase intention formula according to the user historical purchase data and the user historical intention data, wherein the user purchase intention formula specifically comprises the following steps:
wherein Q is the purchase intention coefficient, N is the historical purchase times of the user, M is the historical purchase total amount of the user, V is the purchase probability of the user, T is the total browsing time of the user, Y is the total clicking times of the user, and alpha is the clicking similarity coefficient of the user;
and calculating the historical purchase data of the user and the historical purchase intention data of the user through the user purchase intention formula so as to obtain the historical purchase intention data of the user.
8. The method according to claim 7, wherein step S5 is specifically:
calculating the first intention evaluation data through a user purchase intention formula so as to obtain current intention data of the user;
clustering calculation is carried out on the historical purchase intention data of the user, so that user intention threshold value data are obtained;
and classifying and calculating the current intention data of the user based on the user intention threshold data, so as to obtain the data of the user which can develop.
9. The method according to claim 8, wherein the intent assessment in step S7 is specifically:
acquiring recent browsing data of a user, and extracting features of the recent browsing data of the user so as to generate recent feature data of the user;
constructing a user portrait according to the user recent feature data so as to obtain user portrait data;
and classifying and calculating the developable user data by optimizing a second pushing model of the user based on the user portrait data so as to obtain second intention evaluation data.
10. The big data analysis system of cloud service propelling movement, characterized by comprising:
at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor;
the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the cloud service push big data analysis method of any of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310574565.XA CN116595255A (en) | 2023-05-20 | 2023-05-20 | Big data analysis method and system for cloud service pushing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310574565.XA CN116595255A (en) | 2023-05-20 | 2023-05-20 | Big data analysis method and system for cloud service pushing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116595255A true CN116595255A (en) | 2023-08-15 |
Family
ID=87593373
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310574565.XA Withdrawn CN116595255A (en) | 2023-05-20 | 2023-05-20 | Big data analysis method and system for cloud service pushing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116595255A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116862625A (en) * | 2023-09-05 | 2023-10-10 | 武汉森全科技有限公司 | Online recommendation method for fresh fruits based on Internet big data |
| CN117557346A (en) * | 2024-01-11 | 2024-02-13 | 华高数字科技有限公司 | Full-link intelligent business decision analysis method based on dynamic consumption data |
| CN118250516A (en) * | 2024-05-27 | 2024-06-25 | 长沙艾珀科技有限公司 | Hierarchical processing method for users |
-
2023
- 2023-05-20 CN CN202310574565.XA patent/CN116595255A/en not_active Withdrawn
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116862625A (en) * | 2023-09-05 | 2023-10-10 | 武汉森全科技有限公司 | Online recommendation method for fresh fruits based on Internet big data |
| CN116862625B (en) * | 2023-09-05 | 2023-11-21 | 武汉森全科技有限公司 | Online recommendation method for fresh fruits based on Internet big data |
| CN117557346A (en) * | 2024-01-11 | 2024-02-13 | 华高数字科技有限公司 | Full-link intelligent business decision analysis method based on dynamic consumption data |
| CN117557346B (en) * | 2024-01-11 | 2024-04-02 | 华高数字科技有限公司 | Full-link intelligent business decision analysis method based on dynamic consumption data |
| CN118250516A (en) * | 2024-05-27 | 2024-06-25 | 长沙艾珀科技有限公司 | Hierarchical processing method for users |
| CN118250516B (en) * | 2024-05-27 | 2024-08-30 | 长沙艾珀科技有限公司 | Hierarchical processing method for users |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Qiu et al. | Predicting customer purchase behavior in the e-commerce context | |
| Hasan et al. | Dominance of AI and Machine Learning Techniques in Hybrid Movie Recommendation System Applying Text-to-number Conversion and Cosine Similarity Approaches | |
| Liu et al. | Finding the right consumer: optimizing for conversion in display advertising campaigns | |
| Wang et al. | Improving neighbor-based collaborative filtering by using a hybrid similarity measurement | |
| Almazro et al. | A survey paper on recommender systems | |
| Lee et al. | Web personalization expert with combining collaborative filtering and association rule mining technique | |
| US8775416B2 (en) | Adapting a context-independent relevance function for identifying relevant search results | |
| US8412648B2 (en) | Systems and methods of making content-based demographics predictions for website cross-reference to related applications | |
| Li et al. | Interest-based real-time content recommendation in online social communities | |
| Hu et al. | Bayesian personalized ranking based on multiple-layer neighborhoods | |
| CN116595255A (en) | Big data analysis method and system for cloud service pushing | |
| Choi et al. | A hybrid recommendation method with reduced data for large-scale application | |
| Ead et al. | Intelligent Systems of Machine Learning Approaches for developing E-services portals | |
| Huang et al. | Handling sequential pattern decay: Developing a two-stage collaborative recommender system | |
| US20110119209A1 (en) | Method and system for developing a classification tool | |
| CN107066589B (en) | Entity semantics and word frequency ordering method and device based on comprehensive knowledge | |
| Kouadria et al. | A multi-criteria collaborative filtering recommender system using learning-to-rank and rank aggregation | |
| Gras et al. | Identifying grey sheep users in collaborative filtering: a distribution-based technique | |
| Moradi et al. | A trust-aware recommender algorithm based on users overlapping community structure | |
| Ma et al. | A sub-linear, massive-scale look-alike audience extension system a massive-scale look-alike audience extension | |
| Wang et al. | Low-rank and sparse matrix factorization with prior relations for recommender systems | |
| Suchacka et al. | A k-nearest neighbors method for classifying user sessions in e-commerce scenario | |
| Li et al. | Sentiment based multi-index integrated scoring method to improve the accuracy of recommender system | |
| Moneera et al. | Click through Rate Effectiveness Prediction on Mobile Ads Using Extreme Gradient Boosting. | |
| Ulian et al. | Exploring the effects of different Clustering Methods on a News Recommender System |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20230815 |
|
| WW01 | Invention patent application withdrawn after publication |