CN110807051B - Context-aware real-time service recommendation method based on cloud platform - Google Patents
Context-aware real-time service recommendation method based on cloud platform Download PDFInfo
- Publication number
- CN110807051B CN110807051B CN201911064443.6A CN201911064443A CN110807051B CN 110807051 B CN110807051 B CN 110807051B CN 201911064443 A CN201911064443 A CN 201911064443A CN 110807051 B CN110807051 B CN 110807051B
- Authority
- CN
- China
- Prior art keywords
- context
- database
- user
- service
- sliding window
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000005065 mining Methods 0.000 claims abstract description 21
- 230000008569 process Effects 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012163 sequencing technique Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000000638 solvent extraction Methods 0.000 claims description 2
- 230000006872 improvement Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 4
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2465—Query processing support for facilitating data mining operations in structured databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/55—Push-based network services
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Data Mining & Analysis (AREA)
- Signal Processing (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Remote Sensing (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a context-aware real-time service recommendation method based on a cloud platform, belonging to the technical field of service information recommendation methods; the technical problem to be solved is as follows: the improvement of a context-aware real-time recommendation service method based on a cloud platform is provided; the technical scheme for solving the technical problem is as follows: the method comprises the following steps: defining a context model by adopting a vector model, mining association rules of data, mining the association rules and recommending matching services by adopting a distributed association rule mining method based on a sliding window: the invention is applied to service information recommendation.
Description
Technical Field
The invention discloses a context-aware real-time service recommendation method based on a cloud platform, and belongs to the technical field of service information recommendation methods.
Background
The rapid development of information network technology and the wide popularization of mobile intelligent terminals enable the application of mobile internet to be favored by people, mass data are generated in information service, and users have higher personalized requirements on mobile service quality in the face of mass mobile service resources in a dynamic network environment.
The current mobile application service can not add or delete service according to the actual needs of users, can be flexibly adjusted, and when the user actually experiences, the user usually depends on the subjective experience of the user, the user actively selects and triggers the service, and when the mobile user uses the internet, its expected behavior is often closely related to contextual information such as time, location, surrounding people, activity status, network conditions, for example, some users prefer to browse the news when they get up in the morning rather than going to bed at night, some users prefer to shop online in the office rather than going home from work, therefore, the context information of the user is fully integrated into the mobile service recommendation process, the current actual requirements of the user can be better adapted, the user experience is improved, these adaptation functions are not available to current network operation providers.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to solve the technical problems that: the improvement of the context-aware real-time recommendation service method based on the cloud platform is provided.
In order to solve the technical problems, the invention adopts the technical scheme that: a method for context-aware real-time recommendation service based on a cloud platform comprises the following steps:
the method comprises the following steps: the context model is defined by using a vector model as follows:
context model: c ═ C1,C2,…,Cn);
In the above formula, C is a context consisting of n context attributes, CiRepresents a specific type of context, wherein ciAs a context attribute Ci1, 2, …, n, the specific value of each set of attributes is an example of a context C, which is expressed as: c ═ c1,c2,...,cn);
Step two: after defining the context, mining association rules of the data, and adopting a distributed association rule mining method based on a sliding window:
recording a context C of a user and a selected service S as an object, wherein an association rule needs to dig out a rule with stronger association between the context C and the selected service S and is recorded as C → S;
when the user is in a context environment C ', constructing a combinable candidate service set which accords with C', matching C and C 'in all rules, and providing selectable services for the user in the context of C';
the steps of the matching process of the mining of the association rules and the service recommendation are as follows:
step 2.1: traversing a historical database of a cloud end, horizontally dividing and storing data in Spark RDD, locally counting respective data pieces by each node, and summarizing by a reduce to obtain a frequent 1-item set FList after all the data pieces are calculated;
step 2.2: partitioning the Flist, and solving the problem of load balancing of the Fp-Growth algorithm by adopting a polynomial time degree algorithm;
taking the support degree of the frequent item set as an index of load measurement, keeping the running time of all nodes approximately consistent, dividing Flist into k groups, and the average time of each group isIn the above formula, m is the number of frequent items, sup (i)(q)) For the q-th item i(q)Such that the processing load of each node is approximately
Grouping the items in the Flist to form a projection database of each group, and broadcasting the database to each corresponding node;
step 2.3: dividing the projection database into k data blocks with equal size for each node to obtain a basic sliding window:
BSW1,BSW2......BSWk;
performing a second scan on the grouped projection database on each node;
step 2.4: constructing a pattern tree by using an expanded FP-Growth algorithm for each node, and designing the support degree for each item:
setting relatively low support for important but low-frequency contexts;
setting a relatively high degree of support for contexts that are not important but occur frequently;
each node establishes a mode tree of the node;
step 2.5: based on the basic sliding window obtained in the step 2.3, performing context updated incremental updating mining:
when the database is newly added, taking the newly added transaction as a new basic sliding window, wherein the size of the window is equal to that of the basic sliding window of the original database;
obtaining new basic sliding window BSW based on expanded Fp-Growthm+1And the last sliding window BSWmThe frequent item set is recorded as F1, the current frequent item set is F2, the new frequent item set is compared with the frequent item set in the original database, and the pattern tree is updated according to the following conditions:
case one, when the frequent entries exist in both the original database and the newly added database at F1 n F2, adding the counts of the two entry sets, and then updating the pattern tree;
in case II, when the frequent item exists in the original database but does not exist in the newly added database F1- (F1 n F2), the support count of the item in the newly added database needs to be estimated according to the similarity, and the estimated value is added to the original frequent item set count, and then the mode tree is updated;
case three, the frequent item does not exist in the original database but exists in the newly added database F2- (F1 n F2), it is necessary to look up whether there is the item in the candidate information table, if there is the item, it directly reads its count, if there is not, it is necessary to estimate the count of the item in the original basic sliding window according to the similarity, and add the estimated value and the current frequent item set count, then update the mode tree;
the similarity estimation in the second case and the third case is calculated by using a locality sensitive hashing algorithm:
the last sliding window BSWmDefined as Fwi, with dimension n, its hamming operation is changed to a vector x' with dimension n;
then, the new basic sliding window BSW is usedmCorrespond toDefined as Fwj, its hamming operation is changed into a vector y of dimension n;
step 2.6: and (3) performing service matching based on the calculation result of the step 2.5:
in the mode trees of all nodes, each path contains a context C and a service set S corresponding to the context, when a mobile phone of a user returns a context C ' where the current user is located, a service set conforming to the current context C ' needs to be constructed through the current user context C ', and the step of constructing the service set is as follows:
step 2.6.1: processing data of a context c ' transmitted by a user, filtering items which are not in the Flist in the context c ', and sequencing the rest items which are not filtered in a descending order according to the support degree of the Flist to form a new context c ';
step 2.6.2: hamming a new context c ' to be a vector X ' with the dimension of n, extracting t bits from the vector X ' to be used as a hash value of the vector X ', comparing the t bits with the hash values of the processed k groups of nodes, and sending the context c ' to the nodes with higher similarity;
step 2.6.3: and (3) solving intersection of the context c' and the path p in the mode tree of the node, and endowing each path with different weight coefficients according to the intersection result, wherein the specific weight calculation formula is as follows:
step 2.6.4: recommending services to the user according to the weight, sequencing the services according to the support degree, recommending the first services with the maximum support degree to the user, and recording the set of the services as s;
step 2.6.5: and returning the c → s event data selected by the user to the cloud end and storing the c → s event data in a cloud end database for next mining.
Compared with the prior art, the invention has the beneficial effects that: the invention provides a real-time service recommendation method based on a cloud platform, which can capture the current personalized requirements of users in time and generate near real-time recommendation for massive online users, is more sensitive to contextual information such as time, scene, position and the like according to the characteristics of mobile application and portability, and can provide personalized services such as peripheral catering, leisure, entertainment and the like for the users by sensing the information of the users; based on the cloud platform, the invention utilizes the historical records of the service selected by the user in different context environments to dig out the association relationship between the context and the service, and provides the service set available in the current environment for the user by sensing the change of the context, thereby improving the pertinence and timeliness of service recommendation and further improving the user experience of the mobile application service.
Drawings
The invention is further described below with reference to the accompanying drawings:
FIG. 1 is a schematic diagram of the system of the present invention;
FIG. 2 is a diagram of a system for recommending services based on association rules according to the present invention;
FIG. 3 is a packet topology of the FP-growth algorithm of the present invention;
FIG. 4 is a diagram of a schema tree structure constructed using the extended FP-Growth algorithm of the present invention.
Detailed Description
The method adopts a cloud platform to store and process data for the processing of the current massive mobile service resources and the service recommendation of massive online users, the system architecture diagram is as shown in figure 1, aiming at the service recommendation, the method adopts an association rule supporting a dynamic data set to update and mine a distributed algorithm, and as the algorithm is combined with the context environment of the user, namely, the current situation of the user and the requirements of the user are continuously improved, the requirements of the client are gradually added into the current situation, only possible service options are provided for the user according to the current situation of the user; the invention adopts the association rule updating algorithm based on the sliding window, pays more attention to the personalized expression and the instant demand of the current user in a period of time, gradually adds new services to the current application, and the application does not need to integrate excessive services in advance.
Another key point of the context-aware user service recommendation adopted by the invention is to establish a correlation between a user context and a candidate service, and then find a service set recommended to the user according to the correlation, so that in the execution process of the distributed algorithm, the current browsing or purchasing records and the like of the user (namely the context where the user is located) are used as the front part of a rule, the service capable of being provided for the user is used as the back part of the rule to form a correlation rule, and the rule is used for recommending the selectable service set under the current application scene to the current user, so as to provide more intelligent help and support for the selection and use of services for massive online users, wherein the recommendation process is shown in fig. 2.
The method realizes that the instant service recommendation judges the current state of the user through the context, and adopts a vector model to define the context model as follows in order to adapt to the subsequent rule mining:
context model: c ═ C1,C2,…,Cn);
Where C is a context consisting of n context attributes, Ci(i ═ 1, 2, …, n) represents a particular type of context, such as current time, user age, location, etc.
The specific value of each set of attributes is an example of the context C, and is expressed as:
c=(c1,c2,…,cn),(i=1,2,…,n);
for example: four values are defined (11:00 am, at home, …, listen to music) to indicate that the user selected a certain mobile service while listening to music at home in the morning.
After defining the context, mining association rules of the data, and adopting a distributed association rule mining method based on a sliding window; recording the context C of the user and the selected service S as one thing, wherein the association rule needs to dig out a rule with stronger association between the context C and the service S and is recorded as C → S, and when the user is in the context environment C ', constructing a combinable candidate service set conforming to the context C', namely matching the context C and the context S 'in all the rules, so that the user can be provided with the selectable service in the context C'.
The matching process of mining of the association rules and service recommendation comprises the following steps:
step one, traversing a cloud historical database, horizontally dividing and storing data in Spark RDD, locally counting respective data pieces by each node, and summarizing by reduce to obtain a frequent 1-item set FList after all the data pieces are calculated.
And secondly, segmenting the Flist, wherein the load balancing problem of the distributed FP-Growth algorithm is met because the running time of the distributed FP-Growth algorithm depends on a node which completes a task at the latest, and the load balancing problem of the Fp-Growth algorithm is solved by adopting an algorithm using polynomial time.
The support degree of the frequent item set is used as an index of load measurement, because the higher the support degree is, the item where the support degree is located has the highest frequency in all the items in the whole, the time occupied by processing is the longest, and the aim is to keep the running time of all the nodes approximately consistent as much as possible.
Divide Flist into k groups, each group having an average load ofIn the above formula, m is the number of frequent items, sup (i)(q)) For the q-th item i(q)Is to make the processing load of each node approximately equal
According to the idea, after the items in the Flist are grouped, a projection database of each group is formed and is broadcasted to each corresponding node, and the process is shown in FIG. 3;
step three, each node divides the projection database into k data blocks with equal size to obtain a basic sliding window: BSW1,BSW2......BSWk(ii) a The division is mainly used for the incremental updating process in the step five.
Performing a second scan on the grouped projection database on each node;
and step four, constructing a pattern tree by using the expanded FP-Growth algorithm for each node, wherein in actual application, the traditional algorithm causes data item loss with low frequency, and further influences the mining result. The main reason for this problem is that the conventional FP-Growth algorithm uses a uniform support, and automatically filters some items that appear less frequently. In order to solve the problem, a unified support degree method is not adopted, and the support degree of each item is designed in advance, so that the problem that mining results are inaccurate due to the fact that some important contexts are lost is solved. A relatively low support is set for contexts that are important but occur less frequently, and a relatively high support is set for contexts that are not important but occur more frequently, each node builds a pattern tree of its own node, as shown in fig. 4.
Step five, an incremental updating process based on the sliding window:
when the database is newly added, taking the newly added transaction as a new basic sliding window, wherein the size of the window is equal to that of the basic sliding window of the original database;
obtaining a new basic sliding window BSW according to Fp-Growthm+1And the last sliding window (i.e. the last window BSW of the original mining process)m) Comparing the new frequent item set with the frequent item set in the original database, three situations can occur, namely, the original frequent item set is recorded as F1, and the current frequent item set is recorded as F2.
Case one, when the frequent entries exist in both the original database and the newly added database (F1 ≠ F2), the counts of the two entry sets are added, and then the pattern tree is updated;
in case II, when the frequent item exists in the original database but does not exist in the newly added database F1- (F1 n F2), the support count of the item in the newly added database needs to be estimated according to the similarity, and the estimated value is added to the original frequent item set count, and then the mode tree is updated;
case three, the frequent item does not exist in the original database but exists in the newly added database F2- (F1 n F2), it is necessary to look up whether there is the item in the candidate information table, if there is the item, it directly reads its count, if there is not, it is necessary to estimate the count of the item in the original basic sliding window according to the similarity, and add the estimated value and the current frequent item set count, then update the mode tree;
the similarity estimation in the second case and the third case is calculated by using a locality sensitive hashing algorithm:
the last sliding window BSWmDefined as Fwi, with dimension n, its hamming operation is changed to a vector x' with dimension n;
then, the new basic sliding window BSW is usedm+1Defining Fwj as the corresponding high-dimensional vector, and converting the Hamming operation into a vector y with the dimension n;
step six, performing service matching based on the calculation result: in the mode tree of all nodes, each path contains a context C and a service S in the context ring, and when the mobile phone of the user returns a context C ' where the current user is located, a service set conforming to the current context C ' needs to be constructed through the current user context C '.
The steps of constructing the service set specifically are as follows:
(1) processing a context c ' transmitted by a user, filtering items which are not in the Flist in the context c ', and sequencing the rest items which are not filtered in a descending order according to the support degree of the Flist, thereby forming a new context c ';
(2) hamming a new context c 'to form a vector X' with dimension n, extracting t bits from the X 'as a hash value, comparing the hash value with the hash values of the processed k groups of nodes, and sending the c' to the nodes with higher similarity;
(3) and the context c 'and the path p in the mode tree of the node where the context c' is located calculate intersection, each path is endowed with different weight coefficients according to the intersection result, the weight of the subset path is the maximum, then the ultra-true subset path is followed, finally the path is the cross path, and the irrelevant path has no meaning and is not necessarily endowed with grade.
The specific weight calculation formula is as follows:
(4) recommending services to the user according to the weight, sequencing the services according to the support degree, and recommending the first services with the maximum support degree to the user;
(5) and returning the c → s event data selected by the user to the cloud end and storing the c → s event data in a cloud end database for next mining.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (1)
1. A method for context-aware real-time service recommendation based on a cloud platform is characterized by comprising the following steps: the method comprises the following steps:
the method comprises the following steps: the context model is defined by using a vector model as follows:
context model: c ═ C1,C2,…,Cn);
In the above formula, C is a context consisting of n context attributes, CiRepresents a specific type of context, wherein ciAs a context attribute Ci1, 2, …, n, the specific value of each set of attributes is an example of a context C, which is expressed as: c ═ c1,c2,...,cn);
Step two: after defining the context, mining association rules of the data, and adopting a distributed association rule mining method based on a sliding window:
recording a context C of a user and a selected service S as an object, wherein an association rule needs to dig out a rule with stronger association between the context C and the selected service S and is recorded as C → S;
when the user is in a context environment C ', constructing a combinable candidate service set which accords with C', matching C and C 'in all rules, and providing selectable services for the user in the context of C';
the steps of the matching process of the mining of the association rules and the service recommendation are as follows:
step 2.1: traversing a historical database of a cloud end, horizontally dividing and storing data in Spark RDD, locally counting respective data pieces by each node, and summarizing by a reduce to obtain a frequent 1-item set FList after all the data pieces are calculated;
step 2.2: partitioning the Flist, and solving the problem of load balancing of the Fp-Growth algorithm by adopting a polynomial time degree algorithm; taking the support degree of the frequent item set as an index of load measurement, keeping the running time of all nodes approximately consistent, dividing Flist into k groups, and the average time of each group isIn the above formula, m is the number of frequent items, sup (i)(q)) For the q-th item i(q)Such that the processing load of each node is approximately
Grouping the items in the Flist to form a projection database of each group, and broadcasting the database to each corresponding node;
step 2.3: dividing the projection database into k data blocks with equal size for each node to obtain a basic sliding window:
BSW1,BSW2......BSWk;
performing a second scan on the grouped projection database on each node;
step 2.4: constructing a pattern tree by using an expanded FP-Growth algorithm for each node, and designing the support degree for each item:
setting relatively low support for important but low-frequency contexts;
setting a relatively high degree of support for contexts that are not important but occur frequently;
each node establishes a mode tree of the node;
step 2.5: based on the basic sliding window obtained in the step 2.3, performing context updated incremental updating mining:
when the database is newly added, taking the newly added transaction as a new basic sliding window, wherein the size of the window is equal to that of the basic sliding window of the original database;
obtaining new basic sliding window BSW based on expanded Fp-Growthm+1And the last sliding window BSWmThe frequent item set is recorded as F1, the current frequent item set is F2, the new frequent item set is compared with the frequent item set in the original database, and the pattern tree is updated according to the following conditions:
case one, when the frequent entries exist in both the original database and the newly added database at F1 n F2, adding the counts of the two entry sets, and then updating the pattern tree;
in case II, when the frequent item exists in the original database but does not exist in the newly added database F1- (F1 n F2), the support count of the item in the newly added database needs to be estimated according to the similarity, and the estimated value is added to the original frequent item set count, and then the mode tree is updated;
case three, the frequent item does not exist in the original database but exists in the newly added database F2- (F1 n F2), it is necessary to look up whether there is the item in the candidate information table, if there is the item, it directly reads its count, if there is not, it is necessary to estimate the count of the item in the original basic sliding window according to the similarity, and add the estimated value and the current frequent item set count, then update the mode tree;
the similarity estimation in the second case and the third case is calculated by using a locality sensitive hashing algorithm:
the last sliding window BSWmDefined as Fwi, with dimension n, its hamming operation is changed to a vector x' with dimension n;
then, the new basic sliding window BSW is usedm+1Defining Fwj as the corresponding high-dimensional vector, and converting the Hamming operation into a vector y with the dimension n;
step 2.6: and (3) performing service matching based on the calculation result of the step 2.5:
in the mode trees of all nodes, each path contains a context C and a service set S corresponding to the context, when a mobile phone of a user returns a context C ' where the current user is located, a service set conforming to the current context C ' needs to be constructed through the current user context C ', and the step of constructing the service set is as follows:
step 2.6.1: processing data of a context c ' transmitted by a user, filtering items which are not in the Flist in the context c ', and sequencing the rest items which are not filtered in a descending order according to the support degree of the Flist to form a new context c ';
step 2.6.2: hamming a new context c ' to be a vector X ' with the dimension of n, extracting t bits from the vector X ' to be used as a hash value of the vector X ', comparing the t bits with the hash values of the processed k groups of nodes, and sending the context c ' to the nodes with higher similarity;
step 2.6.3: and (3) solving intersection of the context c' and the path p in the mode tree of the node, and endowing each path with different weight coefficients according to the intersection result, wherein the specific weight calculation formula is as follows:
step 2.6.4: recommending services to the user according to the weight, sequencing the services according to the support degree, recommending the first services with the maximum support degree to the user, and recording the set of the services as s;
step 2.6.5: and returning the c → s event data selected by the user to the cloud end and storing the c → s event data in a cloud end database for next mining.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911064443.6A CN110807051B (en) | 2019-10-31 | 2019-10-31 | Context-aware real-time service recommendation method based on cloud platform |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911064443.6A CN110807051B (en) | 2019-10-31 | 2019-10-31 | Context-aware real-time service recommendation method based on cloud platform |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110807051A CN110807051A (en) | 2020-02-18 |
CN110807051B true CN110807051B (en) | 2021-03-26 |
Family
ID=69500986
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911064443.6A Expired - Fee Related CN110807051B (en) | 2019-10-31 | 2019-10-31 | Context-aware real-time service recommendation method based on cloud platform |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110807051B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116682543B (en) * | 2023-08-03 | 2023-11-10 | 山东大学齐鲁医院 | Sharing method and system of regional rehabilitation information |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102609468A (en) * | 2012-01-17 | 2012-07-25 | 电子科技大学 | Matching method for spatial information services based on context awareness and user preferences |
CN106570008A (en) * | 2015-10-09 | 2017-04-19 | 阿里巴巴集团控股有限公司 | Recommendation method and device |
US20170147154A1 (en) * | 2015-11-19 | 2017-05-25 | Travis William Steiner | Context-aware recommendations of relevant presentation content displayed in mixed environments |
CN109800249A (en) * | 2018-12-24 | 2019-05-24 | 北京航天智造科技发展有限公司 | A kind of knowledge services method and system based on the perception of industrial service cloud platform user behavior |
-
2019
- 2019-10-31 CN CN201911064443.6A patent/CN110807051B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102609468A (en) * | 2012-01-17 | 2012-07-25 | 电子科技大学 | Matching method for spatial information services based on context awareness and user preferences |
CN106570008A (en) * | 2015-10-09 | 2017-04-19 | 阿里巴巴集团控股有限公司 | Recommendation method and device |
US20170147154A1 (en) * | 2015-11-19 | 2017-05-25 | Travis William Steiner | Context-aware recommendations of relevant presentation content displayed in mixed environments |
CN109800249A (en) * | 2018-12-24 | 2019-05-24 | 北京航天智造科技发展有限公司 | A kind of knowledge services method and system based on the perception of industrial service cloud platform user behavior |
Non-Patent Citations (2)
Title |
---|
MapReduce编程模型下的上下文离群数据挖掘算法;杨海峰 等;《太原科技大学学报》;20170815;第38卷(第4期);第265-270页 * |
Supporting Context-Aware Cloud-Based Media Recommendations for Smartphones;Abayomi Moradeyo Otebolaku等;《2014 2nd IEEE International Conference on Mobile Cloud Computing, Services, and Engineering》;20140401;第1卷;第109-116页 * |
Also Published As
Publication number | Publication date |
---|---|
CN110807051A (en) | 2020-02-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP5560367B2 (en) | Method, system and server for managing dynamic information of friends in a network | |
CN105608117B (en) | Information recommendation method and device | |
CN111262953B (en) | Method and device for pushing information in real time | |
US20120047152A1 (en) | System and method for profile tailoring in an aggregate profiling system | |
CN106815260B (en) | Index establishing method and equipment | |
TW201342088A (en) | Digital content reordering method and digital content aggregator | |
WO2014133489A1 (en) | Data synchronization | |
KR20110069067A (en) | Method, system, and apparatus for arranging content search results | |
CN112311612A (en) | Family portrait construction method and device and storage medium | |
US20160277538A1 (en) | Method and system for matching profile records | |
CN106452858A (en) | Method and device for identifying network user and computing device | |
CN110807051B (en) | Context-aware real-time service recommendation method based on cloud platform | |
Li et al. | BP-growth: Searching strategies for efficient behavior pattern mining | |
CN112148942B (en) | Business index data classification method and device based on data clustering | |
CN113556368A (en) | User identification method, device, server and storage medium | |
CN113259470B (en) | Data synchronization method and data synchronization system | |
CN110677269B (en) | Method and device for determining communication user relationship and computer readable storage medium | |
CA3108870A1 (en) | Generating real-time aggregates at scale for inclusion in one or more modified fields in a produced subset of data | |
CN112559899A (en) | User portrait generation method | |
CN105893515B (en) | Information processing method and server | |
CN112269726A (en) | Data processing method and device | |
CN109299368B (en) | Method and system for intelligent and personalized recommendation of environmental information resources AI | |
CN111723201A (en) | Method and device for clustering text data | |
CN101552689B (en) | User interest drift detection method and system based on network structure | |
WO2011067454A1 (en) | Methods and apparatuses for relevance calculation |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210326 |