CN104679992B - The design method of Markov model based on customer service usage time - Google Patents
The design method of Markov model based on customer service usage time Download PDFInfo
- Publication number
- CN104679992B CN104679992B CN201510051107.3A CN201510051107A CN104679992B CN 104679992 B CN104679992 B CN 104679992B CN 201510051107 A CN201510051107 A CN 201510051107A CN 104679992 B CN104679992 B CN 104679992B
- Authority
- CN
- China
- Prior art keywords
- time
- service
- markov model
- user
- attribute
- 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.)
- Active
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of design method of the Markov model based on customer service usage time, this method indexes quantity algorithm (i.e. with value difference first：VDM sliding-model control) is carried out to the usage time of customer service, so as to reduce the complexity of calculating, time attribute domain is divided into section, with the label in section come instead of actual time value, tendency weight is accessed on this basis using customer service to be adjusted the state-transition matrix in Markov model, so as to effectively improving prediction algorithm accuracy.
Description
Technical Field
The invention relates to a Markov model design method based on user service use time in a wireless ubiquitous environment, and belongs to the technical field of communication.
Background
Since the 21 st century, with the development of microelectronics, chips and information communication technologies, mankind will gradually enter the ubiquitous network era, in which wireless ubiquitous networks, namely: the wireless ubiquitous environment is that people are in ubiquitous networks, information exchange between people and any person and any object at any time and any place by using any network is realized, and ubiquitous information services and applications are provided for the individuals and the society by utilizing the existing network technology and the new network technology based on the requirements of the individuals and the society. With the rapid development of internet technology, business resources also show an explosive growth situation, and how to quickly and efficiently acquire a business required by a user from massive business resources makes the user obtain higher user experience, which becomes a key field of attention. Under a wireless ubiquitous environment, users are used as centers, the generation of services directly comes from the requirements of the users, the multimode terminal can be accessed into different wireless networks, the effective utilization of network resources and terminal capacity among different wireless access networks is realized, the information transmission service quality is improved, the requirements of the users are met from different angles and different levels, the user experience quality is improved, and intelligent services are provided for the users, so that the user behavior prediction needs to be researched.
In the future wireless ubiquitous environment, 2G/3G/4G cellular mobile communication networks, IEEE 802.11 Wireless Local Area Networks (WLAN) and IEEE 802.16 wireless metropolitan area networks (WiMAX) coexist as mainstream wireless access modes, the access networks are interconnected and intercommunicated through wired backbone networks or wireless Mesh networks, ad Hoc connection is realized between user terminals by virtue of series technologies such as IEEE 802.11 or IEEE 802.15, and thus ubiquitous wireless access is provided for ubiquitous service requirements; on the other hand, with the rapid development of mobile intelligent terminals and mobile operating systems in recent years, the user experience of the mobile terminals is further enhanced. Almost all the current wired internet service types can be transplanted to the mobile internet, and equipment conditions are provided for realizing wireless services.
The wireless service personalization has become a new hotspot of the current service development, and with the increasingly obvious personalized features of users, personalized push even develops into a new marketing concept and trend of operators and service providers. Therefore, the personalized push technology is developed, and can timely and appropriately push services meeting the requirements of users from a large amount of user behavior historical information through various mining algorithms and prediction models, provide personalized service customization services and service recommendations for the users, and provide decision bases for business providers to launch new services.
At present, the research on user behaviors at home and abroad mainly focuses on mining and analyzing user behavior characteristics and rules and predicting the user behaviors. Tools used for relevant research include data mining algorithms and Markov prediction models, etc. Wherein the Markov model is a simple and efficient model. Because of the non-aftereffect of the Markov chain, the Markov model theory is mostly adopted to predict the user behavior at present. The Markov model becomes a typical model of the research in the aspect of the aspect due to the large information retention amount, thereby ensuring the accurate prediction characteristic.
However, by analyzing the existing various user service prediction methods by using the Markov model, a very important factor in the construction process of the model is that the service time of the user service is not taken into consideration, that is, only the type of the user service is taken into consideration in the existing method, and the service time of each user is considered to be the same. However, different service usage times are important to express the interest of the user, and should not be considered in the probability when the service usage time is small. The prediction methods only consider the service types, often obtain a plurality of results with the same probability, are not good for uniquely determining the services required by the users in the future, and have low prediction accuracy. The present invention can solve the above problems well.
Disclosure of Invention
The invention aims to provide a method for designing a Markov model based on user service use time, which solves the problem of the user service use time on a prediction result. On one hand, the learning process of the novel Markov model prediction method in the method has good theoretical performance and high prediction accuracy; on the other hand, the method is simple to operate, easy to implement and good in application prospect.
The technical scheme adopted by the invention for solving the technical problems is as follows: a Markov model design method based on user service use time includes carrying out discretization treatment on user service use time by using value difference metric algorithm (VDM) to reduce complexity of calculation, dividing time attribute domain into intervals, using labels of intervals to replace actual time values, and utilizing user service visit tendency weight to adjust state transition matrix in Markov model on the basis of actual time value to effectively improve accuracy of prediction algorithm.
The method comprises the following steps:
the invention provides a Markov model design method based on user service use time, which comprises the following steps:
step 1: according to a historical user service sample sequence and in combination with a common habit of using services by a user, the service using time of the user is generally divided into 3 modes of short stay, common use and long-time use, and the discretization target is determined to obtain three discrete intervals;
and 2, step: the method comprises the steps that an application value difference metric algorithm (namely, VDM) is used for carrying out discretization processing on service use time of a user;
and 3, step 3: constructing a Markov model according to the user service sample sequence so as to obtain a corresponding state transition matrix;
and 4, step 4: on the basis of the step 3, the state transition matrix is adjusted according to the user service access tendency weight, so that a Markov model based on the service time of the user service is constructed;
and 5: and predicting the future service of the user according to the Markov model.
The value difference metric algorithm (VDM) is applied to the discretization of the continuous attribute, has good technical effect, applies the discretization technology to the representation of the service use time of the user, divides the time attribute domain into intervals, and replaces the actual time value with the label of the interval, thereby reducing the complexity of calculation.
Has the advantages that:
1. the method adjusts the state transition matrix in the Markov model by using the user service access tendency weight, thereby effectively improving the accuracy of the prediction algorithm.
2. The method takes the service use time of the user into consideration, and adds the service use time when constructing the Markov model, thereby achieving the aim of improving the prediction accuracy of the user behavior.
3. The method for designing the Markov model based on the service life of the user service is simple to operate, easy to implement and good in application prospect.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The invention is further described in detail with reference to the accompanying drawings.
Example one
As shown in FIG. 1, the invention provides a Markov model design method based on user service use time, which integrates a value difference metric algorithm (namely VDM) and a Markov model, and effectively improves the accuracy of user behavior prediction by adding the time factor into the construction process of the Markov model through the discretization treatment of the user service use time.
The invention mainly comprises two aspects: on one hand, a value difference metric algorithm (namely VDM) is used for carrying out discretization processing on the service time of the user service, the discretization technology is applied to the representation of the service time of the user service, a time attribute domain is divided into intervals, and the actual time value is replaced by the labels of the intervals, so that the complexity of calculation is reduced; and on the other hand, the construction of the Markov model is realized, and the state transition matrix in the Markov model is adjusted by using the user service access tendency weight to construct a new Markov model.
1. Discretization of user service use time
Considering the complexity of calculation, the invention proposes to apply a discretization method to the representation of the service use time of the user, divide the time attribute domain into intervals, and replace the actual time value with the label of the interval. The value difference scalar algorithm (i.e., VDM) can also be applied to discretization of continuous attributes with considerable technical effect. In general, the distance between discrete attributes is defined as:
let the current attribute have t discrete values c 1 ,c 2 ,…,c t S classes d 1 ,d 2 ,…,d s Then a list table as shown below can be constructed between the attributes and the category labels.
Wherein n is ij Representing a collection attribute value of c j With a class of d i Number of examples of (3). n is i* Representing a category in the collection as d i Number of examples, n *j The representation attribute takes the value of c j Number of examples of (2), n ** Is the total number of all examples.
When the discrete attribute has the category mark, the distance between different values of the attribute can be measured more accurately by observing the relationship between the attribute value and the category. The value difference metric method is as follows:
in the formula c p And c q Representing two values of the attribute. Intuitively, if c p And c q If the distributions of the two attributes are uniform, the difference between the two attributes can be considered to be 0. The more inconsistent the distribution, the greater the difference. It is clear that the value range of the above metric is 0. Ltoreq. Delta. Ltoreq.1. When the distributions are uniform, the difference is 0; the difference is 1, which is the largest when the attribute and the category are mutually sufficient requirements.
Therefore, the distance between the discrete values calculated by the value difference metric algorithm can be used as a criterion when merging or dividing the continuous attributes, and the method includes the following steps, taking merging as an example:
inputting: a set of consecutive attributes being different from each other, arranged in ascending order x 1 ,x 2 ,…,x n And their corresponding category labels; setting the maximum attribute discrete value number V max The smallest metric value M can be merged min . Parameter V max Is the result expected from the discretization, M min For adjusting the discrete effect.
And (3) outputting: several adjacent intervals, each interval representing a discrete value.
Step 1: by x 1 ,x 2 ,…,x n Forming n intervals, each interval representing a discrete value, wherein min represents a continuumThe lower limit of sex, max, represents the maximum possible value of the attribute.
c 1 ＝[min,(x 1 +x 2 )/2]
c 2 ＝[(x 1 +x 2 )/2,(x 2 +x 3 )/2]
…
c n ＝[(x n-1 +x n )/2,max]
Step 2: calculating the best adjacent interval with the minimum metric
And 3, step 3: merging best neighbor cells
If(m min Less than M min Or k is greater than V max ) /if the interval distances can be pooled or the number of discrete intervals is greater than the expected pooled result, the pooled interval/Then
The invention uses a value difference scalar algorithm (namely VDM) to carry out discretization processing on the service time of the user service, the service time (in seconds) of each service record is regarded as c in the VDM algorithm according to the value corresponding to each parameter in the VDM algorithm j And the visitor IP (representing a different user) of each record is treated as d in the VDM algorithm i . According to historical research results, in combination with common habits of users in using services, the service using time of the users is generally divided into 3 types of short stay, common use and long-time use, so that the discretization aims to obtain three discrete intervals, and then V in the VDM algorithm max Was defined as 3. The value difference scalar algorithm is used to realize the service pairingUsing discretization, and obtaining the boundary values of the three intervals as T 1 Second, T 2 And second. The browsing time is T 1 Short dwell, T, below second 2 The long-term use was performed in seconds or more, and the ordinary use was performed in the middle of the two, and the discretization results are shown in the following table.
| Discrete value | Service usage time (s represents seconds) |
| 1 | 0s<time<＝T 1 s |
| 2 | T 1 s<time<＝T 2 s |
| 3 | time>T 2 s |
Construction of Markov model
The Markov model consists of three parts, including:
(1) A set of states Status. Corresponding to the set of all traffic, it is written Status = { S, S = { (S, S) } 1 ,s 2 ,...,s n F }. S, F are two virtual states, corresponding to the start and end states.
(2) A state transition matrix a. The corresponding element is p ij Is defined as P(s) j |s i ) I.e. subscriber from current service s i Transfer to service s j The probability of (c).
(3) A set of initial probabilities. Corresponding to the initial probability for each state.
The state of the user at time t is represented by the vector H (t) = (0, 0.. 1), if the user is in state s 1 Then the ith dimension of the vector is equal to 1 and the remaining dimensions are 0. The state probability vector of the system at time t is represented by a vector V (t), each dimension representing the probability of a different state. A prediction can be made of the user's state at time t according to:
V(t)＝H(t-1)×A
in the vector V (t), the state corresponding to the dimension with the largest probability value is the most likely state of the user at the time t.
And on the basis of the discretization of the user service use time, adjusting the state transition matrix in the Markov model by using the user service access tendency weight. Wherein the user service access tendency weight w i The calculation formula of (2) is as follows:wherein t is 1 ,t 2 ,…,t n And taking discrete values of the service use time of the user, wherein n is the total number of types of the service. Adjustment formula of Markov model transfer matrixIs the number of states.
The Markov model construction process based on the service use time of the user comprises the following steps:
(1) And generating a state diagram G according to the user service type sequence set: if s is i And s j Is a continuous access service in a sequence, an s is added to the state diagram G i To s j Edge e of i,j If e is i,j If so, adding 1 to the count value of the edge;
(2) According to the aboveGenerating a state transition matrix A by the state diagram G constructed in the step 1: p is a radical of formula ij Has a value of e i,j Is compared with s i The sum of the count values of all outgoing edges;
(3) And (3) adjusting A according to the user service access tendency weight: according to the formulaRecalculating p in conjunction with user service access propensity weights ij To generate a new state transition matrix.
In summary, the present invention provides a method for designing a Markov model based on user service lifetime, which comprises the following steps:
step 1: according to a historical user service sample sequence and by combining the common habit of using services by a user, the service using time of the user is divided into 3 types, namely short stay, common use and long-time use, and the discretization target is determined to obtain three discrete intervals;
and 2, step: the operation value difference scalar algorithm (namely VDM) is used for carrying out discretization processing on the service time of the user service;
and step 3: constructing a Markov model according to the user service sample sequence so as to obtain a corresponding state transition matrix;
and 4, step 4: on the basis of the step 3, the state transition matrix is adjusted according to the access tendency weight of the user service, so that a Markov model based on the service time of the user service is constructed
And 5: and predicting the future service of the user according to the Markov model.
Example two
The specific implementation method of the invention comprises the following steps:
a. in consideration of the complexity of calculation, the invention proposes to apply a discretization technique to the representation of the service time of the user, divide the time attribute domain into intervals, and replace the actual time values with the labels of the intervals.
The distance between the discrete values calculated by the value difference metric algorithm can be used as a criterion when merging or dividing the continuous attributes, taking merging as an example, the algorithm is as follows:
inputting: a set of consecutive attributes being different from each other, arranged in ascending order x 1 ,x 2 ,…,x n And their corresponding category labels; setting the maximum attribute discrete value number V max The smallest metric value M can be merged min . Parameter V max Is the result expected from the discretization, M min For adjusting the discrete effect.
And (3) outputting: several adjacent intervals, each interval representing a discrete value.
Step 1: by x 1 ,x 2 ,…,x n N intervals are formed, each interval representing a discrete value, where min represents the lower limit of the continuous attribute and max represents the possible maximum of the attribute.
c 1 ＝[min,(x 1 +x 2 )/2]
c 2 ＝[(x 1 +x 2 )/2,(x 2 +x 3 )/2]
…
c n ＝[(x n-1 +x n )/2,max]
Step 2: calculating the best adjacent interval with the minimum metric
And step 3: merging best neighbor cells
If(m min Less than M min Or k is greater than V max ) Then/' if the interval distances can be combined or the number of discrete intervals is greater than the expected result of the combination, then combine the interval
b. Discretizing the service time of the user service by using a value difference metric algorithm (VDM), wherein the value corresponding to each parameter in the VDM algorithm is that the service time (in seconds) of each service record is regarded as c in the VDM algorithm j And the visitor IP (representing a different user) of each record is treated as d in the VDM algorithm i . According to historical research results, in combination with common habits of users in using services, the service using time of the users is generally divided into 3 types of short stay, common use and long-time use, so that the discretization aims to obtain three discrete intervals, and then V in the VDM algorithm max Was defined as 3. The discretization of service use can be realized by using a value difference metric algorithm, and the boundary values of three intervals are T respectively 1 Second, T 2 And second. Browsing time is T 1 Short dwell, T, below seconds 2 The long-term use was performed in seconds or more, and the ordinary use was performed in the middle of the two, and the discretization results are shown in the following table.
| Discrete value | Service usage time (s represents seconds) |
| 1 | 0s<time<＝T 1 s |
| 2 | T 1 s<time<＝T 2 s |
| 3 | time>T 2 s |
c. And on the basis of the discretization of the service use time of the user, adjusting the state transition matrix in the Markov model by using the access tendency weight of the user service. Wherein the user service access tendency weight w i The calculation formula of (2) is as follows:wherein t is 1 ,t 2 ,…,t n And taking discrete values of the service use time of the user, wherein n is the total number of types of the service. Adjustment formula of Markov model transfer matrixIs the number of states.
In view of the above detailed description of the method for designing a Markov model based on user service lifetime provided by the present invention, those skilled in the art may change the concept of the embodiment of the present invention in the specific implementation manner and the application scope.
Claims (1)
1. A Markov model design method based on user service use time is characterized in that the method uses a value difference scalar algorithm to carry out discretization processing on the user service use time, applies a discretization technology to the representation of the user service use time, divides a time attribute domain into intervals, and replaces actual time values with labels of the intervals, and comprises the following steps:
step 1: according to a historical user service sample sequence and in combination with the common habit of using services by a user, the service using time of the user is divided into short stay, common use and long-time use, and the discretization target is determined to obtain three discrete intervals;
and 2, step: carrying out discretization processing on the service use time of the user service by using an application value difference scalar algorithm;
and step 3: constructing a Markov model according to the user service sample sequence so as to obtain a corresponding state transition matrix;
and 4, step 4: on the basis of the step 3, the state transition matrix is adjusted according to the user service access tendency weight, so that a Markov model based on the service time of the user service is constructed;
and 5: predicting the future service of the user according to the Markov model;
the discretization of the user service use time comprises the following steps: the time attribute domain is divided into intervals, the actual time value is replaced by the label of the interval, a value difference scalar algorithm (i.e. VDM) is applied to the discretization of the continuous attribute, and the distance between the discrete attributes is defined as:
let the current attribute have t discrete values c 1 ,c 2 ,…,c t S classes d 1 ,d 2 ,…,d s Then constructing a list table as shown in the following between the attribute and the category label;
wherein n is ij Represents a collection attribute value of c j With a class of d i Number of examples, n i* Representing a category in the collection as d i Number of examples of (2), n *j The representation attribute takes the value of c j Number of examples of (2), n ** Is the total number of all examples;
when the discrete attribute has a category mark, measuring the distance between different values of the attribute by investigating the relationship between the attribute value and the category, wherein the value difference metric method comprises the following steps:
in the formula c p And c q Two values representing the attribute, if c p And c q If the distribution is consistent, the difference between the two attributes is 0, if the distribution is inconsistent, the difference is larger, the value range of the measurement is that delta is more than or equal to 0 and less than or equal to 1, and when the distribution is consistent, the difference is 0; when the attribute and the category are mutually sufficient and necessary conditions, the difference is 1, and the maximum is obtained at the moment;
the Markov model of the method comprises:
(1) A set of Status states; corresponding to the set of all traffic, it is written Status = { S, S = { (S, S) } 1 ,s 2 ,...,s n F, S, F are two virtual states, corresponding to the start and end states;
(2) A state transition matrix A with corresponding elements p ij Is defined as P(s) j |s i ) I.e. subscriber from current service s i Transfer to service s j The probability of (d);
(3) A set of initial probabilities corresponding to the initial probabilities of each state;
constructing a Markov model, and adjusting a state transition matrix in the Markov model by using the user service access tendency weight to construct a new Markov model;
the method is based on the user service access tendency weight w i The calculation formula of (2): wherein t is 1 ,t2,…,t n Using discrete value of time for user service, n is total number of service types, and then using adjustment formula of Markov model transfer matrixn isConstructing a new Markov model by the state number;
the value difference scalar algorithm of the method is applied to discretization of continuous attributes; applying a discretization technique to the representation of user traffic usage time; the time attribute field is divided into intervals, and the actual time value is replaced by the label of the interval.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510051107.3A CN104679992B (en) | 2015-01-30 | 2015-01-30 | The design method of Markov model based on customer service usage time |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510051107.3A CN104679992B (en) | 2015-01-30 | 2015-01-30 | The design method of Markov model based on customer service usage time |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104679992A CN104679992A (en) | 2015-06-03 |
| CN104679992B true CN104679992B (en) | 2018-06-05 |
Family
ID=53315025
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510051107.3A Active CN104679992B (en) | 2015-01-30 | 2015-01-30 | The design method of Markov model based on customer service usage time |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104679992B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106685721B (en) * | 2016-12-30 | 2019-12-13 | 深圳新基点智能股份有限公司 | method and system for calculating predictability of user online activity outbreak time |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009252229A (en) * | 2008-04-08 | 2009-10-29 | Junichi Yamazaki | Subjective value preservation calculation of markov interest rate prediction |
| CN103400040A (en) * | 2013-07-31 | 2013-11-20 | 中国人民解放军国防科学技术大学 | Fault diagnosis and prediction method utilizing multistep time domain difference value learning |
| CN103544850A (en) * | 2013-09-13 | 2014-01-29 | 中国科学技术大学苏州研究院 | Collision prediction method based on vehicle distance probability distribution for internet of vehicles |
| CN103905439A (en) * | 2014-03-25 | 2014-07-02 | 重庆邮电大学 | Webpage browsing accelerating method based on home gateway |
| CN103996084A (en) * | 2014-06-06 | 2014-08-20 | 山东大学 | Wind power probabilistic forecasting method based on longitudinal moment Markov chain model |
| CN104134159A (en) * | 2014-08-04 | 2014-11-05 | 中国科学院软件研究所 | Method for predicting maximum information spreading range on basis of random model |
-
2015
- 2015-01-30 CN CN201510051107.3A patent/CN104679992B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009252229A (en) * | 2008-04-08 | 2009-10-29 | Junichi Yamazaki | Subjective value preservation calculation of markov interest rate prediction |
| CN103400040A (en) * | 2013-07-31 | 2013-11-20 | 中国人民解放军国防科学技术大学 | Fault diagnosis and prediction method utilizing multistep time domain difference value learning |
| CN103544850A (en) * | 2013-09-13 | 2014-01-29 | 中国科学技术大学苏州研究院 | Collision prediction method based on vehicle distance probability distribution for internet of vehicles |
| CN103905439A (en) * | 2014-03-25 | 2014-07-02 | 重庆邮电大学 | Webpage browsing accelerating method based on home gateway |
| CN103996084A (en) * | 2014-06-06 | 2014-08-20 | 山东大学 | Wind power probabilistic forecasting method based on longitudinal moment Markov chain model |
| CN104134159A (en) * | 2014-08-04 | 2014-11-05 | 中国科学院软件研究所 | Method for predicting maximum information spreading range on basis of random model |
Non-Patent Citations (2)
| Title |
|---|
| 基于内容和结构的Markov模型在网页预取中的应用;徐燕;《计算机工程与科学》;20071231;第29卷(第4期);第25-27页 * |
| 基于改进的Markov模型预测准确度研究;张友志等;《微电脑与应用》;20081231;第24卷(第9期);第38-41页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104679992A (en) | 2015-06-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Si et al. | Mobility prediction in cellular network using hidden markov model | |
| Le et al. | A practical model for traffic forecasting based on big data, machine-learning, and network KPIs | |
| CN103987056A (en) | Wireless network telephone traffic prediction method based on big-data statistical model | |
| WO2016033969A1 (en) | Method and system for predicting traffic data amount and/or resource data amount | |
| Mohajer et al. | Big data based self-optimization networking: A novel approach beyond cognition | |
| CN110322067A (en) | Location of mobile users prediction technique based on factor graph model | |
| Wang et al. | Small cell transmit power assignment based on correlated bandit learning | |
| CN114118748B (en) | Service quality prediction method and device, electronic equipment and storage medium | |
| Paul et al. | Learning probabilistic models of cellular network traffic with applications to resource management | |
| Hu et al. | A study of LTE network performance based on data analytics and statistical modeling | |
| Shu et al. | Group behavior-based collaborative caching for mobile edge computing | |
| Nabi et al. | Deep learning based fusion model for multivariate LTE traffic forecasting and optimized radio parameter estimation | |
| Liu et al. | A SDN-based intelligent prediction approach to power traffic identification and monitoring for smart network access | |
| Wu et al. | A novel call admission control policy using mobility prediction and throttle mechanism for supporting QoS in wireless cellular networks | |
| Jain et al. | Data-prediction model based on stepwise data regression method in wireless sensor network | |
| CN104679992B (en) | The design method of Markov model based on customer service usage time | |
| CN104616077A (en) | Markov chain and association rule based user service behavior prediction method | |
| Rizwan et al. | Mobile internet activity estimation and analysis at high granularity: SVR model approach | |
| Ma et al. | Cellular Network Traffic Prediction Based on Correlation ConvLSTM and Self-Attention Network | |
| Yan et al. | Mining individual mobile user behavior on location and interests | |
| Bastidas-Puga et al. | Evaluation of SINR prediction in cellular networks | |
| Wang et al. | Communication network time series prediction algorithm based on big data method | |
| Tiwana et al. | Statistical learning for automated RRM: Application to eUTRAN mobility | |
| CN112559854A (en) | Classification method and device | |
| Tang et al. | Network availability evaluation based on markov chain of qos-aware |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |