CN102629341A - Web service QoS (Quality of Service) on-line prediction method based on geographic position information of user - Google Patents

Web service QoS (Quality of Service) on-line prediction method based on geographic position information of user Download PDF

Info

Publication number
CN102629341A
CN102629341A CN 201210110910 CN201210110910A CN102629341A CN 102629341 A CN102629341 A CN 102629341A CN 201210110910 CN201210110910 CN 201210110910 CN 201210110910 A CN201210110910 A CN 201210110910A CN 102629341 A CN102629341 A CN 102629341A
Authority
CN
Grant status
Application
Patent type
Prior art keywords
qos
user
matrix
target
information
Prior art date
Application number
CN 201210110910
Other languages
Chinese (zh)
Other versions
CN102629341B (en )
Inventor
吴健
吴朝晖
尹建伟
李莹
罗威
邓水光
Original Assignee
浙江大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date

Links

Abstract

The invention discloses a Web service QoS (Quality of Service) on-line prediction method based on geographic position information of a user. The Web service QoS on-line prediction method comprises the following steps of: (11) collecting QoS historical data and IP (Internet Protocol) information provided by the user; (12) generating a geometric geographical position coordinate of the user by the collected IP information, calculating relative distance of the geographic position of the user according to the geometric geographical position coordinate, and generating a user relative distance information matrix; (13) receiving a QoS query request of a target user and requesting the target user to customize a neighbor threshold value theta; and (14) judging the QoS query request received by the step (13), retransmitting the QoS information fed back previously to the target user if the target user has called the QoS query request, and performing QoS prediction if the QoS query request has not been called. According to the Web service QoS on-line prediction method, the accuracy for prediction is effectively improved by using a matrix decomposition algorithm combining geographic characteristic; and in addition, individual QoS query requests of a plurality of users can be responded in real time by using an optimized matrix decomposition algorithm.

Description

—种基于用户地理位置信息的Web服务QoS在线预测方法 - kind of Web-based user location information online forecasting service QoS

技术领域 FIELD

[0001] 本发明属于web服务领域,尤其涉及ー种基于用户地理位置信息的Web服务QoS在线预测方法。 [0001] The present invention pertains to a web service, and more particularly relates to a kind of QoS ー line forecasting method based on user location information in Web service.

背景技术 Background technique

[0002] 随着Web 2.0时代科技革命的不断发展,互联网环境下软件方法的主要形态、运行方式、生产方式和使用方式正发生着巨大的变化。 [0002] With the development of Web 2.0 era of technological revolution, the main form of software methods in the Internet environment, operation mode, the mode of production and use is undergoing great changes. 基于Web服务动态聚合,自动组合和弹性伸縮的分布式软件方法成为了未来网络应用开发的重要趋势。 Web services-based dynamic aggregation, distributed software and methods automatically combined elastically stretchable become an important trend in the future development of network applications. 这些Web服务技术应用都是在QoS研究基础上展开的。 These Web services technology is based on research launched in QoS. 近年来,Web服务的QoS研究成为了エ业界和学术界关注的重点。 In recent years, QoS in Web services became Ester industry and academia focus of attention. [0003]目前关于Web服务QoS研究都假设所有Web服务针对目标客户端用户的所有QoS都是已知的,然后通过数学工具来解决此问题。 [0003] Currently on the Web QoS research services assume that all Web services QoS target for all client users are known, and then to resolve this issue by mathematical tools. 然而,在真实情况下,上述假设是不实际的,原因如下:(I)当代企业组织的Web服务架构复杂。 However, in the real case, the assumption is not practical for the following reasons: (I) the contemporary business organization's Web services architecture complex. 对于最終用户来说,需要花费昂贵的时间成本才能调用所有Web服务获取QoS。 For end users, the need to spend costly time cost to call all Web service to get QoS. (2)当代互联网拓扑结构复杂,致使用户在更多时候无法获取准确的Web服务QoS。 (2) Internet topology contemporary complex, resulting in more often users will not have accurate Web service QoS. 因此,在真实的应用场景中,存在着大量Web服务针对目标用户的QoS是未知的。 Thus, in a real application scenarios, there are a lot of Web services for the target user's QoS is unknown. 这些未知QoS的存在动摇了以前服务计算领域研究的基础。 These QoS unknown presence shook the foundations of previous research service computing. 因此,针对未知的QoS进行预测是Web服务研究的重要前提。 Therefore, the forecast for unknown Web service QoS is an important prerequisite for the study.

[0004] 现有技术中,对未知的QoS进行预测主要使用基于Pearson CorrelationCoefficient(PCC)方法来计算客户端用户之间或者Web服务之间的相似度。 [0004] In the prior art, the unknown QoS mainly used to predict the degree of similarity between users or between a Web client service is calculated based on Pearson CorrelationCoefficient (PCC) method. 然而,这种计算方法存在以下几点不足: However, there are several limitations to this method of calculation the following:

[0005] I. PCC方法需要对历史记录中的QoS作统计学习,严重依赖于数据的准确性和完备性。 [0005] I. PCC method requires history of QoS for statistical learning, it relies heavily on the accuracy and completeness of the data. 然而由于当代互联网环境的复杂性,QoS记录并不一定都是准确的,致使PCC方法在服务计算场景下相似性计算准确率下降。 However, due to the complexity of the modern Internet environment, QoS records are not necessarily accurate, resulting PCC method of calculating the similarity calculation accuracy rate of decline in service at the scene.

[0006] 2.传统的PCC方法广泛地应用在推荐方法领域。 [0006] 2. PCC traditional method widely used in the field of the recommended method. 然而,推荐方法和服务计算的应用场景存在着本质的区別。 However, the recommended method of application scenarios and computing services there are essential differences. 在QoS历史记录中,每ー项都是由用户的实际使用网络环境决定的。 In the QoS history, every ー items are determined by the user's actual network environment. 这种数据客观性的特点直接降低了PCC相似度计算的精度。 Objectivity characteristics such data directly reduces the precision of similarity calculation PCC.

[0007] 3.传统的PCC算法需要反复遍历QoS记录才能产生用户关系矩阵。 [0007] 3. PCC traditional algorithm needs to repeatedly traverse the QoS recording in order to produce customer relationship matrix. 然而在实际应用场景下,PCC算法无法对海量的用户历史记录作在线预测,因此只能通过离线方式进行预测,无法响应用户实时QoS查询请求。 However, in practical application scenarios, PCC algorithm can not make online predictions of massive user history, it can only be predicted by off-line, real-time QoS can not respond to user queries.

发明内容 SUMMARY

[0008] 针对上述技术缺陷,本发明提出ー种基于用户地理位置信息的Web服务QoS在线预测方法。 [0008] In view of the above technical defects, the present invention provides QoS ー species line forecasting method based on user location information in Web service.

[0009] 为了解决上述技术问题,本发明的技术方案如下: [0009] To solve the above technical problem, the technical solution of the present invention is as follows:

[0010] ー种基于用户地理位置信息的Web服务QoS在线预测方法,包括如下步骤: [0010] Web-based user location ー species QoS information service line prediction method, comprising the steps of:

[0011] 11)收集用户提供的QoS历史数据和IP信息;[0012] 12)根据步骤11)收集的IP信息产生用户的几何地理位置坐标,根据所述几何地理位置坐标计算用户地理位置的相对距离,产生用户相对距离信息矩阵; QoS history data [0011] 11) to collect information and IP users; [0012] 12) according to step 11) generates a user IP information collected geometrical location coordinates calculated based on the user location coordinates relative geometrical location distance, relative distance information generating user matrix;

[0013] 13)接受目标用户QoS查询请求,并请求目标用户自定义邻居阈值0 ; [0013] 13) the user accepts the target QoS query request, and requests the user-defined target neighbor threshold 0;

[0014] 14)对步骤13)接受的QoS查询请求进行判断,如目标用户曾经调用过该QoS查询请求,则把上次反馈的QoS信息重新发送给目标用户;如果该QoS查询请求是未曾调用过,则进行QoS预测; [0014] 14) in step 13 is) QoS request acceptance query is determined as the target user has previously called the query request QoS, QoS feedback information put last retransmitted to the target user; if the request is a QoS query have not been called too, then the QoS prediction;

[0015] 所述QoS预测包括如下步骤: [0015] The QoS prediction comprising the steps of:

[0016] 141)根据步骤13)接收到目标用户自定义邻居阈值e,为目标用户选择合适的邻居;所述目标用户合适的邻居选择策略如下: [0016] 141) in accordance with step 13) is received from the user defined target neighbor threshold e, select an appropriate neighbor the target user; the target user right neighbor selection strategy is as follows:

[0017] G(i) = {j Idist (i, j)彡0,i 关j} (a) [0017] G (i) = {j Idist (i, j) San 0, i Off j} (a)

[0018] 其中dist(i,j)为用户地理位置的相对距离,满足上述关系的用户j都可以定义为目标用户i的合适邻居; User j [0018] where dist (i, j) is the relative distance of the user's geographic location, satisfying the above relation can be defined as the target neighbor suitable user i;

[0019] 142)基于用户地理位置信息建立约束条件: [0019] 142) to establish the geographic location information based on user constraints:

Figure CN102629341AD00051

[0021] 根据公式(b)的约束条件,建立满足Web服务预测场景的最优化方程: [0021] According to the constraint equation (b), the establishment of Web services to meet the optimization equation predicted scenario:

[0022] [0022]

2 2

Figure CN102629341AD00052

[0023] 所述R为用户-服务的QoS矩阵;Ri j是用户i对服务j的QoS使用情況。 [0023] R is the user - Matrix QoS services; Ri of i j is the use of the user service QoS of j.

[0024] Iij是原始矩阵的指示符即当Rij存在QoS记录时Iij = I,当Rij不存在QoS记录时= 0 ;U和S分别为用户的隐式特征矩阵和服务的隐式特征矩阵,所述Ui是目标用户i的隐式特征向量,X1和X2是规则因子,a是控制公式(b)地理规则项參与程度的因子; [0024] Iij is an indicator of the original matrix, i.e., when there is a QoS recording Rij Iij = I, when the recording QoS absence Rij = 0; U and S are implicit characteristics implicit user features and services matrix and matrix, the feature vector Ui implicit target user i, X1 and X2 are rules factor, a is the control equation (b) the extent of geographic factors involved in rules;

[0025] 143)对公式(C)采用梯度下降法进行求解; [0025] 143) of the formula (C) using the gradient descent method to solve;

[0026] 144)求解后,对于满足終止条件的特征矩阵U和S,通过矩阵内积方式重构原始矩阵R的所有QoS信息,即R ^ UtS,目标用户i对Web服务j的QoS预测值为Rij ; [0026] 144) After solving for the terminating condition is satisfied and characterized in matrix U S, reconstruct all the QoS information of the original matrix R by way of the matrix product, i.e. R ^ UtS, user I target Web service QoS predicted value of j as Rij;

[0027] 所述终止条件为J-J'彡£ ; [0027] The termination condition is J-J 'San £;

[0028] 所述e为迭代阈值;所述J'为迭代后得出的新J值; [0028] e is the iteration threshold; the new value of J J 'is obtained after iteration;

[0029] 15)将预测值反馈给目标用户。 [0029] 15) the predicted value back to the target user.

[0030] 进ー步的,所述步骤143)的梯度下降法进行求解包括如下步骤: For [0030] the intake ー step, step 143) to solve the gradient descent method comprising the steps of:

[0031] 21)对公式(C)分别矩阵变量U和S对偏导数方程求解,得到: [0031] 21) of the formula (C) are variable matrices U and S partial derivative of the equation is solved to give:

Figure CN102629341AD00053

[0034] 然后,梯度下降法进入了矩阵变量的迭代过程: [0034] Then, the iterative gradient descent process into the variable matrix:

Figure CN102629341AD00061

[0037] 其中n W迭代因子,用来控制梯度下降速率; [0037] wherein n W iteration factor for controlling the gradient descent rate;

[0038] 22)对于每一次迭代产生的矩阵变量U和S,将其代入公式(C)中得到J',迭代终止条件为为JJ' ( e ;所述e为迭代阈值,e =0.001。 [0038] 22) For a matrix variable U and S each iteration produced, which is substituted into the formula (C) obtained in J ', iteration termination condition is JJ' (e; a e is the iteration threshold, e = 0.001.

[0039] 进ー步的,所述反馈的QoS预测信息Rij包装成html页面格式,通过前端显示引擎把结果展现给目标用户。 [0039] ー step into the feedback prediction information of QoS Rij packaged html page format, the display engine through the distal end result to the target user.

[0040] 进ー步的,所述目标用户自定义邻居阈值e作出以下限制:若e < 0%接受e ;若e彡0%接受0%所述e*是阈值的上限,该值设置为1000。 [0040] into ー step, the target user-defined neighbor threshold value e to the following limitations: if e <0% to accept e; if 0% e San receiving 0 the% e * is the upper limit threshold value, the value is set to 1000.

[0041] 本发明的有益效果在于:通过使用用户的地理位置信息动态选择请求用户的相似用户群。 [0041] Advantageous effects of the present invention is characterized in: dynamically selecting similar users requesting user by using a user's location information. 同时,通过使用了结合地理特征的矩阵分解算法有效地提高预测的准确性。 Meanwhile, by using matrix decomposition algorithm using geographic characteristic effectively improve the accuracy of prediction. 另外,通过使用优化的矩阵分解算法可实时响应多用户的个性化QoS查询请求。 Further, by using matrix decomposition algorithm can be optimized in real time in response to a query request personalized multi-user QoS.

附图说明 BRIEF DESCRIPTION

[0042] 图I是基于用户地理位置信息的Web服务QoS在线预测方法流程图; [0042] Figure I is a flow-line prediction method based on Web QoS user location information services;

[0043] 图2是QoS在线预测引擎LBR的内部流程图。 [0043] FIG 2 is a flowchart showing an internal engine LBR line prediction of QoS.

具体实施方式 detailed description

[0044] 下面将结合附图和具体实施例对本发明做进ー步的说明。 [0044] The accompanying drawings and the following embodiments of the present invention is made into the specific embodiment described ー step.

[0045] 如图I所示,本发明的整体流程图包括以下部分: [0045] As shown in FIG I, the overall flowchart of the present invention comprises the following parts:

[0046] 流程I :收集用户提供的QoS历史数据和IP信息。 [0046] Process I: collect historical data provided by the user QoS and IP information. 假设一共有m个用户和n个服务,那么:(1)使用历史数据产生ー个m*n的用户-服务的QoS矩阵R :其中每ー项Rij是用户i对服务j的QoS使用情況。 Assuming a total of m users and n-service, then: (1) using the history data generating ー a m * n users - QoS matrix R and services: wherein each ー entry Rij is user i usage of the QoS service of j. (2)使用用户的IP信息可以产生用户的几何地理位置坐标。 (2) using the user's IP information may be generated geometric coordinates of the user's geographic location. 其中姆ー个用户对应元组(alt(i), Iat (i)), alt(i)表示用户i的经度位置,Iat (i)表示用户i的纬度位置。 Wherein a corresponding user Farm ー tuple (alt (i), Iat (i)), alt (i) represents the longitude of user i, Iat (i) represents the latitude of user i.

[0047] 流程2 :根据几何地理元组计算用户间的相对距离。 [0047] Scheme 2: calculate the relative distance between the user geographic tuple geometry. 距离公式如下: Distance formula is as follows:

[0048] [0048]

Figure CN102629341AD00062

[0049] 其中c是从经纬度单位转化成単位米的常数。 [0049] where c is the conversion from latitude and longitude to be constant. Unit Unit meters. 假设地球是球形,那么c近似等于111261。 Assuming that the earth is spherical, then c is approximately equal to 111,261.

[0050] 对所有用户计算相对距离,产生ー个m*m的用户相对距离信息矩阵D :其中每ー项dist(i, j)表示用户i和用户j的相对距离信息。 [0050] The relative distance calculation for all the users, generating a m * m ー user relative distance information matrix D: wherein each item ー dist (i, j) represents the relative distance information of the user i and user j.

[0051] 流程3 :门户页面接受目标用户QoS查询请求,并需要目标用户自定义邻居阈值9 O [0051] Process 3: QoS portal page to accept the target user queries and needs of the target user-defined thresholds neighbor 9 O

[0052] 流程4 :对流程3采集的查询请求作分析: [0052] Scheme 4: Scheme 3 query request collected for analysis:

[0053] (I)假若目标用户已经在以前调用过该Qos请求服务,那么历史记录就会存在相应的QoS记录,那么就可以把原来的QoS信息重新发送到流程6的前端显示引擎,生成结果返回页面。 [0053] (I) if the target user has been called before the Qos request service, then history will record the presence of the appropriate QoS, then you can put the original QoS information is sent back to the front of the display process 6 engine generates results return to the page.

[0054] (2)假若用户以前没有调用过该请求服务,那么需要进行流程5的QoS预测。 [0054] (2) If the user has not been called before the service request, then the need for QoS flow forecast 5.

[0055] 流程5 :在线预测LBR算法引擎是QoS预测的执行实体。 [0055] Scheme 5: LBR-line prediction algorithm execution entity QoS engine is predicted. 如图2所示,LBR算法引擎的子流程包括以下几部分: 2, the subprocess LBR algorithm engine comprises the following parts:

[0056] I)根据流程3接收到目标用户自定义邻居阈值0,为目标用户选择合适的邻居。 [0056] I) Scheme 3 receives the target user based on custom thresholds neighbors 0, select an appropriate neighbor the target user.

[0057]目标用户合适的邻居选择策略如下: [0057] Suitable target users neighbor selection strategy is as follows:

[0058] [0058]

Figure CN102629341AD00071

[0059] 其中dist(i,j)由公式⑴定义,满足上述关系的用户j都可以定义为目标用户i的合适邻居。 [0059] where dist (i, j) defined by the formula ⑴, user j satisfying the aforementioned relationship can be defined as a suitable target neighbor user i.

[0060] 实验表明,0的取值对最終的预测精度会产生影响。 [0060] Experiments show that the value of 0 will have an impact on the accuracy of the final prediction. 为了避免出现0过大导致预测性能下降的情况发生,本发明对于用户自定义的阈值e作出以下限制: 0 In order to avoid too large prediction performance degradation occurs, the present invention is limited to the threshold value e User-defined:

[0061] •若0 < 0%本发明接受0。 [0061] • If 0 <0% 0 of the present invention is acceptable.

[0062] •若0彡9%本发明接受9' [0062] • If San 9 0 9% Invention acceptable '

[0063] 0*是阈值的上限。 [0063] 0 * is the upper limit threshold value. 通过实验结果验证,本发明把0*设置为1000。 By experimental verification result, the present invention is set to 0 * 1000. 在这种限制下,既可以保证引擎可以为目标用户找到合适的邻居,同时也过滤那些不相关的邻居,从而提升预测精度。 In this limit, both to ensure that the engine can find the right target audience for the neighbors, but also filter out irrelevant neighbors, so as to enhance the prediction accuracy.

[0064] 2)为了实现算法能在线精确地预测目标用户的QoS值,本发明首先引入了经典的SVD预测技术作为算法模板: [0064] 2) In order to implement the algorithm can accurately predict online user QoS target value, the present invention is first introduced classical prediction techniques SVD algorithm as a template:

[0065] [0065]

Figure CN102629341AD00072

[0066] 其中Iij是原始矩阵的指示符(Iij = I当Rij存在QoS记录,反之Iij = 0)。 [0066] where Iij is the original matrix indicator (Iij = I Rij present when recording QoS, and vice versa Iij = 0). 根据SVD的定义,U和S分别为用户的隐式特征矩阵和服务的隐式特征矩阵。 The definition of the SVD, U and S matrices are implicit characteristics of the user and the implicit characteristic matrix and services. 后两项为规则项,避免U和S过拟合于原始算法模型。 After two terms as a rule, avoid over-fitting U and S to the original algorithm model. I |U| |F是矩阵U的弗罗贝尼乌斯范数(Frobeniusnorm),定义为: I | U | | F is the Frobenius norm of the matrix U (Frobeniusnorm), defined as:

Figure CN102629341AD00073

[0068] 其中Uu为矩阵U的第i行第j列的元素。 [0068] Uu wherein the i-th row of the matrix U of the j-th column element. 对I |S| If同理,在公式(3)中,入i和入2为规则因子,控制矩阵U和S的拟合速率。 For I | S | If Likewise, in the formula (3), the rule i and the factor of 2, the rate controlling matrix U and S is fitted. SVD预测技术通过最小化公式⑶产生满足条件的U和S,使用矩阵U和S的内积还原矩阵R的所有QoS信息。 SVD ⑶ generated prediction techniques by minimizing the condition formulas U and S, all the QoS information within the matrix U and the reduction product of S matrix R.

[0069] 对于流程I)得到目标用户合适的邻居,本发明提出如下假设:“目标用户与邻居共同调用Web服务,他们得到的服务体验应该是类似的。”这个假设符合直观感受:因为目标用户和邻居由于处于同一地区,他们将共同使用相同/类似的IT基础设施(网络带宽和网络拓扑结构等)。 [0069] obtained for process I) target user right neighbor, the invention proposes the following hypothesis: "target users and neighbors together to invoke the Web service, they get service experience should be similar." This assumption counterintuitive: because the target user Since the neighbors are in the same area, they will jointly use the same / similar IT infrastructure (network bandwidth and network topology, etc.). 也正因为服务体验是和IT基础设施密切相关的,因此目标用户和邻居的隐式特征应该是相似的。 It is because of service experience and is closely related to the IT infrastructure, so the target user and implicit characteristics neighbors should be similar. 实验结果亦表明这种假设是合理。 The results also show that this assumption is reasonable.

[0070] 基于上述假设,提出了基于用户地理位置信息的约束条件: [0070] Based on the above assumptions, the user is proposed constraint on geographical location information:

Figure CN102629341AD00081

[0072] 其中G(i)由公式⑵定义,Ui是目标用户i的隐式特征向量。 [0072] where G (i) defined by the formula ⑵, Ui implicit target eigenvector of user i. 在得到上述约束条件后,本发明把目标用户和邻居的地缘关系和传统的SVD预测技术融合起来,产生满足Web服务预测场景的最优化方程: After obtaining the above constraints, the present invention is the relationship between the target user and geographical neighbors and traditional technologies integrate SVD prediction, generation of the optimization equation satisfying Web service predicted scenario:

[0073] [0073]

Figure CN102629341AD00082

[0074] 其中入ェ和入2是规则因子,a是控制公式⑷地理规则项參与程度的因子。 [0074] and the S Factory 2 wherein the factor is the rule, a is the degree of control formula ⑷ geographic Rules participation factor.

[0075] 3)对于公式(6),本发明需要使用梯度下降法进行求解。 [0075] 3) for the equation (6), the present invention needs to be solved using a gradient descent method.

[0076] 梯度下降法首先对目标损失函数即公式(6)进行偏导数方程求解,公式(6)存在2个未知的矩阵变量U和S,故需要分别求偏导: [0076] First, the gradient descent method that is the target loss function equation (6) for solving partial derivative equations, equation (6) there are two unknown variables matrices U and S, respectively, so that the partial derivatives needed:

[0077] [0077]

Figure CN102629341AD00083

[0078] [0078]

[0079] 然后,梯度下降法进入了矩阵变量的迭代过程: [0079] Then, the iterative gradient descent process into the variable matrix:

[0080] [0080]

Figure CN102629341AD00084

[0081] [0081]

[0082] 其中n W迭代因子,用来控制梯度下降速率。 [0082] wherein n W iteration factor, used to control the rate of gradient descent.

[0083] 4)对于每一次迭代产生的矩阵变量U和S,将新产生的矩阵变量U和S代入公式 [0083] 4) The matrices U and S variables generated each iteration, the matrix U and generates a new variable into the equation S

(6)中,计算出J'从而更新公式(6)的結果。 (6), the computed J 'to update the result of Equation (6). 迭代的終止条件为: Iteration termination condition is:

[0084] J-J'彡e (11) [0084] J-J 'San e (11)

[0085] e为迭代阈值,通常e =0.001。 [0085] e is the iteration threshold, typically e = 0.001.

[0086] 假若损失函数符合上述終止条件,则迭代过程终止。 [0086] If the termination condition loss function meets the above, the iteration process terminates. 假若不满足,则重回子流程 If not satisfied, then return to the sub-processes

(3)梯度下降法继续迭代,直到满足条件为止。 (3) gradient descent method iterations continue until the condition is satisfied.

[0087] 5)根据经典SVD的定义,对于满足条件的特征矩阵U和S,本发明通过公式(12)矩阵内积方式重构原始矩阵R的所有QoS信息: [0087] 5) The classic definition of the SVD, for satisfying the condition wherein the matrix U and S, embodiment of the present invention within the matrix product of all QoS information of the original reconstruction matrix R by the equation (12):

[0088] [0088]

Figure CN102629341AD00085

[0089] 那么,目标用户i对Web服务j的QoS信息值为Rij。 [0089] Then, the target user i value Rij of QoS information of Web services j.

[0090] 在线预测算法引擎是响应用户QoS查询的核心。 [0090] Online prediction algorithm engine is the core user QoS query response. 在真实情况下,引擎需要面对众多用户的实时查询请求,这要求算法必须在提高预测精度的同时降低计算时间复杂度。 In a real situation, real-time query engine needs to face the request of many users, it must reduce the computational requirements of the algorithm time complexity while improving forecasting accuracy. 本发明的算法时间复杂度主要在于公式(7)和公式(8)。 Calculation complexity of the present invention is mainly characterized equation (7) and (8). 数学证明,每一次迭代时间复杂度为o(ed):其中e为原始QoS矩阵的密度,d是常数,为隐式特征空间的维度。 Mathematical proof, each iteration time complexity is o (ed): wherein e is the density of the original QoS matrix, d is a constant, implicit feature space dimensions. 可以看出每一次迭代的时间复杂度和原始矩阵的密度成线性关系。 As can be seen each iteration time complexity and density of the original matrix is ​​linear. 通常,原始矩阵非常稀疏,因此单次迭代的时间复杂度很低。 Typically, the original matrix is ​​very sparse, so a single iteration time complexity is low. 同吋,实验证明本发明的预测算法通常于15次左右迭代即可符合预设条件。 Same-inch, proved prediction algorithm iteration of the present invention generally at about 15 times to meet the preconditions. 综上所述,本发明的预测算法可实时响应多用户的在线QoS查询请求。 In summary, the algorithm of the present invention may predict response to a multi-user online query request QoS in real time.

[0091] 流程6 :负责接收QoS信息Rij并包装成html页面格式,并通过前端显示引擎把结果展现给用户。 [0091] Scheme 6: Rij responsible for receiving QoS information and packaged into html page format, and the results presented to the engine via the front display user.

[0092] 测试结果: [0092] Test results:

[0093] 为了量化的展现本发明中提出的QoS预测方法和传统预测方法之间的优劣,使用MAE (Mean Absolute Error)来评估预测的准确度。 [0093] In order to show the advantages and disadvantages between the quantized prediction method and the conventional QoS prediction method proposed in the present invention, the use of MAE (Mean Absolute Error) to assess the accuracy of prediction. 为了更好地解释NMAE,先简单定义一下MAE : To better explain NMAE, briefly define what MAE:

Figure CN102629341AD00091

[0095] 其中,ru,s表示客户端用户u调用Web服务s的缺失QoS预测值,Is表示客户端用户u调用Web服务s的真实QoS值,N表示进行预测的缺失QoS总数,MAE即为所有预测结果同各自真实值之间误差的平均值 [0095] where, ru, s indicates that the client calls the user u missing QoS predictive value of Web services s, Is that the client user u call real QoS value of Web services s, N represents the total number of missing QoS to predict, MAE is the All results of the error between the prediction value of the true average value of each

[0096] 实验使用的数据集中包含了339个客户端用户对5825个Web服务的详细调用信息,因此使用ー个339*5825大小的客户端用户-Web服务矩阵来存储。 [0096] experimental data set used contains 339 client users detailed information on the 5825 call Web services, so using a user client ー 5825 * 339 size matrix storage -Web service. 在实验中,这个矩阵被分成了两部分:训练样本和测试样本。 In the experiment, the matrix is ​​divided into two parts: the training and test samples. 为了尽量真实地模拟实际的使用环境,随机地抽取若干密度的矩阵样本作为训练样本,剰余的作为测试样本。 In order to try to realistically simulate the actual use of the environment, a number of randomly drawn sample matrix density as training samples for Surplus remaining as the test sample.

[0097]同时,设置 0 = 100、a = 0. 001、d = 10。 [0097] Meanwhile, a 0 = 100, a = 0. 001, d = 10.

[0098] [0098]

Figure CN102629341AD00092

[0099] 与现在的方法UMEAN,MEAN,UPCC和IPCC相比,本发明中的方法LBR的MAE值更小,即预测結果更为精确。 [0099] Compared with current methods UMEAN, MEAN, UPCC and IPCC, MAE value LBR method of the present invention is smaller, i.e., more accurate prediction results.

[0100] 以上所述仅是本发明的优选实施方式,应当指出,对于本技术领域的普通技术人员,在不脱离本发明构思的前提下,还可以做出若干改进和润饰,这些改进和润饰也应视为本发明的保护范围内。 [0100] The above are only preferred embodiments of the present invention, it should be noted that for those of ordinary skill in the art, without departing from the inventive concept premise, can make various improvements and modifications, improvements and modifications of these also it is considered within the scope of the present invention.

Claims (4)

  1. 1. ー种基于用户地理位置信息的Web服务QoS在线预测方法,其特征在于,包括如下步骤: 11)收集用户提供的QoS历史数据和IP信息; 12)根据步骤11)收集的IP信息产生用户的几何地理位置坐标,根据所述几何地理位置坐标计算用户地理位置的相对距离,产生用户相对距离信息矩阵; 13)接受目标用户QoS查询请求,并请求目标用户自定义邻居阈值0 ; 14)对步骤13)接受的QoS查询请求进行判断,如目标用户曾经调用过该QoS查询请求,则把上次反馈的QoS信息重新发送给目标用户;如果该QoS查询请求是未曾调用过,则进行QoS预测; 所述QoS预测包括如下步骤: 141)根据步骤13)接收到目标用户自定义邻居阈值0,为目标用户选择合适的邻居;所述目标用户合适的邻居选择策略如下: G(i) = {j dist(i, j)彡0 , i 关j} (a) 其中dist(i,j)为用户地理位置的相对距离,满足上述关系的用 1. ー species line forecasting method based QoS Web user location information services, characterized by comprising the steps of: 11) to collect QoS history data provided by the user and the IP information; 12) produced according to step 11 the user) information collected IP geometric location coordinates, relative distances is calculated according to the user's geographic location coordinate geometry, relative distance information to generate the user matrix; 13) receiving the target QoS user query request, and requests the user-defined target neighbor threshold 0; 14) step 13) receiving a query request for QoS determination, such as the target user has previously called the query request QoS, QoS feedback information put last retransmitted to the target user; if the query request is a QoS has not been called, the prediction of QoS ; the QoS prediction comprises the steps of: 141) step 13) receiving a user-defined neighbor to the target threshold value in accordance with 0, to select the appropriate neighbor the target user; the target user right neighbor selection strategy is as follows: G (i) = { j dist (i, j) San 0, i off j} (a) where dist (i, j) is the relative distance of the user's geographic location, by satisfying the above relationship j都可以定义为目标用户i的合适邻居; 142)基于用户地理位置信息建立约束条件: Suitable neighbor j can be defined as the target user i; 142) to establish the geographic location information based on user constraints:
    Figure CN102629341AC00021
    根据公式(b)的约束条件,建立满足Web服务预测场景的最优化方程: According to the constraint equation (b), the establishment of Web services to meet the optimization equation predicted scenario:
    Figure CN102629341AC00022
    所述R为用户-服务的QoS矩阵;Ri j是用户i对服务j的QoS使用情况; Iij是原始矩阵的指示符即当Rij存在QoS记录时Iij = I,当Rij不存在QoS记录时Iij = 0 ;U和S分别为用户的隐式特征矩阵和服务的隐式特征矩阵,所述Ui是目标用户i的隐式特征向量,X1和X2是规则因子,a是控制公式(b)地理规则项參与程度的因子; 143)对公式(C)采用梯度下降法进行求解; 144)求解后,对于满足終止条件的特征矩阵U和S,通过矩阵内积方式重构原始矩阵R的所有QoS信息,即R ^ UtS,目标用户i对Web服务j的QoS预测值为Rij ; 所述终止条件为JJ' ^ £ ; 所述e为迭代阈值;所述J'为迭代后得出的新J值; 15)将预测值反馈给目标用户。 R is the user - Matrix QoS services; Ri of i j is the user's use of the QoS service j; ??? Iij is an indicator of the original matrix, i.e., when there is a QoS recording Rij Iij = I, Iij recorded when QoS absence Rij = 0; U and S are the users of the service and the implicit characteristic matrix implicit feature matrix, the eigenvectors Ui implicit target user i, the X1 and X2 is a regular factor, a is a control formula (b) geographic degree of participation factor rules; 143) of the formula (C) is solved using the gradient descent method; 144) after solving for the terminating condition is satisfied characteristic matrix U and S, all QoS reconstruct the original matrix R by way of the matrix product information, i.e., R ^ UtS, the target QoS of user i j the predicted value Rij of Web services; the termination condition is JJ '^ £; e is the iteration threshold; the J' is obtained after iteration new J value; 15) back to the predicted value of the target user.
  2. 2.根据权利要求I所述的ー种基于用户地理位置信息的Web服务QoS在线预测方法,其特征在于,所述步骤143)的梯度下降法进行求解包括如下步骤: 21)对公式(c)分别矩阵变量U和S对偏导数方程求解,得到: The I according to claim ー kinds of Web-based user location information service QoS line prediction method, wherein said step 143) to solve the gradient descent method comprising the steps of: 21) of the formula (c) each variable U and S matrix solving partial derivative equations to give:
    Figure CN102629341AC00023
    Figure CN102629341AC00031
    然后,梯度下降法进入了矩阵变量的迭代过程: Then, gradient descent into the matrix iterative process variables:
    Figure CN102629341AC00032
    其中nり迭代因子,用来控制梯度下降速率; 22)对于每一次迭代产生的矩阵变量U和S,将其代入公式(C)中得到J',迭代终止条件为为J-J'≤e ;所述e为迭代阈值,e =0.001。 Wherein n ri iteration factor, used to control the rate of descent gradient; 22) for each iteration of the matrix generated variable U and S, which is substituted into the formula (C) obtained in J ', iteration termination condition is J-J'≤e ; e is the iteration threshold, e = 0.001.
  3. 3.根据权利要求I所述的ー种基于用户地理位置信息的Web服务QoS在线预测方法,其特征在于,所述反馈的QoS预测信息Rij包装成html页面格式,通过前端显示引擎把结果展现给目标用户。 According to claim I of the Web-based user location ー species QoS information service line prediction method, wherein, the QoS feedback prediction information Rij packaging into html page format, through the distal end result to the display engine Target users.
  4. 4.根据权利要求I所述的ー种基于用户地理位置信息的Web服务QoS在线预测方法,其特征在于,所述目标用户自定义邻居阈值0作出以下限制: 若9 < 9%接受e ;若e彡0%接受0%所述0*是阈值的上限,该值设置为1000。 According to claim I of the Web-based user location ー species QoS information service line prediction method, wherein said target neighbor user-defined threshold value 0 to the following limitations: if 9 <9% received E; if receiving e San 0% 0% 0 * is the upper limit of the threshold value, the value is set to 1000.
CN 201210110910 2012-04-13 2012-04-13 Based on user location information online Web service (QoS) prediction method CN102629341B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201210110910 CN102629341B (en) 2012-04-13 2012-04-13 Based on user location information online Web service (QoS) prediction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201210110910 CN102629341B (en) 2012-04-13 2012-04-13 Based on user location information online Web service (QoS) prediction method

Publications (2)

Publication Number Publication Date
CN102629341A true true CN102629341A (en) 2012-08-08
CN102629341B CN102629341B (en) 2015-09-09

Family

ID=46587599

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201210110910 CN102629341B (en) 2012-04-13 2012-04-13 Based on user location information online Web service (QoS) prediction method

Country Status (1)

Country Link
CN (1) CN102629341B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103684850A (en) * 2013-11-25 2014-03-26 浙江大学 Service neighborhood based Web Service quality prediction method
CN103840985A (en) * 2014-02-28 2014-06-04 浙江大学 Web service quality prediction method and device based on user neighborhoods
CN104270281A (en) * 2014-09-03 2015-01-07 河海大学 Web Service QoS (Quality of Service) combined forecasting method
CN104601385A (en) * 2013-10-31 2015-05-06 浙江大学 WebService service quality prediction method based on geographic location
CN105024886A (en) * 2015-07-31 2015-11-04 浙江大学 Rapid Web server QoS (Quality of Service) prediction method based on user metadata
CN103647673B (en) * 2013-12-24 2017-03-15 博元森禾信息科技(北京)有限公司 Quality of Service Web service QoS prediction method and apparatus

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101217427A (en) * 2008-01-11 2008-07-09 清华大学 A network service evaluation and optimization method under uncertain network environments
US20080243735A1 (en) * 2007-03-26 2008-10-02 International Business Machines Active sampling collaborative prediction method for end-to-end performance prediction
CN102170449A (en) * 2011-04-28 2011-08-31 浙江大学 Web service QoS prediction method based on collaborative filtering

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080243735A1 (en) * 2007-03-26 2008-10-02 International Business Machines Active sampling collaborative prediction method for end-to-end performance prediction
CN101217427A (en) * 2008-01-11 2008-07-09 清华大学 A network service evaluation and optimization method under uncertain network environments
CN102170449A (en) * 2011-04-28 2011-08-31 浙江大学 Web service QoS prediction method based on collaborative filtering

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
M. ALRIFAI, D. SKOUTAS, AND T. RISSE: "Selecting skyline services for qos-based web service composition", 《PROCEEDINGS OF THE 19TH INTERNATIONAL CONFERENCE ON WORLD WIDE WEB》, 31 December 2010 (2010-12-31), pages 465 - 471, XP032303097, DOI: doi:10.1109/CIS.2012.140 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104601385A (en) * 2013-10-31 2015-05-06 浙江大学 WebService service quality prediction method based on geographic location
CN103684850A (en) * 2013-11-25 2014-03-26 浙江大学 Service neighborhood based Web Service quality prediction method
CN103684850B (en) * 2013-11-25 2017-02-22 浙江大学 Web Service based forecasting service quality neighborhood
CN103647673B (en) * 2013-12-24 2017-03-15 博元森禾信息科技(北京)有限公司 Quality of Service Web service QoS prediction method and apparatus
CN103840985A (en) * 2014-02-28 2014-06-04 浙江大学 Web service quality prediction method and device based on user neighborhoods
CN104270281A (en) * 2014-09-03 2015-01-07 河海大学 Web Service QoS (Quality of Service) combined forecasting method
CN104270281B (en) * 2014-09-03 2017-09-19 河海大学 One kind of Web Service QoS combination forecasting method
CN105024886A (en) * 2015-07-31 2015-11-04 浙江大学 Rapid Web server QoS (Quality of Service) prediction method based on user metadata
CN105024886B (en) * 2015-07-31 2018-04-20 浙江大学 Service with QoS prediction method based on fast Web user metadata

Also Published As

Publication number Publication date Type
CN102629341B (en) 2015-09-09 grant

Similar Documents

Publication Publication Date Title
CN103001805A (en) Web user flow producing method based on user behavior analysis
Bonissone et al. Fast meta-models for local fusion of multiple predictive models
JP2007317068A (en) Recommending device and recommending system
CN102523246A (en) Cloud computation treating system and method
US20130024141A1 (en) Energy consumption analysis using node similarity
WO2000055790A2 (en) Gradient criterion method for neural networks and application to targeted marketing
CN102854968A (en) Real-time energy consumption metering method of virtual machine
CN103106279A (en) Clustering method simultaneously based on node attribute and structural relationship similarity
CN103544663A (en) Method and system for recommending network public classes and mobile terminal
Faltings et al. Incentive mechanisms for community sensing
Hanson et al. Jump-diffusion stock return models in finance: Stochastic process density with uniform-jump amplitude
CN103106616A (en) Community detection and evolution method based on features of resources integration and information spreading
US20140143332A1 (en) Discovering signature of electronic social networks
CN102609533A (en) Kernel method-based collaborative filtering recommendation system and method
CN103606022A (en) Short-term load prediction method
CN103425754A (en) Tile-type map service caching real-time dynamic update method applicable to on-line edit of user
Gündüz et al. A poisson model for user accesses to web pages
CN102663616A (en) Method and system for measuring web advertising effectiveness based on multiple-contact attribution model
Yang et al. Scalable collaborative filtering using incremental update and local link prediction
CN103886047A (en) Distributed on-line recommending method orientated to stream data
CN102682074A (en) Product implicit attribute recognition method based on manifold learning
CN101894369A (en) Real-time method for computing focal length of camera from image sequence
Jia et al. Planned missing data designs with small sample sizes: How small is too small?
CN103440352A (en) Method and device for analyzing correlation among objects based on deep learning
CN103237023A (en) Dynamic trust model establishing system

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model
EE01