CN104579850A

CN104579850A - Quality of service (QoS) prediction method for Web service under mobile Internet environment

Info

Publication number: CN104579850A
Application number: CN201510039675.1A
Authority: CN
Inventors: 王尚广; 王陆葆; 孙其博; 李静林; 刘志晗; 杨放春
Original assignee: Beijing University of Posts and Telecommunications
Current assignee: Beijing University of Posts and Telecommunications
Priority date: 2015-01-27
Filing date: 2015-01-27
Publication date: 2015-04-29

Abstract

Provided is a quality of service (QoS) prediction method for a Web service under a mobile Internet environment. Firstly, QoS original data of a certain Web service repeatedly accessed by a certain user under the mobile Internet environment, and carrying out preprocessing on the QoS original data by using logarithm processing mode is collected; secondly, similar user groups of a target user is found through an improved and user-based collaborative filtering method, and then a normal value section is confirmed according to features of QoS attributes; thirdly, an authentic QoS prediction value can be obtained through the calculation. The quality of service (QoS) prediction method is characterized in that the preprocessing on the original data is carried out through logarithm calculation; the normal value section is selected to avoid an abnormal value generated by QoS volatility; a user personality characteristic shown by the volatility is comprehensively considered; the reliability of QoS data is guaranteed when calculating the similarity; the accurate QoS prediction value is obtained by the calculation through similar users. The quality of service (QoS) prediction method has the advantages that the calculation accurateness is high, the practicality is high, and the value for the application and popularization is good.

Description

Quality of service (QoS) prediction method for Web service in mobile Internet environment

Technical Field

The invention relates to a quality of service (QoS) of Web service under the mobile Internet environment (quality of service) prediction method; belongs to the technical field of computer application.

Background

Due to the rapid development of mobile networks, a great number of various Web services are continuously and continuously appearing, and therefore, when selecting a Web service, a user needs to face more and more candidate Web services with the same function but different performance differences. Moreover, it is not possible for the user to try out each Web service one by one to select the best Web service. This requires a comprehensive consideration of the performance of various candidate Web services on the respective quality of service QoS attributes. The QoS specific performance parameters of the Web service include: availability, response time, bandwidth, throughput, etc., each of which characterizes quality information of the Web service in some aspect. However, various QoS data generated by historical access of any user do not necessarily cover all the candidate Web services, so a method for making QoS prediction is needed in the face of missing QoS data, and a corresponding Web service is selected according to the prediction result. Therefore, accurate prediction of QoS performance parameters is a prerequisite and key for subsequent Web service recommendation, so research on the problem is always the focus of research of relevant scholars at home and abroad.

The state of the art in foreign countries is: balke and Wagner, hannover university, germany, propose a Web services QoS query scheme that integrates user usage patterns, requirements, and user preferences. Balke further proposes a collaborative discovery mechanism of Web services QoS based on user preferences.

Maamar et al, university of Mezajersey, headquartered, proposed a model for Web Service interaction, which highlights the resources of Web Service interaction, in addition to user preferences. The focus of this approach is to provide a mechanism to formalize the service consumer's preferences, history, scenarios, and resources of the service provider.

Kokash et al, the university of leyton, the netherlands, proposed a Web service discovery method based on the past experience of the user. The method formalizes the interaction between the Web service and the consumer into implicit culture, thereby recommending and providing good Web service to the newly registered Web service consumer.

The domestic situation is as follows: the Web service prediction method based on the Euclidean distance, which is proposed by Shaolingguang et al of Beijing university, converts various QoS data into distance vectors, and further calculates the similarity degree between users, thereby realizing the prediction of missing QoS data.

The WSRec method proposed by Zheng and et al of hong Kong Chinese university is to perform deep optimization on the basis of collaborative filtering, and eliminates the error of similarity calculation caused by individual factors of users by analyzing the weight and parameters of various QoS data and calculating the historical QoS data of the users, thereby realizing accurate positioning of similar users and finally realizing accurate QoS prediction.

In summary, most of the existing QoS prediction methods have the following disadvantages:

data processing and user similarity solving of various QoS prediction methods in the prior art are performed based on a network environment which can basically realize stable transmission, such as the traditional Internet, and most of QoS data for analysis are only single QoS raw data collected at a certain time selected randomly. Compared with the traditional internet, the mobile internet is only widely used in recent years, the occurrence time is short, and the transmission has the inherent characteristics and defects: for example, when information is transmitted through a wireless signal, situations such as delay, packet loss, etc. occur frequently, and these situations occur frequently, which causes uncertainty of QoS data to change, such as response time becoming large, throughput becoming small, etc. Even in the case of a good network status, this situation often occurs due to various objective factors in the wireless transmission process (e.g., the location of the user, the surrounding environment, the load of the wireless transmission device, etc.). As a result, the values of the QoS parameters of many Web services are caused to show fluctuating changes, i.e., there are abnormal values and normal values. Therefore, if these prior art prediction methods are directly applied to QoS prediction of a Web service in a mobile internet environment, a large prediction error occurs.

The QoS prediction error is mainly caused by the following two reasons:

(1) due to the influence of the volatility, the QoS data in the data set used in the prediction may not truly reflect the QoS condition of the node in a normal condition, i.e., the QoS data is unreliable. Therefore, such unreliable QoS data may cause error in similarity calculation, resulting in that the non-similar users are treated as similar users by mistake.

(2) Also because the QoS data due to the volatility is unreliable, the determination of the QoS data prediction value is affected by the volatility even if the correct similar user is selected.

Therefore, how to improve the above two problems causing the QoS prediction error of the Web service becomes a new topic focused by the technicians in the industry, and a lot of research and study are performed on the new topic.

Disclosure of Invention

In view of the above, the present invention provides a method for predicting quality of service (QoS) of a Web service in a mobile internet environment, which can accurately predict QoS values in a mobile internet environment with strong value fluctuation of QoS parameters, and the method first preprocesses collected QoS raw data in a simple and easy manner, so as to reduce the fluctuation of QoS data obtained in the mobile internet environment well, and further finds out a similar user group of a target user through calculation, determines a QoS normal value interval thereof, and avoids and deletes abnormal values in the QoS data, thereby obtaining a more accurate QoS predicted value, which is used as a key reference factor for reasonably selecting the Web service, and contributes to popularization and application of the Web service.

In order to achieve the above object, the present invention provides a method for predicting quality of service qos (quality of service) of a Web service in a mobile internet environment, comprising: firstly, QoS (quality of service) original data of a certain Web service repeatedly accessed by a certain user under the environment of a mobile internet are collected, and the QoS original data are preprocessed in a logarithmic processing mode; then, finding out a similar user group of a target user through an improved user-based collaborative filtering method, further determining a normal value interval according to the characteristics of QoS attributes, and finally calculating to obtain a credible QoS predicted value; the method comprises the following operation steps:

step 1, preprocessing collected QoS raw data from the real world: due to the fact that a plurality of uncertain interference factors exist in the data of the mobile internet in the wireless transmission process, the QoS original data when the Web service is accessed presents fluctuation changes of different degrees, and particularly the numerical value fluctuation of response time is more obvious; therefore, the multiple QoS raw data of each Web service accessed by each user is obtained firstly, and then the QoS raw data is preprocessed: using logarithm calculation to reduce QoS data difference caused by volatility, and obtaining QoS data used for subsequent calculation;

step 2, calculating the similarity between users: calculating a Pearson Correlation coefficient PCC (Pearson Correlation coefficient) between multiple times of QoS data obtained by two users on each Web service commonly accessed by the two users by adopting a user-based collaborative filtering method, and converting the Pearson Correlation coefficient calculated by the two users on all the Web services commonly accessed by the two users into a similarity value for representing the user similarity between the two users; then, the similarity between all users is obtained by adopting the mode;

step 3, determining the normal value interval of the similar user group and the QoS data thereof, and avoiding the abnormal value: determining a similar user group according to the calculation result of the step 2, and selecting a change interval of the QoS data normal value according to the QoS attribute characteristics for shielding an abnormal value, thereby determining a normal value interval of the QoS data of each similar user in the similar user group;

step 4, solving the QoS predicted value: and performing weighting processing on the determined multiple similar users and QoS data thereof, and shielding abnormal values through normal value intervals to obtain a final QoS predicted value.

The key innovative technology of the QoS prediction method of the Web service under the mobile Internet environment is as follows: the abnormal value is processed by adopting a two-step method: firstly, preprocessing QoS (quality of service) original data by using logarithm calculation, and then calculating the similarity between users by using improved PCC (policy and charging control), thereby not only reducing the influence of abnormal values on similarity calculation, but also comprehensively considering the user personality characteristics embodied by volatility. And secondly, deleting abnormal values generated by QoS fluctuation by selecting a normal value interval to perform data screening, thereby thoroughly eliminating the influence of the abnormal values on the calculation of predicted values and realizing accurate QoS prediction.

In addition, the innovative technology of the method of the invention also comprises the following three points:

(A) the acquired QoS raw data is preprocessed in a simple logarithmic calculation mode, the realization effect is good, the method is simple, and the method is suitable for the requirement of big data calculation.

(B) The PCC correlation coefficient is calculated by repeatedly accessing the QoS data of the same Web service by two users, and the correlation coefficient obtained by the Web service commonly accessed by the two users is weighted, so that the similarity of the two users is obtained. The improved PCC calculation method considers the personalized characteristics of the fluctuating QoS data when the user repeatedly accesses the Web service, and improves the accuracy of the calculation of the similarity of the user.

(C) According to the characteristics of the QoS data attributes, the normal value interval is determined, the influence of abnormal values is further avoided, the method is simple and effective, and all the abnormal values can be deleted to the greatest extent.

The method of the invention has the advantages that: the method can preprocess the fluctuating QoS original data in the mobile internet in a simple and easy way to obtain reliable QoS data to carry out similarity calculation, thereby obtaining a reliable similar user group. And then, determining a normal value interval according to the characteristics of the QoS attributes, and further avoiding and deleting the influence of QoS data fluctuation so as to obtain a reliable QoS predicted value.

The method has strong applicability and can be used for QoS value prediction of various types of Web services. Most importantly, the processing mode of the method is simple to operate, easy to implement, low in calculation complexity, very suitable for big data calculation, and capable of effectively guaranteeing efficiency of large-scale user data calculation.

Therefore, the method has higher calculation accuracy and stronger practicability and has good popularization and application values.

Drawings

Fig. 1 is a flowchart illustrating the operation steps of a QoS prediction method for a Web service in a mobile internet environment according to the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the accompanying drawings.

The invention relates to a QoS prediction method of Web service in mobile internet environment, which comprises the steps of firstly collecting QoS raw data of a certain Web service repeatedly accessed by a certain user in the mobile internet environment, and preprocessing the QoS raw data in a logarithmic processing mode; and then, finding out the most similar user group of the target user by an improved user-based collaborative filtering method, further determining a normal value interval according to the attribute characteristics of the QoS data, and finally calculating to obtain a credible QoS predicted value.

Referring to fig. 1, the specific operation steps of the method of the present invention are described:

step 1, preprocessing collected QoS raw data from the real world: because there are a plurality of uncertain interference factors in the wireless transmission process of network data in the mobile internet; especially, when the network is busy, packet loss, timeout and retransmission occur continuously, so that the QoS original data when accessing the Web service shows fluctuation in different degrees, and especially the numerical fluctuation of the response time is more obvious. For example, normal and abnormal values may exist in the collected QoS raw data set. Although the number of abnormal values is small, the abnormal values have a much larger value (e.g., response time) than the normal values. If a simple operation is performed on the basis of only the QoS raw data, the deviation of the obtained result must be very large. Therefore, after acquiring multiple times of QoS raw data of each Web service accessed by each user, preprocessing the QoS raw data: and reducing the QoS data difference caused by the volatility by utilizing logarithmic calculation to obtain QoS data used for subsequent calculation. The step 1 comprises the following operation contents:

(11) setting the QoS raw data set of t repeated accesses of the ith user to the jth Web service to be Q^Original _i,j＝{q¹ _i,j,q² _i,j,…,q^r _i,j,…,q^t _i,jThe natural numbers i, j and r are respectively the user serial number, the Web service serial number accessed by the user and the serial number of the repeated access times of the Web service serial number, and the maximum value of r is t; q. q.s^r _i,jAnd executing the QoS raw data of the access of the r time for the ith user to the jth Web service.

(12) Is provided with the firstQoS raw data set Q of t repeated accesses of i users to jth Web service^Original _i,jThe QoS data set of each QoS raw data after being preprocessed is Q_i,j＝{p¹ _i,j,p² _i,j,…,p^r _i,j,…,p^t _i,j}; wherein p is^r _i,jQoS raw data q for ith user to jth Web service access r^r _i,jAnd performing the preprocessed QoS data.

(13) QoS raw data q for t repeated accesses of ith user to jth Web service^r _i,jCalculating formula p according to natural logarithm^r _i,j＝ln(q^r _i,j) Preprocessing of a base natural logarithm operation is performed.

Step 2, calculating the similarity between users: calculating a Pearson correlation coefficient PCC between multiple times of QoS data obtained by two users on each Web service commonly accessed by the two users by adopting an improved user-based collaborative filtering method, and converting the Pearson correlation coefficient calculated by the two users on all the Web services commonly accessed by the two users into a similarity numerical value for representing the user similarity between the two users; then, the similarity between all users is obtained in the above manner.

The improved user-based collaborative filtering method is used for discovering similar user individuals of the target object based on the neighborhood among the users according to the historical behavior data of the users, so that the behavior data of the target object is predicted, and the purpose of individual recommendation is achieved.

In the step 2, because the calculation is simple and the precision is high, the similarity between users adopts a method for calculating a Pearson correlation coefficient PCC; however, due to the characteristics of QoS data and its data set, the PCC calculation method needs to be improved: the personality characteristics of the user of the QoS data when the user accesses the Web service are considered, and the personality characteristics of the volatility of the QoS data when the user repeatedly accesses the Web service are considered.

The operation of calculating the similarity between users in step 2 includes the following steps:

(21) when calculating the similarity between two users a and b, firstly, find out the set US of the Web service commonly visited by the two users_a∩US_b＝{s₁,s₂,s₃,…,s_j,…,s_nIn the formula, s_jFor the j-th Web service commonly accessed by the two users a and b, the natural numbers j and n are the sequence number of the Web service commonly accessed by the two users a and b and the total number of the Web services commonly accessed, i.e. the maximum value of j is n.

(22) When the correlation degree of the two users a and b on each Web service which is commonly accessed by the two users a and b is calculated, the Pearson correlation coefficient of the two users on each Web service which is commonly accessed by the two users is obtained; then, the obtained correlation coefficients are converted into a similarity value by adopting an arithmetic mean mode, and the similarity value is used as the unweighted user similarity of the two users:in the formula, sim_a,bFor the unweighted processed user similarity values of the two users a and b,andrespectively repeating t times for user a and user b to obtain the expected value of QoS data when accessing jth Web service,

and

the variance of the QoS data obtained when the user a and the user b access the jth Web service repeatedly t times respectively,is all t QoS data p¹ _a,j+p² _a,j+p³ _a,j+…+p^t _a,jThe accumulated sum of (c).

(23) Since the calculation process of the above step (22) only considers the PCC of the QoS data obtained when the two users a and b commonly access the Web service and calculates the average value thereof, it is not excluded that when the two users a and b commonly access the Web service, the PCC of the QoS data is obtainedWhen the number n of the accessed Web services is small, the influence on the calculation result is caused; therefore, the calculation result sim of step (22) is required_a,bBy the following formulaCarrying out weighting processing; in formula (II) sim'_a,bFor the calculation result sim_a,bThe numerical value after weighting processing is used for representing the final user similarity of the two users a and b; | US_a∩US_bI is the number of Web services that the two users have accessed together, | US_aI and I US_bAnd | is the number of the Web services respectively accessed by the user a and the user b.

(24) And (5) returning to execute the steps (21) to (23), and solving to obtain the similarity between all the users.

Step 3, determining the normal value interval of the similar user group and the QoS data thereof, and avoiding the abnormal value: and (3) determining a similar user group according to the calculation result of the step (2), and selecting a change interval of the QoS data normal value according to the QoS attribute characteristics for shielding the abnormal value, thereby determining a normal value interval of the QoS data of each similar user in the similar user group. This is because there are two types of QoS attributes for Web services: a positive attribute and a negative attribute, wherein when the network environment becomes bad and is not beneficial to information transmission, the QoS value of the positive attribute (such as the throughput of the system) becomes small, and the QoS value of the negative attribute (such as the response time and the packet loss rate) becomes large; therefore, the variation interval of the QoS data normal value is selected according to the QoS attribute characteristics for masking the abnormal value, so as to determine the normal value interval of the QoS data of each similar user in the similar user group.

The step 3 comprises the following operations:

(31) if the missing QoS data of a certain user a to a certain Web service w is to be predicted, firstly, the similarity data between the user a and all users is obtained according to the step 2, and then all users having historical access to the Web service w are arranged in a descending order according to the similarity value of the users a.

(32) And (4) according to the descending order obtained in the step (31), the first K users form a similar user group of the user a: s_sam(a)＝{u₁,u₂,u₃,…,u_c,…,u_KIn the formula, the natural number c is the serial number of the similar user in the similar user group of the user a, and the maximum value is K, u_cIs a similar group of users S_sam(a)C-th similar user in (1); the size of the number K of users depends on the data set from which the calculation is based. The number K of the similar user groups of the user a is the value K when the prediction error value is minimum after a plurality of experiments are carried out on the used QoS data set.

(33) Setting QoS data set Q of K similar users in user a similar user group_uc,jIs a set of

S_{sim (a, j)} = {Q_{u_{1}, j}, Q_{u_{2}, j}, . . ., Q_{u_{c}, j}, . . ., Q_{u_{K}, j}},

Wherein,and repeating t times for accessing the QoS data set of the jth Web service for the c-th similar user in the similar user group of the user a.

(34) Solving each similar user u in the similar user group_cQoS data set obtained when repeatedly accessing jth Web serviceExpected value ofWherein,is composed ofThe r-th QoS data element of (1); the natural number r is the number of access times for the user to repeatedly access a certain Web service, and the maximum value is t.

(35) Since the QoS data attributes include positive attributes or negative attributes, this characteristic can also be found from the collected QoS raw data: when the network condition becomes worse to affect the information transmission, the response time value tends to fluctuate more and the throughput value tends to fluctuate less; therefore, the normal value interval of the forward attribute QoS data is (p_max) The normal value interval of the negative attribute QoS data is (p)_min，) (ii) a In the formula,for the c-th similar user u in the similar user group_cQoS data set obtained when repeatedly accessing jth Web serviceExpected value of p_minAnd p_maxRespectively corresponding QoS data setsMinimum and maximum values of (d).

(36) Screening according to the QoS data normal value interval to obtain a QoS data normal value set:

and is

<math> <mrow> <msubsup> <mi>Q</mi> <mrow> <msub> <mi>u</mi> <mi>c</mi> </msub> <mo>,</mo> <mi>j</mi> </mrow> <mo>′</mo> </msubsup> <mo>&Subset;</mo> <msub> <mi>Q</mi> <mrow> <msub> <mi>u</mi> <mi>c</mi> </msub> <mo>,</mo> <mi>j</mi> </mrow> </msub> <mo>,</mo> <msubsup> <mi>p</mi> <mrow> <msub> <mi>u</mi> <mi>c</mi> </msub> <mo>,</mo> <mi>j</mi> </mrow> <mrow> <mo>′</mo> <mi>e</mi> </mrow> </msubsup> <mo>&Element;</mo> <mo>{</mo> <mrow> <mo>(</mo> <msub> <mi>p</mi> <mi>min</mi> </msub> <mo>,</mo> <msub> <mi>E</mi> <mrow> <msub> <mi>u</mi> <mi>c</mi> </msub> <mo>,</mo> <mi>j</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow> </math>

Or (a)，p_max) }; wherein,to representIs composed ofA subset of (1), i.e.Is a slave QoS data setThe medium is selected from the raw materials of the raw materials,and, if setQoS data inIs positive attribute, its value is in the interval (p)_min，) Internal; if setQoS data inA negative attribute, its value is in the interval (p_max) In the table, the natural number e is the number of QoS data, and the maximum value is v.

The step 4 comprises the following operations:

(41) set of sequential pairs

Q_{sim (a, j)} = {Q_{u_{1}, j}, Q_{u_{2}, j}, . . ., Q_{u_{c}, j}, . . ., Q_{u_{K}, j}}

Screening abnormal value data according to the step (36) for each parameter to obtain a screened QoS data set

(42) According to the formulaCalculating to obtain K QoS predicted values F (u)_cJ); wherein u is_c∈Q_sim(a)＝{u₁,u₂,…,u_c,…,u_K}，v isTotal number of QoS data in (1);

(43) for the K predicted value results F (u)_cJ) obtaining a predicted value Forecast of the missing QoS data when the user a accesses the jth Web service after weighting processing according to the following formula:

<math> <mrow> <mi>Forecast</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mi>K</mi> </mfrac> <munderover> <mi>Σ</mi> <mrow> <mi>c</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>K</mi> </munderover> <mi>F</mi> <mrow> <mo>(</mo> <msub> <mi>u</mi> <mi>c</mi> </msub> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>+</mo> <mfrac> <mrow> <munderover> <mi>Σ</mi> <mrow> <mi>c</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>K</mi> </munderover> <msubsup> <mi>sim</mi> <mrow> <mi>a</mi> <mo>,</mo> <msub> <mi>u</mi> <mi>c</mi> </msub> </mrow> <mo>′</mo> </msubsup> <mrow> <mo>(</mo> <mi>F</mi> <mrow> <mo>(</mo> <msub> <mi>u</mi> <mi>c</mi> </msub> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>-</mo> <mfrac> <mn>1</mn> <mi>K</mi> </mfrac> <munderover> <mi>Σ</mi> <mrow> <mi>c</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>K</mi> </munderover> <mi>F</mi> <mrow> <mo>(</mo> <msub> <mi>u</mi> <mi>c</mi> </msub> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>)</mo> </mrow> </mrow> <mrow> <munderover> <mi>Σ</mi> <mrow> <mi>c</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msubsup> <mi>sim</mi> <mrow> <mi>a</mi> <mo>,</mo> <msub> <mi>u</mi> <mi>c</mi> </msub> </mrow> <mo>′</mo> </msubsup> </mrow> </mfrac> <mo>,</mo> <mo>;</mo> </mrow> </math>

wherein,for user a and similar user u in similar user group_cThe natural number subscript c is the serial number of the similar users in the similar user group, and the maximum value is K.

The invention has carried on many times of example tests on the basis of QoS data set collected in the mobile Internet, the result of the test is successful, have achieved the purpose of the invention.

Claims

1. A method for predicting quality of service (QoS) of Web service in mobile Internet environment is characterized in that: firstly, QoS (quality of service) original data of a certain Web service repeatedly accessed by a certain user under the environment of a mobile internet are collected, and the QoS original data are preprocessed in a logarithmic processing mode; then, finding out a similar user group of a target user through an improved user-based collaborative filtering method, further determining a normal value interval according to the characteristics of QoS attributes, and finally calculating to obtain a credible accurate QoS predicted value; the method comprises the following operation steps:

step 2, calculating the similarity between users: calculating a Pearson Correlation coefficient PCC (Pearson Correlation coefficient) between multiple times of QoS data obtained by two users on each Web service commonly accessed by the two users by adopting an improved user-based collaborative filtering method, and converting the Pearson Correlation coefficient calculated by the two users on all the Web services commonly accessed by the two users into a similarity value for representing the user similarity between the two users; then, the similarity between all users is obtained by adopting the mode;

2. The method of claim 1, wherein: the step 1 comprises the following operation contents:

(11) setting the QoS raw data set of t repeated accesses of the ith user to the jth Web service asThe natural numbers i, j and r are respectively the user serial number, the Web service serial number accessed by the user and the serial number of the repeated access times of the Web service serial number, and the maximum value of r is t; q. q.s^r _i,jQoS raw data for executing the access of the ith time to the jth Web service for the ith user;

(12) setting a QoS raw data set Q of t repeated accesses of ith user to jth Web service^Original _i,jThe QoS data set of each QoS raw data after being preprocessed is Q_i,j＝{p¹ _i,j,p² _i,j,…,p^r _i,j,…,p^t _i,j}; wherein p is^r _i,jQoS raw data q for ith user to jth Web service access r^r _i,jQoS data after preprocessing is carried out;

3. The method of claim 1, wherein: the improved user-based collaborative filtering method is that similar user individuals of a target object are found based on neighborhoods among users according to historical behavior data of the users, so that the behavior data of the target object is predicted, and the purpose of individual recommendation is achieved.

4. The method of claim 3, wherein: in the step 2, because the calculation is simple and the precision is high, the similarity between users adopts a method for calculating a Pearson correlation coefficient PCC; however, due to the characteristics of QoS data and its data set, the PCC calculation method needs to be improved: the personality characteristics of the user of the QoS data when the user accesses the Web service are considered, and the personality characteristics of the volatility of the QoS data when the user repeatedly accesses the Web service are considered.

5. The method of claim 4, wherein: the operation of calculating the similarity between users in step 2 includes the following steps:

(21) when calculating the similarity between two users a and b, firstly, find out the set US of the Web service commonly visited by the two users_a∩US_b＝{s₁,s₂,s₃,…,s_j,…,s_nIn the formula, s_jFor the j-th Web service commonly accessed by the two users a and b, the natural numbers j and n are respectively the serial number of the Web service commonly accessed by the two users a and b and the total number of the Web services commonly accessed, namely the maximum value of j is n;

and

the variance of the QoS data obtained when the user a and the user b access the jth Web service repeatedly t times respectively,is all t QoS data p¹ _a,j+p² _a,j+p³ _a,j+…+p^t _a,jThe cumulative sum of;

(23) the calculation process of the step (22) only considers the two users a andb, PCC of QoS data obtained when the Web services are accessed together is solved, and the average value of the PCC is solved, so that the influence on the calculation result when the number n of the Web services which are accessed together is small is not eliminated; therefore, the calculation result sim of step (22) is dealt with_a,bBy the following formulaCarrying out weighting processing; in formula (II) sim'_a,bFor the calculation result sim_a,bThe numerical value after weighting processing is used for representing the final user similarity of the two users a and b; | US_a∩US_bI is the number of Web services that the two users have accessed together, | US_aI and I US_bI is the number of Web services respectively accessed by the user a and the user b respectively;

6. The method of claim 4, wherein: the step 3 comprises the following operation contents:

(31) if missing QoS data of a certain user a to a certain Web service w is to be predicted, firstly, similarity data between the user a and all users is obtained according to the step 2, and then all users having historical access to the Web service w are arranged in a descending order according to the similarity value of the users a;

(32) and (4) according to the descending order obtained in the step (31), the first K users form a similar user group of the user a: s_sam(a)＝{u₁,u₂,u₃,…,u_c,…,u_KIn the formula, the natural number c is the serial number of the similar user in the similar user group of the user a, and the maximum value is K, u_cIs a similar group of users S_sam(a)C-th similar user in (1); the size of the user number K depends on a data set according to which the user number K is calculated;

(33) setting QoS data sets of K similar users in user a similar user groupIs a set ofWherein,repeatedly accessing the QoS data set of the jth Web service for the jth similar user in the c similar user group of the user a for t times;

(34) solving each similar user u in the similar user group_cQoS data set obtained when repeatedly accessing jth Web serviceExpected value ofWherein,is composed ofThe r-th QoS data element of (1); the natural number r is the access frequency serial number of the user repeatedly accessing a certain Web service, and the maximum value is t;

(35) because the QoS data attribute comprises a positive attribute or a negative attribute, wherein the response time is the negative attribute, and the throughput is the positive attribute; therefore, from the collected QoS raw data, it can be found that: when the network condition becomes worse to affect the information transmission, the response time value tends to become larger, and the throughput value tends to become smaller; therefore, the normal value interval of the forward attribute QoS data isThe normal value interval of the negative attribute QoS data isIn the formula,for the c-th similar user u in the similar user group_cQoS data set obtained when repeatedly accessing jth Web serviceExpected value of p_minAnd p_maxRespectively corresponding QoS data setsMinimum and maximum values of;

and isWherein,to representIs composed ofA subset of (1), i.e.Is a slave QoS data setSelected from the group consisting ofQoS data inIs positive attribute, its value is in the intervalInternal; if setQoS data inNegative attribute, its value is in the intervalIn the table, the natural number e is the number of QoS data, and the maximum value is v.

7. The method of claim 6, wherein: in the step (31), the number K of the similar user groups of the user a is a value when the predicted error value is verified and confirmed to be minimum after a plurality of times of experimental selections are performed on the data set.

8. The method of claim 1, wherein: the step 4 comprises the following operation contents:

(41) set of sequential pairsScreening abnormal value data according to the step (36) for each parameter to obtain a screened QoS data set

wherein,for user a and similar user u in similar user group_cThe natural number subscript c is in the similar user groupLike the user number, the maximum value is K.