JP2007053679A - System and method for traffic amount prediction - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To predict also about a traffic amount in a network with imbalanced degree distribution. <P>SOLUTION: A traffic information acquisition unit 21 acquires traffic information from a traffic information collecting device 10 for every collecting period. A network structure information calculation unit 22 calculates distribution of a terminal number and an average value of speech time for each number of called parties in an object period based on the acquired traffic information as network structure information. A prediction network structure information calculation unit 23 calculates prediction network structure information based on the calculated network structure information. A prediction traffic amount calculation unit 24 calculates a prediction traffic amount of the all network based on the calculated prediction network structure information. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a traffic volume prediction system and a traffic volume prediction method.

Conventionally, as a method for predicting the future communication traffic volume, for example, the traffic volume is classified by time attribute such as month, day of the week, or by user attribute such as gender and age, and the average value of the traffic volume in each attribute There is a method for predicting the future traffic volume by calculating the variance of these values after calculating the variance. Patent Document 1 and Patent Document 2 listed below disclose techniques for predicting the total traffic volume by predicting the average traffic volume for each time attribute and summing the average values of all attributes.
JP-A-4-329052 JP 2004-23114 A

Recently, many networks such as acquaintances and the Internet do not belong to a so-called random network according to a normal distribution, but belong to a so-called scale-free network according to a power law (power law). (Reference: "Science of complex networks", Naoki Masuda and Norio Imano, Sangyo Tosho, "Chapter 2 2.2 Entrance to Graph Theory"). As shown in FIG. 10, the degree distribution P (k) [k: degree] of the random network is represented by a bell-shaped graph centered on the average value. On the other hand, the degree distribution P (k) of the scale-free network is represented by a relationship proportional to k ^−γ (γ is a power exponent), and is represented by a graph as shown in FIG. Comparing the degree distributions of both networks, it can be seen that in the scale-free network, a value greatly deviating from the average value exists with a higher probability than in the random network.

  Therefore, as in the prior art, in the method of predicting the future traffic volume based on the average value of the past traffic volume, it is possible to predict the traffic volume in a network having a biased degree distribution represented by a scale-free network. It becomes difficult.

  Therefore, in order to solve the above-described problems, an object of the present invention is to provide a traffic amount prediction system and a traffic amount prediction method that can also predict a traffic amount in a network having a biased degree distribution.

  A traffic volume prediction system according to the present invention includes a traffic information collecting unit that collects traffic information for each link in a network formed by a plurality of communication nodes and a link connecting the communication nodes, and a traffic collected by the traffic information collecting unit. Based on the information, it is calculated by network structure information calculating means for calculating the average number of communication nodes and communication time for each number of links in a predetermined period as network structure information indicating the structure of the network, and network structure information calculating means. Prediction network structure information calculating means for calculating predicted network structure information as future network structure information based on network structure information in a plurality of different periods, and a prediction network calculated by the prediction network structure information calculating means. Based on the work structure information, characterized in that it comprises a prediction traffic amount calculating means for calculating the predicted amount of traffic is the amount future traffic.

  The traffic volume prediction method of the present invention is collected in a traffic information collecting step for collecting traffic information for each link in a network formed by a plurality of communication nodes and links connecting the communication nodes, and a traffic information collecting step. Network structure information calculating step for calculating the average value of the number of communication nodes and the communication time for each number of links in a predetermined period as network structure information indicating the network structure based on the traffic information, and in the network structure information calculating step A predicted network structure information calculating step for calculating predicted network structure information as future network structure information based on the calculated network structure information in a plurality of different periods; Tsu based on the prediction network structure information calculated in-flop, characterized in that it comprises the prediction traffic calculating step of calculating the predicted amount of traffic is the amount future traffic.

  According to these inventions, the average number of communication nodes and communication time for each number of links in a predetermined period is calculated as network structure information. Thereby, the link relationship between each communication node and the change of the communication time in each link can be grasped as the change of the network structure. Further, the predicted network structure information is calculated based on the calculated network structure information in a plurality of different periods, and the future traffic amount is calculated based on the calculated predicted network structure information. Thereby, the traffic amount according to the change of the network structure can be calculated. Therefore, it is possible to predict not only the traffic amount in a random network but also the traffic amount in a network having a biased degree distribution represented by a scale-free network.

  In the traffic volume prediction system according to the present invention, the predicted network structure information calculation unit calculates a change ratio or difference of the network structure information in a plurality of different periods, and calculates the calculated change ratio or difference as predetermined network structure information. The predicted network structure information is preferably calculated by multiplying or adding to.

  In the traffic volume prediction system according to the present invention, the predicted traffic volume calculation means calculates the traffic volume for each number of links by multiplying the number of communication nodes for each number of links included in the predicted network structure information by the average value of communication time. After that, it is preferable to calculate the predicted traffic volume by totaling the traffic volume for every number of links.

  The traffic volume prediction system of the present invention further comprises network structure information storage means for storing the network structure information calculated by the network structure information calculation means, wherein the predicted network structure information calculation means is calculated by the network structure information calculation means. It is preferable to calculate the predicted network structure information based on the network structure information and the network structure information in one or more periods stored by the network structure information storage means.

  According to the traffic amount prediction system and the traffic amount prediction method according to the present invention, it is possible to predict the traffic amount in a network having a biased degree distribution typified by a scale-free network with high accuracy.

  DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments of a traffic volume prediction system and a traffic volume prediction method according to the present invention will be described with reference to the drawings. In addition, in each figure, the same code | symbol is attached | subjected to the same element and the overlapping description is abbreviate | omitted.

  FIG. 1 is a diagram illustrating a system configuration of a traffic amount prediction system according to an embodiment. As shown in FIG. 1, the traffic volume prediction system 1 includes a traffic information collection device 10 (traffic information collection means) and a traffic volume prediction device 20 installed inside a mobile communication network.

  The traffic information collection device 10 acquires voice call history information of the mobile phone 50 (communication node) managed in the mobile communication network. Thereby, voice call history information is collected as traffic information. The traffic information collection device 10 sends the acquired traffic information to the traffic amount prediction device 20. Here, the traffic information is voice call history information for each link connecting the mobile phones 50, and examples of the traffic information include a transmission terminal ID, a reception terminal ID, a call start time, and a call time.

  The traffic amount prediction device 20 includes a traffic information acquisition unit 21 (traffic information collection unit), a network structure information calculation unit 22 (network structure information calculation unit), and a prediction network structure information calculation unit 23 (prediction network structure information calculation unit). And a predicted traffic amount calculation unit 24 (predicted traffic amount calculation means) and an output unit 25. The traffic volume prediction device 20 includes a traffic information database 27 and a network structure information database 28.

  The traffic information acquisition unit 21 acquires the traffic information collected in the traffic information collection device 10 every predetermined collection period. The predetermined collection period can be arbitrarily set according to the operation, for example, 1 hour, 1 day, 1 month, or the like. The traffic information acquisition unit 21 stores the traffic information acquired from the traffic information collection device 10 in the traffic information database 27.

  Here, the data structure of the traffic information database 27 will be described with reference to FIG. As shown in FIG. 2, the traffic information database 27 includes, for example, a transmission terminal ID, a reception terminal ID, a call start time, and a call time as data items. In the calling terminal ID, an ID for uniquely specifying the calling-side mobile phone 50 is stored, and in the receiving terminal ID, an ID for uniquely specifying the receiving-side mobile phone 50 is stored. The call start time stores the time when the call is started between the calling-side mobile phone 50 and the called-side mobile phone 50, and the calling time includes the calling-side mobile phone 50 and the called-side mobile phone. The call time between 50 is stored.

  Based on the traffic information stored in the traffic information database 27, the network structure information calculation unit 22 shown in FIG. 1 determines the number of terminals (number of mobile phones) for each number of call partners (number of links) in a predetermined target period. An average value of the call time of the mobile phone 50 for each number of call destinations in a predetermined target period is calculated. The calculated average number of terminals and call duration for each number of call destinations is network structure information indicating the network structure. The predetermined target period can be arbitrarily set, for example, one day, one month, or one year.

  The method for calculating the network structure information will be specifically described. First, the network structure information calculation unit 22 extracts traffic information for a predetermined target period from the traffic information database 27. Specifically, for example, when the predetermined target period is August 2005, traffic information including the call start time in August 2005 is extracted as shown in FIG.

  Next, the network structure information calculation unit 22 classifies the extracted traffic information into sets of a transmission terminal ID and a reception terminal ID, and calculates a total call time in each group (see FIG. 3). More specifically, for example, it is assumed that a pair of a mobile phone with a calling terminal ID “1” and a mobile phone with a receiving terminal ID “2” shown in FIG. Assuming that “79 seconds”, which is the sum of the call time “42 seconds” of 8/1 and the call time “37 seconds” of 8/9, is calculated as the total call time in this group (see FIG. 3).

  Next, the network structure information calculation unit 22 calculates the number of call partners for each calling terminal ID and the total call time based on the calculated total call time for each set (see FIG. 4). More specifically, for example, a mobile phone having a calling terminal ID “1” shown in FIG. 3 is a mobile phone having a receiving terminal ID “2”, “4”, “11” in August 2005. When it is assumed that only the telephone call is made, “3” is calculated as the number of the other party of the calling terminal ID “1”, and “79 seconds”, “291 seconds”, “ “472 seconds”, which is the sum of “102 seconds”, is calculated as the total call time of the calling terminal ID “1” (see FIG. 4).

  Next, the network structure information calculation unit 22 calculates the total number of terminals and the average call time for each number of call destinations based on the calculated number of call destinations for each calling terminal ID and the total call time (FIG. 5). reference). More specifically, the total number of terminals for each number of call destinations shown in FIG. 5 is calculated by counting the number of mobile phone terminals for each number of call destinations shown in FIG. By calculating the total call time for each number of destinations and dividing by the total number of terminals for each number of call destinations, the average call time for each number of call destinations shown in FIG. 5 is calculated. Thereby, network structure information is calculated.

  The network structure information calculation unit 22 stores the calculated network structure information in the network structure information database 28.

  Here, the data structure of the network structure information database 28 will be described with reference to FIG. As shown in FIG. 6, the network structure information database 28 includes, for example, year, month, number of call destinations, total number of terminals, and average call time as data items. In the network structure information database 28 composed of such data items, the network structure information shown in FIG. 5 is classified and stored for each generation date of traffic information which is the basis of the network structure information.

  The predicted network structure information calculation unit 23 shown in FIG. 1 is based on the network structure information calculated by the network structure information calculation unit 22 and the network structure information in one or more past periods stored in the network structure information database 28. The predicted network structure information, which is the future network structure information, is calculated. More specifically, the predicted network structure information calculation unit 23 calculates a change rate or difference of network structure information in a plurality of different periods, and multiplies predetermined network structure information by the calculated change rate or difference. Prediction network structure information is calculated by adding.

  Here, the method for calculating the predicted network structure information will be described more specifically. For example, when the network structure information calculated by the network structure information calculation unit 22 is network structure information of the current month, and the past network structure information extracted from the network structure information database 28 is network structure information of the previous month, Based on the change between the two network structure information, the change in the number of terminals, and the expected total number of terminals in the next month, the network structure information of the next month, which is the predicted network structure information, is calculated as follows.

  First, a method for calculating the number of terminals for each number of call destinations included in the predicted network structure information will be described. The graph shown in FIG. 7 is a graph showing the logarithmic relationship when the number of call destinations is k and the number of mobile phone terminals is P (k). As can be seen from this graph, the distribution of the number of terminals according to the number of call destinations is a biased distribution in which the average value and the mode value do not match unlike the random network. In such a network, as the number of terminals increases, the number of call destinations of individual users also increases, so the distribution of the number of terminals by number of call destinations moves in the increasing direction on the horizontal axis (number of call destinations). That is expected.

  Therefore, we decided to calculate the number of terminals for each number of callers in the following month as follows. First, the amount of change in the horizontal direction during this period at multiple points on the curve that continuously approximates the distribution of the number of terminals by the number of call destinations in the previous month and the distribution of the number of terminals by number of call destinations in the current month. Ask for. Next, assuming that the rate of change at each point is the same between the current month and the next month, the distribution of the number of terminals by the number of call partners for the next month is calculated. That is, the amount of change in the horizontal axis direction at each point between the current month and the next month is obtained by multiplying the amount of change in the horizontal axis direction at each point between the previous month and the current month by a certain ratio a. Then, based on the obtained variation, each point on the distribution curve for the current month is moved in the horizontal axis direction, thereby calculating the number of terminals for each number of call destinations for the next month.

  Note that the fixed ratio a is determined so that the expected total number of terminals in the next month matches the total number of terminals determined from the distribution curve in the next month. More specifically, for example, when the amount of change is constant at all points on the distribution curve, the distribution of the number of terminals for each number of call destinations slides in the horizontal axis direction as it is. At this time, the distribution curve of the number of terminals according to the number of call destinations of the current month and the distribution curve of the number of terminals according to the number of call destinations of the next month are P2 (k) and P3 (k), respectively. When the number of terminals is n, P3 (k) = P2 (k / a) holds and ΣP3 (k) = ΣP2 (k / a) = n holds. Thereby, the fixed ratio a can be determined.

  Next, a method for calculating the average call time for each number of call destinations included in the predicted network structure information will be described. The graph shown in FIG. 8 is a graph showing the relationship when the number of call destinations is k and the average call time of the mobile phone is T (k). As can be seen from this graph, the number of call destinations and the average call time are in a proportional relationship. Therefore, a change can be predicted according to the ratio of the difference between the average call time for each number of call destinations in the previous month and the average call time for each number of call destinations in the current month.

  Specifically, first, the difference in the average call time between the previous month and the current month for each number of call partners is the difference between the total number of terminals in the current month and the expected total number of terminals in the next month relative to the difference in the total number of terminals in the previous month and the current month. A value obtained by multiplying the difference ratio is calculated. Next, by adding the calculated value to the average call time of the current month for each number of call destinations, the average call time for the number of call destinations of the next month is calculated. More specifically, referring to FIG. 6, for example, if the current month is August 2005, the difference in average call time when the number of call partners is “1” is (t21− t11). The difference between the total number of terminals in the previous month (July) and the current month (August) is N1, and the difference between the total number of terminals in the current month (August) and the expected total number of terminals in the next month (September) is N2. To do. In this case, the average call time when the number of call partners in the next month (September) is “1” is calculated as t21 + (t21−t11) · N2 / N1.

  The predicted traffic amount calculation unit 24 illustrated in FIG. 1 calculates a predicted traffic amount that is a future traffic amount based on the predicted network structure information calculated by the predicted network structure information calculation unit 23.

  More specifically, the total number of terminals for each number of call destinations included in the predicted network structure information is multiplied by the average call time to calculate the traffic volume for each number of call destinations. Next, the traffic volume for every number of call destinations is totaled, and the predicted traffic volume of the entire network is calculated.

  The output unit 25 outputs the predicted network structure information calculated by the predicted network structure information calculation unit 23 and the predicted traffic amount calculated by the predicted traffic amount calculation unit 24. The output destination corresponds to, for example, a display device, a file stored in a storage device, another external device, or the like.

  Next, with reference to FIG. 9, the flow of traffic amount prediction processing performed in the traffic amount prediction system 1 described above will be described.

  First, the traffic information collection device 10 collects the voice call history information of the mobile phone 50 managed in the mobile communication network as traffic information (step S1).

  Next, the traffic information acquisition unit 21 of the traffic volume prediction device 20 acquires traffic information from the traffic information collection device 10 for each arbitrarily set collection period (for example, one day) (step S2), and traffic. The information is stored in the information database 27 (step S3).

  Next, the network structure information calculation unit 22 of the traffic volume prediction device 20 uses the traffic information stored in the traffic information database 27 to determine the number of call partners in an arbitrarily set target period (for example, one month). The distribution of the number of other terminals and the average value of the call time are calculated as network structure information (step S4) and stored in the network structure information database 28 (step S5).

  Next, the predicted network structure information calculation unit 23 of the traffic amount prediction device 20 calculates the ratio or difference of the change in the network structure information between the previous month and the current month, and uses the calculated change ratio or difference as the network structure information for the current month. Prediction network structure information is calculated by multiplication or addition (step S6).

  Next, the predicted traffic amount calculation unit 24 of the traffic amount prediction device 20 multiplies the total number of terminals for each number of call destinations calculated by the predicted network structure information calculation unit 23 and the average call time to multiply the number of call destinations. The traffic volume for each network is calculated, and the calculated traffic volumes are totaled to calculate the predicted traffic volume for the entire network (step S7).

  Next, the output unit 25 of the traffic amount prediction device 20 displays the predicted network structure information calculated by the predicted network structure information calculation unit 23 and the predicted traffic amount calculated by the predicted traffic amount calculation unit 24 on the display screen of the display device. It is displayed on the top (step S8).

  As described above, since the average value of the number of mobile phones and the call time for each number of call destinations in a predetermined target period is calculated as network structure information, the link relationship between each mobile phone and the call on each link Changes in time can be viewed as changes in the actual network structure.

  Further, since the predicted network structure information is calculated based on the calculated network structure information in a plurality of different periods, and the future traffic amount is calculated based on the calculated predicted network structure information, the conventional traffic is calculated. Changes in the actual network structure that could not be handled by the prediction based on the average value of the amount can be adopted for the traffic amount prediction, and it is highly accurate according to the actual network structure. It becomes possible to predict the traffic volume. Therefore, not only the traffic amount in the random network, but also the traffic amount in the network having a biased degree distribution represented by the scale-free network can be predicted with higher accuracy.

  Note that the predicted network structure information calculation unit 23 in the above-described embodiment has the network structure information calculated by the network structure information calculation unit 22 and the network structure information in one or more past periods stored in the network structure information database 28. However, the method for calculating the predicted network structure information is not limited to this. Based on the network structure information in a plurality of different periods calculated by the network structure information calculation unit 22, it is only necessary to be able to calculate predicted network structure information that is future network structure information.

  In the above-described embodiment, the traffic volume prediction system 1 is configured by the two devices, the traffic information collection device 10 and the traffic volume prediction device 20, but the device configuration of the traffic volume prediction system 1 is as follows. It is not limited. For example, the traffic information collecting device 10 and the traffic amount predicting device 20 may be configured by one device having each function, or the traffic information collecting device 10 and the traffic amount predicting device 20 may be appropriately combined. It is good also as distributing to a plurality of devices.

  In the above-described embodiment, the traffic volume of a voice call is predicted, but the predicted traffic volume is not limited to that due to a voice call. For example, the present invention can also be applied to the case of predicting the traffic volume of video communication such as a TV phone, packet communication such as mail, and other communication in a network between users.

It is a figure which illustrates the system configuration | structure of the traffic amount prediction system in embodiment of invention. It is a figure which illustrates the data structure of a traffic information database. It is a figure for demonstrating the total call time for every group of the transmission terminal ID calculated by the network structure information calculation part, and a receiving terminal ID. It is a figure for demonstrating the number of call partners for every transmission terminal ID calculated by the network structure information calculation part, and the total call time. It is a figure for demonstrating the total number of terminals and the average call time for every number of other call parties calculated by the network structure information calculation part. It is a figure which illustrates the data structure of a network structure information database. It is a figure which illustrates the relationship between the number of other parties and the number of terminals. It is a figure which illustrates the relationship between the number of call partners and average call time. It is a flowchart which illustrates the flow of the traffic amount prediction process performed in a traffic amount prediction system. It is a figure which illustrates the degree distribution of a random network. It is a figure which illustrates degree distribution of a scale free network.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Traffic amount prediction system, 10 ... Traffic information collection apparatus, 20 ... Traffic amount prediction apparatus, 21 ... Traffic information acquisition part, 22 ... Network structure information calculation part, 23 ... Predicted network structure information calculating unit 24... Predicted traffic amount calculating unit 25... Output unit 27... Traffic information database 28.

Claims (5)

Traffic information collecting means for collecting traffic information for each link in a network formed by a plurality of communication nodes and links connecting the communication nodes;
Network structure information for calculating, as network structure information indicating the structure of the network, an average value of the number of communication nodes and the communication time for each number of links in a predetermined period based on the traffic information collected by the traffic information collection unit A calculation means;
Predicted network structure information calculating means for calculating predicted network structure information that is the network structure information in the future based on the network structure information in a plurality of different periods calculated by the network structure information calculating means;
Based on the predicted network structure information calculated by the predicted network structure information calculating means, a predicted traffic amount calculating means for calculating a predicted traffic amount that is a future traffic amount;
A traffic volume prediction system comprising:   The predicted network structure information calculation means calculates a change ratio or difference of the network structure information in a plurality of different periods, and multiplies or adds the calculated change ratio or difference to the predetermined network structure information. The traffic volume prediction system according to claim 1, wherein the prediction network structure information is calculated as described above.   The predicted traffic amount calculating means calculates the traffic amount for each link number by multiplying the communication node number for each link number included in the predicted network structure information and the average value of the communication time, The traffic volume prediction system according to claim 1, wherein the predicted traffic volume is calculated by counting the traffic volume for each of the links. Network structure information storage means for storing the network structure information calculated by the network structure information calculation means;
The predicted network structure information calculating means is configured to perform the prediction based on the network structure information calculated by the network structure information calculating means and the network structure information in one or more periods stored by the network structure information storage means. The traffic volume prediction system according to claim 1, wherein network structure information is calculated. A traffic information collecting step for collecting traffic information for each link in a network formed by a plurality of communication nodes and links connecting the communication nodes;
Network structure information for calculating an average value of the number of communication nodes and communication time for each number of links in a predetermined period as network structure information indicating the structure of the network based on the traffic information collected in the traffic information collecting step A calculation step;
A predicted network structure information calculating step for calculating predicted network structure information that is the network structure information in the future based on the network structure information in a plurality of different periods calculated in the network structure information calculating step;
A predicted traffic amount calculating step for calculating a predicted traffic amount that is a future traffic amount based on the predicted network structure information calculated in the predicted network structure information calculating step;
A traffic amount prediction method comprising:

Images