Network device and model learning method

Abstract

To provide a network device and a model learning method that allow quality information to be estimated with high accuracy.SOLUTION: A network device includes: a connection status calculation unit that calculates first connection status information indicating a connection status between a terminal device and a plurality of application devices on the basis of an acquisition situation in a capture device of packets transmitted from at least one of the plurality of application devices to the terminal device via a network; and a model learning unit that generates a learning model by learning teacher data including the calculated first connection status information, first communication status information indicating a communication status of the packets in the network, and first quality information indicating quality related to output of the packets at the terminal device.SELECTED DRAWING: Figure 23

Description

[Technical field]

[0001] The present invention relates to a network device and a model learning method. [Background technology]

[0002] In a video distribution system (hereinafter also simply referred to as a video distribution system or an information processing system) that distributes video data to a terminal device via a network, communication state information that can be determined to have a large influence on quality information is identified based on a correlation between information indicating a communication state such as a delay time or a packet loss rate in a network (hereinafter also referred to as communication state information) and information indicating the playback quality of the video data in a terminal device or the quality of experience of a user viewing the video data (hereinafter also referred to as quality information).The video distribution system then generates a learning model (hereinafter also simply referred to as a learning model) that can estimate quality information by learning the identified communication state information as an explanatory variable (see, for example, Patent Documents 1 to 4). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2012-157004 A [Patent Document 2] Patent Publication No. 2022-042005 [Patent Document 3] International Publication No. 2019 / 004013 [Patent Document 4] Special Publication No. 2016-500952 Summary of the Invention [Problem to be solved by the invention]

[0004] Here, in the above-mentioned video distribution system, for example, the correlation between the timing when the communication state information changes in the network and the timing when the quality information changes in the terminal device may not be strong enough. Therefore, in the video distribution system, even if the above-mentioned learning model is used, for example, the quality information may not be estimated with high accuracy.

[0005] Therefore, in one aspect, the present invention has an object to provide a network device and a model learning method that enable quality information to be estimated with high accuracy. [Means for solving the problem]

[0006] In one aspect of the embodiment, the device includes a connection status calculation unit that calculates first connection status information indicating a connection status between the terminal device and the multiple application devices based on an acquisition status in a capture device of a packet transmitted from at least one of the multiple application devices to the terminal device via a network, and a model learning unit that generates a learning model by learning teacher data including the calculated first connection status information, first communication status information indicating a communication status of the packet in the network, and first quality information indicating a quality related to the output of the packet in the terminal device. Effect of the Invention

[0007] According to one aspect, it is possible to estimate quality information with high accuracy. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram illustrating the configuration of an information processing system 10. As shown in FIG. [Diagram 2] FIG. 2 is a diagram illustrating the configuration of the information processing system 10. As shown in FIG. [Diagram 3] FIG. 3 is a diagram illustrating the hardware configuration of the information processing device 1. As shown in FIG. [Figure 4]FIG. 4 is a diagram illustrating functions of the information processing device 1 in the first embodiment. [Diagram 5] FIG. 5 is a diagram illustrating functions of the information processing device 1 in the first embodiment. [Figure 6] FIG. 6 is a flow chart for explaining an outline of the learning process in the first embodiment. [Figure 7] FIG. 7 is a flowchart illustrating details of the learning process in the first embodiment. [Figure 8] FIG. 8 is a flowchart illustrating details of the learning process in the first embodiment. [Figure 9] FIG. 9 is a flowchart illustrating details of the learning process in the first embodiment. [Figure 10] FIG. 10 is a flowchart illustrating details of the learning process in the first embodiment. [Figure 11] FIG. 11 is a flowchart illustrating details of the learning process in the first embodiment. [Figure 12] FIG. 12 is a flowchart illustrating details of the learning process in the first embodiment. [Figure 13] FIG. 13 is a flowchart illustrating details of the learning process in the first embodiment. [Figure 14] FIG. 14 is a diagram for explaining a specific example of the quality information 132. As shown in FIG. [Figure 15] FIG. 15 is a diagram for explaining a specific example of the communication state information 133. As shown in FIG. [Figure 16] FIG. 16 is a diagram illustrating details of the learning process in the first embodiment. [Figure 17] FIG. 17 is a diagram illustrating a specific example of the connection state information 134. As shown in FIG. [Figure 18] FIG. 18 is a diagram illustrating a specific example of the connection state information 134. As shown in FIG. [Figure 19]FIG. 19 is a diagram illustrating a specific example of the session information 135. As shown in FIG. [Figure 20] FIG. 20 is a diagram illustrating a specific example of the connection state information 134. As shown in FIG. [Figure 21] FIG. 21 is a diagram illustrating a specific example of the teacher data DT. [Figure 22] FIG. 22 is a flowchart illustrating the estimation process in the first embodiment. [Figure 23] FIG. 23 is a diagram illustrating the estimation process in the first embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. However, such descriptions should not be interpreted in a limiting sense and do not limit the subject matter described in the claims. In addition, various changes, substitutions, and modifications can be made without departing from the spirit and scope of the present disclosure. In addition, different embodiments can be appropriately combined.

[0010] [Configuration of the information processing system in the first embodiment] First, a description will be given of the configuration of the information processing system 10. Figures 1 and 2 are diagrams for explaining the configuration of the information processing system 10.

[0011] As shown in FIG. 1, the information processing system 10 includes, for example, an information processing device 1 (hereinafter also referred to as a network device 1), a video distribution device 2a, a video distribution device 2b, a video distribution device 2c, a video distribution device 2d, a capture device 3, a terminal device 4a, a terminal device 4b, and a terminal device 4c. Hereinafter, the terminal devices 4a, 4b, and 4c are collectively referred to simply as terminal devices 4. Hereinafter, the information processing system 10 will be described as having three terminal devices 4, but the information processing system 10 may have, for example, a number of terminal devices 4 other than three. Hereinafter, the video distribution device 2a, the video distribution device 2b, the video distribution device 2c, and the video distribution device 2d are collectively referred to simply as video distribution device 2 or application device 2. Hereinafter, the information processing system 10 will be described as having four video distribution devices 2, but the information processing system 10 may have, for example, a number of video distribution devices 2 other than four.

[0012] The video distribution device 2 is, for example, a physical machine or a virtual machine, and is a device that distributes video data via a network NW.

[0013] Specifically, the video distribution device 2a is, for example, a video distribution device 2 capable of distributing high-quality video data. The video distribution device 2b is, for example, a video distribution device 2 capable of distributing low-quality video data. The video distribution device 2c is, for example, a video distribution device 2 capable of distributing high-quality video data, and is located closer physically to the terminal device 4 than the video distribution device 2a. The video distribution device 2d is, for example, a video distribution device 2 capable of distributing low-quality video data, and is located closer physically to the terminal device 4 than the video distribution device 2b.

[0014] The terminal device 4 is, for example, a terminal such as a PC (Personal Computer) or a smartphone used by a user (hereinafter simply referred to as the user) who views video data, and receives packets continuously transmitted from the video distribution device 2 via the network NW, and outputs the video data composed of each of the received packets to an output device (not shown).

[0015] Specifically, the terminal device 4 selects, for example, from among a plurality of video distribution devices 2, a video distribution device 2 that corresponds to the communication state in the network NW, and receives packets transmitted from the selected video distribution device 2. That is, the terminal device 4 selects, for example, a video distribution device 2 that is as close as possible and that can distribute video data with the best possible image quality within a range where playback of the video data is not interrupted.

[0016] More specifically, for example, when it is determined that the communication state in the network NW is such that high-quality video data can be acquired and that the video data is stored in the video distribution device 2c, the terminal device 4 requests the video distribution device 2c to distribute the video data (high-quality video data) and receives the packets transmitted from the video distribution device 2c. On the other hand, when it is determined that the communication state in the network NW is such that high-quality video data can be acquired and that the video data is not stored in the video distribution device 2c, the terminal device 4 requests the video distribution device 2a to distribute the video data (high-quality video data) and receives the packets transmitted from the video distribution device 2a.

[0017] Also, for example, when it is determined that the communication state in the network NW is not in a state in which high-quality video data can be acquired (in other words, when it is determined that low-quality video data can be acquired), and when it is determined that the video data is stored in the video distribution device 2d, the terminal device 4 requests the video distribution device 2d to distribute the video data (low-quality video data) and receives the packets transmitted from the video distribution device 2d. On the other hand, when it is determined that the communication state in the network NW is not in a state in which high-quality video data can be acquired, and when it is determined that the video data is not stored in the video distribution device 2d, the terminal device 4 requests the video distribution device 2b to distribute the video data (low-quality video data) and receives the packets transmitted from the video distribution device 2b.

[0018] The capture device 3, for example, captures a packet transmitted from the video distribution device 2 to the terminal device 4. Specifically, the capture device 3, for example, copies the packet transmitted from the video distribution device 2 to the terminal device 4 and captures the copied packet. Then, the capture device 3 transmits the captured packet to the information processing device 1, for example.

[0019] The network NW is, for example, a network including the Internet IN. Specifically, in the network NW, for example, at least a part of the connection between the video distribution device 2 and the terminal device 4 is made by wire. Also, in the network NW, for example, at least a part of the connection between the video distribution device 2 and the terminal device 4 is made by wireless. Note that the wireless connection here may be, for example, a connection in accordance with the communication standard of the 5th Generation Mobile Communication System (5G).

[0020] More specifically, the network NW may include, for example, a core network (CN) 5 and a radio access network (Radio Access Network) 6, as shown in Fig. 2. The core network 5 may be connected to, for example, an Internet network IN and the radio access network 6 by wire. The radio access network 6 may be connected to, for example, each terminal device 4 by wireless. Furthermore, the capture device 3 may be installed between, for example, the core network 5 and the radio access network 6.

[0021] 1 and 2, a case will be described in which the capture device 3 is disposed between the video distribution device 2 and the terminal device 4 in the network NW, and the capture device 3 directly captures packets transmitted from the video distribution device 2 to the terminal device 4, but the present invention is not limited to this. Specifically, the packets transmitted from the video distribution device 2 to the terminal device 4 may be captured by a device such as a TAP (Terminal Access Point) disposed between the core network 5 and the wireless access network 6, or a TAP (e.g., a virtual TAP) constituting a part of the wireless access network 6. The capture device 3 may capture packets transmitted from the video distribution device 2 to the terminal device 4 by receiving packets transmitted from a device such as a TAP.

[0022] The information processing device 1 is, for example, a physical machine or a virtual machine, and performs a process of generating a learning model using teacher data generated based on the acquisition status of packets in the capture device 3 (hereinafter, also referred to as a learning process).

[0023] Specifically, the information processing device 1 calculates connection state information (hereinafter also referred to as first connection state information) indicating a connection state between the terminal device 4 and the multiple video distribution devices 2, for example, based on the acquisition status of packets transmitted from at least one of the multiple video distribution devices 2 to the terminal device 4 via the network NW. Then, the information processing device 1 generates a learning model by learning teacher data including, for example, the calculated first connection state information, communication state information (hereinafter also referred to as first communication state information) indicating a communication state of packets in the network NW, and quality information (hereinafter also referred to as first quality information) indicating quality related to the output of packets in the terminal device 4. The communication state information is, for example, KPIs (Key Performance Indicators) such as delay time and packet loss rate for transmitting and receiving packets in the network NW. In addition, the quality information is, for example, QoE (Quality of Experience) in the terminal device 4.

[0024] Furthermore, the information processing device 1 performs a process of estimating quality information in the network NW (hereinafter also referred to as an estimation process) by using, for example, the learning model generated in the learning process.

[0025] Specifically, the information processing device 1 calculates connection state information (hereinafter also referred to as second connection state information) indicating the connection state between the terminal device 4 and the multiple video distribution devices 2 based on the acquisition status of packets transmitted from the video distribution device 2 to the terminal device 4 via the network NW, for example, as in the case of the learning process. Then, the information processing device 1 acquires quality information (hereinafter also referred to as second quality information) output from the learning model in response to input of the calculated second connection state information and communication state information (hereinafter also referred to as second communication state information) indicating the communication state of packets in the network NW, for example. After that, the information processing device 1 outputs the acquired second quality information, for example.

[0026] That is, in the information processing device 1 of this embodiment, for example, when generating a learning model capable of estimating quality information, learning is performed using teacher data that includes not only communication state information and quality information, but also connection state information indicating the connection state between the terminal device 4 and each video distribution device 2.

[0027] As a result, the information processing device 1 in this embodiment is able to generate a learning model that can accurately estimate quality information such as QoE even when the correlation between the timing at which communication state information changes in the network NW and the timing at which a change in quality information occurs in the terminal device 4 is not sufficiently strong.

[0028] In the following, the learning process and the estimation process are performed in the same information processing device (information processing device 1), but the present invention is not limited to this. Specifically, the learning process and the estimation process may be performed in different information processing devices.

[0029] [Hardware configuration of information processing device] Next, a description will be given of the hardware configuration of each information processing device 1. Fig. 3 is a diagram for explaining the hardware configuration of the information processing device 1.

[0030] 3, the information processing device 1 includes, for example, a CPU (Central Processing Unit) 101, which is a processor, a memory 102, a communication device (I / O interface) 103, and a storage device 104. Each unit is connected to each other via a bus 105.

[0031] The storage device 104 has, for example, a program storage area (not shown) that stores a program 110 for performing the learning process and the estimation process. The storage device 104 also has, for example, a storage unit 130 (hereinafter also referred to as information storage area 130) that stores information used when performing the learning process and the estimation process. The storage device 104 may be, for example, a hard disk drive (HDD) or a solid state drive (SSD).

[0032] The CPU 101 executes, for example, a program 110 loaded from the storage device 104 to the memory 102 to perform learning processing and estimation processing.

[0033] Furthermore, the communication device 103 communicates with, for example, the capture device 3 .

[0034] [Functions of the information processing device in the first embodiment] Next, a description will be given of functions of the information processing device 1 in the first embodiment. Figures 4 and 5 are diagrams for explaining functions of the information processing device 1 in the first embodiment.

[0035] As shown in FIG. 4, the information processing device 1 realizes various functions including a packet acquisition unit 111, an information acquisition unit 112, a communication analysis unit 113, a connection status calculation unit 114, a data generation unit 115, a model learning unit 116, an information estimation unit 117, and an information utilization unit 118 by organic cooperation between hardware such as a CPU 101 and a memory 102 and a program 110.

[0036] 5, for example, packets 131, quality information 132, communication state information 133, connection state information 134, session information 135, teacher data DT, and learning model MD are stored in the information storage area 130. The session information 135 will be described later.

[0037] First, the functions in the learning process will be described.

[0038] The packet acquisition unit 111 receives, for example, packets transmitted from the capture device 3 (packets captured by the capture device 3). Then, the packet acquisition unit 111 extracts, for example, packets 131 transmitted from the video distribution device 2 from among the received packets. Thereafter, the packet acquisition unit 111 stores, for example, the extracted packets 131 in the information storage area 130. Specifically, the packet acquisition unit 111 stores, for example, the acquired packets 131 in the information storage area 130 in a state in which the terminal device 4 to which each packet 131 is sent and the session in which each packet 131 is transmitted and received can be identified.

[0039] More specifically, for example, for each packet transmitted from the capture device 3, the packet acquisition unit 111 determines whether each packet is a packet 131 transmitted from the video distribution device 2 from a combination of a source IP address included in an IP (Internet Protocol) header of each packet and a source port number included in a TCP (Transmission Control Protocol) header of each packet. Also, for each packet 131 determined to have been transmitted from the video distribution device 2, the packet acquisition unit 111 identifies the terminal device 4 to which each packet 131 is to be transmitted and the session in which each packet 131 is transmitted or received from a combination of a destination IP address included in an IP header of each packet 131 and a destination port included in a TCP header of each packet 131. Then, the packet acquisition unit 111 stores the packets 131 determined to have been transmitted from the video distribution device 2 in the information storage area 130 for each terminal device 4 and for each session.

[0040] The information acquiring unit 112 acquires, for example, for each terminal device 4, quality information 132 transmitted from each terminal device 4. Specifically, the information acquiring unit 112 acquires, for example, for each terminal device 4, quality information 132 input by each user via their own terminal device 4. Then, the information acquiring unit 112 stores the acquired quality information 132 in the information storage area 130, for example, for each terminal device 4.

[0041] The communication analysis unit 113 generates communication state information 133 from the acquisition status of the packets 131 by the packet acquisition unit 111, for example, for each terminal device 4. Then, the communication analysis unit 113 stores the generated communication state information 133 in the information storage area 130, for example, for each terminal device 4.

[0042] Specifically, the communication analysis unit 113 may calculate, for example, for each session and for each terminal device 4, the traffic volume per unit time of the packets 131 acquired by the packet acquisition unit 111 (hereinafter, also referred to as traffic volume 133a). Then, the communication analysis unit 113 may calculate, for example, for each terminal device 4, a value calculated from each of the calculated traffic volumes 133a (for example, a total value of each of the calculated traffic volumes 133a), as the communication state information 133.

[0043] Furthermore, the communication analysis unit 113 may calculate, for example, for each terminal device 4 and for each session, a time between the timing when the packet 131 acquired by the packet acquisition unit 111 is transmitted from the video distribution device 2 and passes through the capture device 3 and the timing when the reception acknowledgement (ACK: ACKnowledgement) of each packet 131 is transmitted from the terminal device 4 and passes through the capture device 3 (hereinafter, also referred to as a response required time). Then, the communication analysis unit 113 may calculate, for example, for each terminal device 4 and for each session, a time obtained by subtracting a predetermined threshold from the calculated response required time (hereinafter, also referred to as a delay time). After that, the communication analysis unit 113 may calculate, as the communication state information 133, a time calculated from each of the calculated delay times (for example, an average time of each of the calculated delay times), for example, for each terminal device 4.

[0044] Furthermore, the communication analysis unit 113 may calculate, for example, for each terminal device 4 and for each session, the proportion of packets 131 whose receipt confirmation responses do not pass through the capture device 3 among the packets 131 transmitted from the video distribution device 2 and passed through the capture device 3 (hereinafter also referred to as a packet loss rate).The communication analysis unit 113 may calculate, as the communication state information 133, a value calculated from each of the calculated packet loss rates (for example, an average value of each of the calculated packet loss rates), for example, for each terminal device 4.

[0045] The connection state calculation unit 114 generates, for example, for each terminal device 4, connection state information 134 indicating the connection state between each terminal device 4 and multiple video distribution devices 2. Then, the connection state calculation unit 114 stores the generated connection state information 134 in the information storage area 130, for example, for each terminal device 4.

[0046] Specifically, the connection state calculation unit 114, for example, for each terminal device 4 and for each session, identifies, from among the multiple video distribution devices 2, a video distribution device 2 (hereinafter also referred to as a first video distribution device 2) that first transmitted packets 131 constituting the video data being played on each terminal device 4 (hereinafter also referred to as simply the video data being played) to the terminal device 4, and a video distribution device 2 (hereinafter also referred to as a second video distribution device 2) that transmitted packets 131 constituting the video data being played to the terminal device 4 at a predetermined timing. The predetermined timing may be, for example, the timing when packets 131 constituting the video data being played were last transmitted from the second video distribution device 2 to the terminal device 4. In other words, the predetermined timing may be, for example, the timing immediately before packets 131 constituting the video data being played were transmitted from the second video distribution device 2 to the terminal device 4.

[0047] Then, the connection status calculation unit 114 calculates, for example, for each terminal device 4 and for each session, the similarity between the address information of the identified first video distribution device 2 and the address information of the second video distribution device 2. After that, the connection status calculation unit 114 stores, as connection status information 134, a value calculated from each of the calculated similarities (for example, the minimum value of each of the calculated similarities) for each terminal device 4 in the information storage area 130.

[0048] Furthermore, the connection status calculation unit 114 identifies, for example, each terminal device 4, the number of sessions in which packets 131 were transmitted from any of the multiple video distribution devices 2 to each terminal device 4. Then, the connection status calculation unit 114 stores the identified number of sessions as connection status information 134 in the information storage area 130, for example, each terminal device 4.

[0049] The connection state calculation unit 114 may specify, for example, for each terminal device 4, the number of sessions in which packets 131 have been transmitted in a predetermined time period. The predetermined time period may be, for example, the most recent one minute. The connection state calculation unit 114 may specify, for example, for each terminal device 4, the number of sessions in which packets 131 have been transmitted in a predetermined time period and in which the amount of data transmitted per unit time using each session is equal to or greater than a threshold value.

[0050] Furthermore, the connection status calculation unit 114 identifies the time during which packets 131 were continuously transmitted (hereinafter also referred to as continuous transmission time) for each session for each terminal device 4. Then, the connection status calculation unit 114 stores, as connection status information 134, in the information storage area 130, a time calculated from each of the identified continuous transmission times (for example, an average time of each of the identified continuous transmission times), for each terminal device 4, for example.

[0051] The connection state calculation unit 114 may specify, for example, the time during which the packets 131 are continuously transmitted in a predetermined time period as the continuous transmission time.

[0052] The data generation unit 115 generates a plurality of pieces of teacher data DT each including, for example, the quality information 132 acquired by the information acquisition unit 112, the communication state information 133 generated by the communication analysis unit 113, and the connection state information 134 calculated by the connection state calculation unit 114. Then, the data generation unit 115 stores the generated plurality of pieces of teacher data DT in the information storage area 130, for example.

[0053] Specifically, the data generation unit 115 generates multiple pieces of teacher data DT, for example, so that quality information 132, communication status information 133, and connection status information 134 that correspond to the same terminal device 4 and indicate each state at the same timing (same time period) are included in the same teacher data DT.

[0054] The model learning unit 116 generates a learning model MD, for example, by learning a plurality of pieces of teacher data DT generated by the data generation unit 115. Then, the model learning unit 116 stores the generated learning model MD in the information storage area 130, for example.

[0055] Next, the functions in the estimation process will be described.

[0056] The packet acquisition unit 111 extracts and acquires new packets 131 transmitted from the video distribution device 2, for example, from among new packets transmitted from the capture device 3. Then, the packet acquisition unit 111 stores the acquired new packets 131 in the information storage area 130, for example, for each terminal device 4 and for each session.

[0057] The communication analysis unit 113 generates new communication state information 133 from the acquisition status of the new packet 131 by the packet acquisition unit 111, for example, for each terminal device 4. Then, the communication analysis unit 113 stores the generated new communication state information 133 in the information storage area 130, for example, for each terminal device 4.

[0058] The connection state calculation unit 114 generates new connection state information 134 based on the acquisition status of the new packet 131 by the packet acquisition unit 111, for example, for each terminal device 4. Then, the connection state calculation unit 114 stores the generated new connection state information 134 in the information storage area 130, for example, for each terminal device 4.

[0059] The information estimation unit 117 inputs new communication state information 133 generated by the communication analysis unit 113 and new connection state information 134 calculated by the connection state calculation unit 114 to the learning model MD, for example, for each terminal device 4. Then, the information estimation unit 117 acquires new quality information 132 output from the learning model MD, for example, for each terminal device 4.

[0060] That is, the information estimation unit 117 inputs, for example, new communication state information 133 and new connection state information 134 that correspond to the same terminal device 4 and indicate the respective states at the same timing to the learning model MD.

[0061] The information using unit 118 controls each network device (not shown) included in the information processing system 10, for example, based on the new quality information 132 (the new quality information 132 for each terminal device 4) acquired by the information estimating unit 117.

[0062] Specifically, for example, when the quality information 132 acquired by the information estimation unit 117 does not satisfy a predetermined quality condition (hereinafter also simply referred to as the quality condition), the information utilization unit 118 may change various settings of each network device included in the information processing system 10 (for example, the wireless access network 6 shown in FIG. 2) so that the quality information 132 satisfies the quality condition.

[0063] In addition, the information utilization unit 118 outputs, for example, new quality information 132 (new quality information 132 for each terminal device 4) acquired by the information estimation unit 117 to an operation terminal (not shown) to be viewed by an administrator of the information processing device 1 (hereinafter simply referred to as the administrator).

[0064] [Outline of learning process in the first embodiment] Next, an outline of the first embodiment will be described with reference to Fig. 6, which is a flow chart for explaining an outline of the learning process in the first embodiment.

[0065] As shown in FIG. 6, the information processing device 1 calculates connection status information 134 indicating the connection status between the terminal device 4 and multiple video distribution devices 2, for example, based on the acquisition status of packets 131 transmitted from the video distribution device 2 to the terminal device 4 via the network NW (S1).

[0066] Then, the information processing device 1 generates a learning model MD by learning teacher data DT including, for example, connection status information 134 calculated in the processing of S1, communication status information 133 indicating the communication status of the packet 131 in the network NW, and quality information 132 indicating the quality related to the output of the packet 131 in the terminal device 4 (S2).

[0067] As a result, the information processing device 1 in this embodiment is able to generate a learning model MD capable of accurately estimating quality information 132 such as QoE even when, for example, there is not a sufficiently strong correlation between the timing at which communication state information 133 changes in the network NW and the timing at which a change in quality information 132 occurs in the terminal device 4.

[0068] [Details of the learning process in the first embodiment] Next, the learning process in the first embodiment will be described in detail. Fig. 7 to Fig. 13 are flow charts for explaining the learning process in the first embodiment in detail. Fig. 14 to Fig. 21 are diagrams for explaining the learning process in the first embodiment in detail.

[0069] [Packet acquisition process] First, a process of acquiring the packet 131 captured by the capture device 3 (hereinafter, also referred to as a packet acquisition process) will be described in the learning process. Fig. 7 is a flow chart for explaining the packet acquisition process.

[0070] As shown in FIG. 7, the packet acquisition unit 111 waits until it receives a packet transmitted from the capture device 3 (NO in S11).

[0071] Then, when a packet transmitted from the capture device 3 is received (YES in S11), the packet acquisition unit 111 determines, for example, whether the received packet is a packet 131 transmitted from the video distribution device 2 (S12).

[0072] Specifically, the packet acquisition unit 111 determines, for example, whether the combination of the source IP address contained in the IP header of the packet received in the processing of S11 and the source port number contained in the TCP header is a combination of an IP address and a port number corresponding to the video distribution device 2.

[0073] As a result, if it is determined that the packet received in the processing of S11 is a packet 131 sent from the video distribution device 2 (YES in S12), the packet acquisition unit 111 identifies, for example, the terminal device 4 that is the destination of the packet 131 received in the processing of S11 and the session in which the packet 131 received in the processing of S11 is sent and received (S13).

[0074] Specifically, the packet acquisition unit 111 identifies the terminal device 4 and the session corresponding to the combination of the destination IP address included in the IP header and the destination port number included in the TCP header of the packet 131 received in the process of S11, for example.

[0075] Then, the packet acquisition unit 111 stores, for example, the packet 131 received in the process of S11 in the information storage area 130 in association with the identification information of the terminal device 4 and the session identified in the process of S13 (S14).

[0076] On the other hand, if it is determined in the process of S11 that the received packet is not the packet 131 transmitted from the moving image distribution device 2 (NO in S12), the packet acquisition unit 111 does not perform the process of S13 and the process of S14, for example.

[0077] [Information acquisition process] Next, the process of acquiring the quality information 132 (hereinafter, also referred to as information acquisition process) of the learning process will be described. Fig. 8 is a flow chart for explaining the information acquisition process.

[0078] As shown in FIG. 8, the information acquisition unit 112 waits until it receives the quality information 132 transmitted from the terminal device 4 (NO in S21).

[0079] Specifically, the information acquiring unit 112 waits, for example, until the user inputs the quality information 132 via the terminal device 4. Note that the user inputs the quality information 132, for example, at predetermined time intervals (for example, every minute).

[0080] Then, when the quality information 132 transmitted from the terminal device 4 is received (YES in S21), the information acquisition unit 112 stores, for example, the received quality information 132 in the information storage area 130 (S22).

[0081] Specifically, the information acquiring unit 112 stores, for example, the quality information 132 received in the process of S21 in association with the identification information of the terminal device 4 to which the quality information 132 received in the process of S21 was transmitted in the information storage area 130. A specific example of the quality information 132 will be described below.

[0082] [Specific examples of quality information] Fig. 14 is a diagram for explaining a specific example of the quality information 132. Specifically, Fig. 14 is a diagram for explaining a specific example of the quality information 132 stored in the information storage area .

[0083] 14 has, for example, items such as "time" in which each time is set, "QoE" in which the QoE at each time is set, and "terminal ID" in which identification information of the terminal device 4 corresponding to each QoE is set. For example, "5", "4", "3", "2", or "1" is set in order of decreasing quality to "QoE".

[0084] Specifically, in the first line of quality information 132 shown in FIG. 14, for example, “12:01:00” is set as “Time”, “5” is set as “QoE”, and “T001” is set as “Terminal ID”.

[0085] That is, the information in the first row of the quality information 132 shown in FIG. 14 indicates that, for example, the QoE at 12:01:00 of the terminal device 4 having a terminal ID of T001 is 5.

[0086] In addition, in the second line of the quality information 132 shown in FIG. 14, for example, “12:02:00” is set as the “time”, “2” is set as the “QoE”, and “T001” is set as the “terminal ID”.

[0087] That is, the information in the second row of the quality information 132 shown in FIG. 14 indicates that, for example, the QoE at 12:02:00 of the terminal device 4 having the terminal ID T001 is 2.

[0088] In addition, in the information on the third line of the quality information 132 shown in FIG. 14, for example, “12:03:00” is set as the “time”, “1” is set as the “QoE”, and “T001” is set as the “terminal ID”.

[0089] That is, the information in the third row in the quality information 132 shown in Fig. 14 indicates that, for example, the QoE at 12:03:00 of the terminal device 4 having the terminal ID T001 is 2. Description of other information included in Fig. 14 will be omitted.

[0090] [First information generation process] Next, the process of generating communication state information 133 (hereinafter, also referred to as first information generating process) of the learning process will be described. Fig. 9 is a flow chart for explaining the first information generating process.

[0091] 9, the communication analysis unit 113 waits until the first information generation timing comes (NO in S31), for example. The first information generation timing may be a regular timing such as every minute.

[0092] Then, when it is time to generate the first information (YES in S31), the communication analysis unit 113 generates communication status information 133 (S32), for example, for each terminal device 4, by using each packet 131 stored in the information storage area 130 that was transmitted from the capture device 3 between the previous first information generation timing (e.g., one minute ago) and the current first information generation timing.

[0093] Thereafter, the communication analysis unit 113 stores the communication state information 133 generated in the process of S32 in the information storage area 130 for each terminal device 4 (S33). A specific example of the communication state information 133 will be described below.

[0094] [Examples of communication status information] Fig. 15 is a diagram for explaining a specific example of the communication state information 133. Specifically, Fig. 15 is a diagram for explaining a specific example of the communication state information 133 stored in the information storage area .

[0095] The communication status information 133 shown in FIG. 15 has, for example, items such as a "terminal ID" in which identification information of each terminal device 4 is set, a "time" in which each time is set, a "delay time" in which the delay time for transmitting and receiving packets 131 at each time is set, a "loss rate" in which the packet loss rate of packets 131 at each time is set, and a "traffic volume" in which the traffic volume at each time is set.

[0096] Specifically, in the information in the first line of the communication status information 133 shown in FIG. 15, for example, the "Terminal ID" is set to "T001," the "Time" is set to "12:01:00," the "Delay time" is set to "0 (ms)," the "Loss rate" is set to "0 (%)," and the "Traffic volume" is set to "5.1 (Mbps)."

[0097] In addition, in the information on the second line of the communication status information 133 shown in FIG. 15, for example, "Terminal ID" is set to "T001", "Time" is set to "12:02:00", "Delay time" is set to "1.0 (ms)", "Loss rate" is set to "0.1 (%)", and "Traffic volume" is set to "2.0 (Mbps)".

[0098] In addition, in the information on the third line of the communication status information 133 shown in Fig. 15, for example, "T001" is set as the "terminal ID", "12:03:00" is set as the "time", "2.0 (ms)" is set as the "delay time", "0.2 (%)" is set as the "loss rate", and "0.2 (Mbps)" is set as the "traffic volume". Explanation of other information included in Fig. 15 will be omitted.

[0099] [Second information generation process] Next, a description will be given of a process of generating at least a part of the connection state information 134 (hereinafter, also referred to as a second information generating process) in the learning process. Fig. 10 is a flow chart illustrating the second information generating process.

[0100] 10, the connection state calculation unit 114 waits until the second information generation timing comes (NO in S41). The second information generation timing may be a regular timing such as every minute. Specifically, the second information generation timing may be the same timing as the first information generation timing, for example.

[0101] Then, when the second information generation timing arrives (YES in S41), the connection status calculation unit 114 identifies, for example, for each terminal device 4 and for each session, from among the multiple video distribution devices 2, the first video distribution device 2 that first transmitted a packet 131 constituting the video data being played on the terminal device 4 to the terminal device 4 (S42).

[0102] Specifically, the connection state calculation unit 114 identifies, for example, for each terminal device 4 and for each session, the first packet 131 (hereinafter also referred to as the initial packet 131) constituting the video data being played back on the terminal device 4, from among the packets 131 stored in the information storage area 130. Then, the connection state calculation unit 114 identifies, for example, for each terminal device 4 and for each session, the video distribution device 2 that is the source of the identified initial packet 131, as the first video distribution device 2.

[0103] In addition, the connection status calculation unit 114 identifies, for example, for each terminal device 4 and for each session, from among the multiple video distribution devices 2, the second video distribution device 2 that last transmitted a packet 131 constituting the video data being played to the terminal device 4 (S43).

[0104] Specifically, the connection state calculation unit 114 identifies, for example, for each terminal device 4 and for each session, the last packet 131 (hereinafter also referred to as the final packet 131) constituting the video data being played back on the terminal device 4, from among the packets 131 stored in the information storage area 130. Then, the connection state calculation unit 114 identifies, for example, for each terminal device 4 and for each session, the video distribution device 2 that is the source of the identified final packet 131, as the second video distribution device 2.

[0105] Thereafter, the connection status calculation unit 114 stores, for example, for each terminal device 4, the similarity (hereinafter simply referred to as similarity) between the address information of the first video distribution device 2 identified in the processing of S42 and the address information of the second video distribution device 2 identified in the processing of S43 in the information storage area 130 as at least a part of the connection status information 134 (S44).

[0106] Specifically, the connection status calculation unit 114 calculates the similarity between the IP address of the first video distribution device 2 identified in the processing of S42 and the IP address of the second video distribution device 2 identified in the processing of S43, for example, for each terminal device 4 and for each session. Then, the connection status calculation unit 114 stores a value calculated from each of the calculated similarities (for example, the minimum value of each of the calculated similarities) in the information storage area 130 as at least a part of the connection status information 134, for example, for each terminal device 4. A specific example of the processing of S44 will be described below.

[0107] [Specific example of S44 processing] In the process of S44, the connection state calculation unit 114, for example, vectorizes each of the IP address of the first moving image distribution device 2 and the IP address of the second moving image distribution device 2.

[0108] Specifically, for example, if the IP address of the first video distribution device 2 is "142.251.84.230", the connection status calculation unit 114 converts "142" to "10001110", "251" to "11111011", "84" to "01010100", and "230" to "11100110", and then combines each of the binary converted values ​​to convert "142.251.84.230" to "1000111011110110101010011100110".

[0109] Then, the connection state calculation unit 114 calculates, for example, the cosine similarity or the Pearson correlation coefficient between the vectorized IP address for the first video distribution device 2 and the vectorized IP address for the second video distribution device 2.

[0110] That is, as shown in Fig. 16, when video data is being distributed from video distribution device 2c to terminal device 4, if traffic volume 133a of packets 131 constituting the video data drops and playback of the video data is interrupted, terminal device 4 attempts to resume the video data by establishing a new session with, for example, another video distribution device 2 capable of distributing the video data (hereinafter also simply referred to as another video distribution device 2) as shown in Fig. 16. Specifically, in this case, terminal device 4 attempts to receive packets 131 by establishing a new session with, for example, each of the other video distribution devices 2 including video distribution device 2a (video distribution device 2 located farther from terminal device 4 than video distribution device 2c and video distribution device 2d), as shown in Fig. 16.

[0111] Here, it is possible to determine that the similarity of the address information between the video distribution device 2c and the video distribution device 2a is smaller than the similarity of the address information between the video distribution device 2c and another video distribution device 2 (e.g., video distribution device 2d) that is located closer to the terminal device 4 than the video distribution device 2a. It is also possible to determine that the behavior of the terminal device 4 establishing new sessions one after another with each of the other video distribution devices 2 including the video distribution device 2a does not occur, for example, when the playback of the video data is not interrupted. In other words, it is possible to determine that the possibility that the terminal device 4 establishes a new session with the video distribution device 2a when the playback of the video data is not interrupted is extremely lower than the possibility that the terminal device 4 establishes a new session with the video distribution device 2a when the playback of the video data is interrupted.

[0112] Therefore, it can be determined that the similarity when the playback of the video data in the terminal device 4 is interrupted is likely to be smaller than the similarity when the playback of the video data in the terminal device 4 is not interrupted, for example.

[0113] Therefore, in the process of S44, the information processing device 1 in this embodiment calculates, for example, a similarity between the address information of the first video distribution device 2 identified in the process of S42 and the address information of the second video distribution device 2 identified in the process of S43. Then, as will be described later, the information processing device 1 uses, for example, the calculated similarity as a part of the teacher data DT.

[0114] This allows the information processing device 1 in this embodiment to generate a learning model MD that learns behavior (e.g., behavior of traffic volume 133a, etc.) when playback of video data is interrupted in the terminal device 4. Therefore, the information processing device 1 can generate a learning model MD that can predict a decrease in quality information 132 such as QoE due to interruption of playback of video data, for example.

[0115] In addition, if the playback of the video data is not resumed despite starting the reception of the packet 131 from the video distribution device 2a, the terminal device 4 may, for example, establish a new session with the video distribution device 2b (a video distribution device 2 capable of distributing low-quality video data and located far from the terminal device 4) as shown in FIG. 16, and attempt to receive the packet 131 from the video distribution device 2b. Furthermore, if the playback of the video data is not resumed despite starting the reception of the packet 131 from the video distribution device 2b, the terminal device 4 may, for example, further establish a new session with the video distribution device 2d as shown in FIG. 16, and attempt to receive the packet 131 from the video distribution device 2d. Thereafter, if the decreased throughput 131a increases (recovers), the terminal device 4 may, for example, re-establish a session with the video distribution device 2c as shown in FIG. 16, and resume the reception of the packet 131 from the video distribution device 2c.

[0116] [Example of connection status information (1)] Next, a specific example of the connection state information 134 will be described. Figs. 17, 18 and 20 are diagrams for explaining a specific example of the connection state information 134. Specifically, Fig. 17 is a specific example of the connection state information 134 (hereinafter also referred to as connection state information 134a) including the similarity between the address information of the first video distribution device 2 and the address information of the second video distribution device 2. Note that Fig. 17 is a specific example of the connection state information 134a for one terminal device 4.

[0117] In the example shown in FIG. 17, the connection status information 134a has, for example, items of "time zone," in which each time zone is set, and "similarity," in which the similarity between the address information of the first video distribution device 2 and the address information of the second video distribution device 2 during each time zone is set.

[0118] Specifically, in the information on the first line of the connection state information 134a shown in FIG. 17, for example, "12:00:00-12:01:00" is set as the "time period" and "1" is set as the "similarity level."

[0119] In addition, in the information on the second line of the connection status information 134a shown in Fig. 17, for example, "12:01:00-12:02:00" is set as the "time period" and "0.8" is set as the "similarity." Explanation of other information included in Fig. 17 will be omitted.

[0120] [Third information generation process] Next, a process of generating at least a part of the connection state information 134 (hereinafter, also referred to as a third information generating process) will be described as part of the learning process. Fig. 11 is a flow chart illustrating the third information generating process.

[0121] 11, the connection state calculation unit 114 waits until the third information generation timing comes (NO in S51). The third information generation timing may be a regular timing such as every minute. Specifically, the third information generation timing may be the same timing as the first information generation timing or the second information generation timing, for example.

[0122] Then, when the timing for generating the third information arrives (YES in S51), the connection status calculation unit 114 stores, for example, for each terminal device 4, the number of sessions (hereinafter simply referred to as the number of sessions) in which packets 131 were transmitted from any of the multiple video distribution devices 2 to each terminal device 4 as at least a part of the connection status information 134 in the information storage area 130 (S52).

[0123] Specifically, in this case, the connection state calculation unit 114 identifies, for example, for each terminal device 4, the number of sessions in which packets 131 were transmitted to each terminal device 4 during a predetermined time period (for example, the most recent one minute).

[0124] 16, when the playback of video data is interrupted, the terminal device 4 establishes, for example, a new session with another video distribution device 2. Therefore, it is possible to determine that the number of sessions when the playback of video data is interrupted (the number of sessions in a given time period) is greater than the number of sessions when the playback of video data is not interrupted, for example.

[0125] Therefore, in the process of S52, the information processing device 1 in this embodiment calculates, for example, the number of sessions in which packets 131 were transmitted from any of the multiple video distribution devices 2 to each terminal device 4. Then, as will be described later, the information processing device 1 uses, for example, the calculated number of sessions as a part of the teacher data DT.

[0126] As a result, the information processing device 1 in this embodiment can generate a learning model MD that has learned behavior when playback of video data is interrupted in the terminal device 4, for example, in the same way as in the second information generation process. Therefore, the information processing device 1 can generate a learning model MD that can predict a decrease in quality information 132 such as QoE due to an interruption in playback of video data, for example.

[0127] In the process of S52, the connection state calculation unit 114 may exclude (not count) sessions in which the amount of data transmitted per unit time using each session is equal to or less than a threshold. That is, the connection state calculation unit 114 may exclude sessions in which data (e.g., various control data, etc.) used for purposes other than the distribution of video data is transmitted.

[0128] [Example of connection status information (2)] Next, a specific example of the connection state information 134 will be described. Fig. 18 is a specific example of the connection state information 134 including the number of sessions (hereinafter, also referred to as connection state information 134b). Fig. 18 is a specific example of the connection state information 134b for one terminal device 4.

[0129] In the example shown in FIG. 18, the connection status information 134b has, for example, items such as a “time period” in which each time period is set, and a “number of sessions” in which the number of sessions in which packets 131 were transmitted from any of the multiple video distribution devices 2 to each terminal device 4 during each time period is set.

[0130] Specifically, in the information on the first line of the connection status information 134b shown in FIG. 18, for example, "12:00:00-12:01:00" is set as the "time period" and "1" is set as the "number of sessions."

[0131] In addition, in the information on the second line of the connection state information 134b shown in Fig. 18, for example, "12:01:00-12:02:00" is set as the "time" and "2" is set as the "number of sessions." Descriptions of other information included in Fig. 18 will be omitted.

[0132] [Fourth information generation process] Next, a process of generating at least a part of the connection state information 134 (hereinafter, also referred to as a fourth information generating process) will be described as part of the learning process. Fig. 12 is a flow chart illustrating the fourth information generating process.

[0133] 12, the connection state calculation unit 114 waits until the fourth information generation timing comes (NO in S61). The fourth information generation timing may be a regular timing such as every minute. Specifically, the fourth information generation timing may be the same timing as the first information generation timing, the second information generation timing, and the third information generation timing, for example.

[0134] Then, when the fourth information generation timing arrives (YES in S61), the connection status calculation unit 114 stores, for example, for each terminal device 4, the continuous transmission time during which packets 131 were continuously transmitted in the information storage area 130 as at least a part of the connection status information 134 (S62).

[0135] Specifically, in this case, the connection state calculation unit 114 identifies a continuous transmission time during which packets 131 were continuously transmitted in a predetermined time period (e.g., the most recent one minute) for each terminal device 4 and for each session. Then, the connection state calculation unit 114 stores a time calculated from each of the identified continuous transmission times (e.g., an average time of each of the identified continuous transmission times) for each terminal device 4 in the information storage area 130 as at least a part of the connection state information 134. A specific example of the process of S62 will be described below.

[0136] [Example of S62 processing] Fig. 19 is a diagram illustrating a specific example of the session information 135. The session information 135 is, for example, information indicating, for each session and for each predetermined time period, a continuous transmission time during which packets 131 are continuously transmitted in each session. Note that Fig. 19 is a specific example of the session information 135 for one terminal device 4. Also, the session information 135 may be, for example, information generated (updated) by the connection state calculation unit 114 in the process of S62.

[0137] Specifically, in the first line of session information 135 shown in FIG. 19 (information corresponding to a session with a "Session ID" of "S001"), for example, "60 (seconds)" is set for "12:00:00-12:01:00" and "30 (seconds)" is set for "12:01:00-12:02:00".

[0138] That is, the information in the first line of session information 135 shown in FIG. 19 indicates, for example, that in a session having a "Session ID" of "S001", the continuous transmission time in the one minute from 12:00:00 to 12:01:00 is 60 seconds, and the continuous transmission time in the one minute from 12:01:00 to 12:02:00 is 30 seconds.

[0139] In addition, in the information in the second line of session information 135 shown in FIG. 19 (information corresponding to a session whose "Session ID" is "S002"), for example, "30 (seconds)" is set for "12:01:00-12:02:00" and "10 (seconds)" is set for "12:02:00-12:03:00".

[0140] That is, the information in the second line of session information 135 shown in FIG. 19 indicates, for example, that in a session having a "Session ID" of "S002", the continuous transmission time in the one minute from 12:01:00 to 12:02:00 was 30 seconds, and the continuous transmission time in the one minute from 12:02:00 to 12:03:00 was 10 seconds.

[0141] In addition, in the information in the third line of the session information 135 shown in FIG. 19 (information corresponding to a session with a "session ID" of "S003"), for example, "12:02:00-12:03:00" is set to "10 (seconds)".

[0142] That is, the information on the third line of the session information 135 shown in FIG. 19 indicates that, for example, in a session with a "Session ID" of "S003", the continuous transmission time in the one minute period from 12:02:00 to 12:03:00 is 30 seconds.

[0143] In addition, in the information on the fourth line of session information 135 shown in FIG. 19 (information corresponding to a session with a "Session ID" of "S004"), for example, "20 (seconds)" is set for "12:02:00-12:03:00", "60 (seconds)" is set for "12:03:00-12:04:00", "60 (seconds)" is set for "12:04:00-12:05:00", and "5 (seconds)" is set for "12:05:00-12:06:00".

[0144] That is, the information on the fourth line of session information 135 shown in FIG. 19 indicates, for example, that in a session having a "Session ID" of "S004", the continuous transmission time for one minute from 12:02:00 to 12:03:00 is 20 seconds, the continuous transmission time for one minute from 12:03:00 to 12:04:00 is 60 seconds, the continuous transmission time for one minute from 12:04:00 to 12:05:00 is 60 seconds, and the continuous transmission time for one minute from 12:05:00 to 12:06:00 is 5 seconds.

[0145] 19, the information corresponding to "12:00:00-12:01:00" is, for example, "60 (seconds)", which is the information corresponding to "S001". Therefore, in the process of S62, the connection state calculation unit 114 calculates 60 seconds as the continuous transmission time for one minute from 12:00:00 to 12:01:00.

[0146] 19, the information corresponding to "12:01:00-12:02:00" is, for example, "30 (seconds)" which is the information corresponding to "S001" and "30 (seconds)" which is the information corresponding to "S002." Therefore, in the process of S62, the connection state calculation unit 114 calculates 30 seconds, which is the average of 30 seconds, as the continuous transmission time for one minute from 12:00:00 to 12:01:00.

[0147] 19, the information corresponding to "12:02:00-12:03:00" is, for example, "10 (seconds)" which is the information corresponding to "S002", "30 (seconds)" which is the information corresponding to "S003", and "20 (seconds)" which is the information corresponding to "S004". Therefore, in the process of S62, the connection state calculation unit 114 calculates 20 seconds, which is the average of 10 seconds, 30 seconds, and 20 seconds, as the continuous transmission time for one minute from 12:02:00 to 12:03:00. Description of other information included in FIG. 19 will be omitted.

[0148] 16, when the playback of the video data is interrupted, the terminal device 4 establishes, for example, a new session with another video distribution device 2. Therefore, it is possible to determine that the continuous transmission time when the playback of the video data is interrupted (the continuous transmission time in a predetermined time period) is shorter than the continuous transmission time when the playback of the video data is not interrupted, for example.

[0149] Therefore, in the process of S62, the information processing device 1 in this embodiment calculates, for example, a continuous transmission time of the packet 131. Then, as will be described later, the information processing device 1 uses, for example, the calculated continuous transmission time as a part of the teacher data DT.

[0150] As a result, similarly to the second information generation process and the third information generation process, the information processing device 1 in this embodiment can generate a learning model MD that has learned behavior when playback of video data is interrupted in the terminal device 4. Therefore, by using the learning model MD, for example, the information processing device 1 can predict a decrease in quality information 132 such as QoE that accompanies interruption of playback of video data in the terminal device 4.

[0151] [Example of connection status information (3)] Next, a description will be given of a specific example of the connection state information 134. Fig. 20 is a specific example of the connection state information 134 (hereinafter, also referred to as connection state information 134c) including the continuous transmission time of the packets 131. Note that Fig. 20 is a specific example of the connection state information 134c for one terminal device 4.

[0152] In the example shown in FIG. 20, the connection state information 134c has, as items, for example, a "time" in which each time period is set, and a "continuous transmission time" in which the continuous transmission time in each time period is set.

[0153] Specifically, in the information on the first line of the connection status information 134c shown in FIG. 20, for example, "12:00:00-12:01:00" is set as the "time period" and "60 (seconds)" is set as the "continuous transmission time."

[0154] In addition, in the information on the second line of the connection status information 134c shown in Fig. 20, for example, "12:01:00-12:02:00" is set as the "time period" and "30 (seconds)" is set as the "continuous transmission time." Explanation of other information included in Fig. 20 will be omitted.

[0155] [Main process of learning process] Next, the main process of the learning process (hereinafter, simply referred to as the main process) will be described with reference to FIG 13.

[0156] 13, the data generation unit 115 waits until the learning timing (NO in S71), for example. The learning timing may be the timing when the administrator inputs information to the information processing device 1 to generate the learning model MD, for example.

[0157] Then, when it is time to learn (YES in S71), the data generation unit 115 generates a plurality of pieces of teacher data DT each including, for example, the quality information 132, the communication state information 133, and the connection state information 134 stored in the information storage area 130 (S72). After that, the data generation unit 115 stores, for example, the generated plurality of pieces of teacher data DT in the information storage area 130.

[0158] Specifically, the data generation unit 115 generates a plurality of pieces of teacher data DT so that, for example, quality information 132, communication state information 133, and connection state information 134 corresponding to the same terminal device 4 and generated at the same timing are included in the same teacher data DT. Specific examples of the plurality of pieces of teacher data DT will be described below.

[0159] [Specific examples of training data] Fig. 21 is a diagram for explaining a specific example of a plurality of pieces of teacher data DT. Specifically, Fig. 21 is a diagram for explaining a specific example of a plurality of pieces of teacher data DT stored in the information storage area 130.

[0160] 21 has, for example, items such as "delay time" in which a delay time for transmitting and receiving each packet 131 is set, "loss rate" in which a loss rate for each packet 131 is set, "traffic volume" in which a traffic volume for each packet 131 is set, and "QoE" in which a QoE in each terminal device 4 is set. Also, each teacher data DT shown in Fig. 21 has, for example, items such as "similarity" in which the similarity calculated in the second information generation process is set, "number of sessions" in which the number of sessions calculated in the third information generation process is set, and "continuous transmission time" in which the continuous transmission time calculated in the fourth information generation process is set.

[0161] Specifically, in the information in the first line of the teacher data DT shown in Fig. 21, for example, "0 (ms)" is set as the "delay time", "0 (%)" is set as the "loss rate", "5.1 (Mbps)" is set as the "traffic volume", and "5" is set as the "QoE". Also, in the information in the first line of the teacher data DT shown in Fig. 21, for example, "1" is set as the "similarity", "1" is set as the "number of sessions", and "60 (seconds)" is set as the "continuous transmission time".

[0162] In addition, in the information in the second line of the teacher data DT shown in Fig. 21, for example, "1.0 (ms)" is set as the "delay time", "0.1 (%)" is set as the "loss rate", "2.0 (Mbps)" is set as the "traffic volume", and "2" is set as the "QoE". In addition, in the information in the second line of the teacher data DT shown in Fig. 21, for example, "0.8" is set as the "similarity", "2" is set as the "number of sessions", and "30 (seconds)" is set as the "continuous transmission time". Description of other data included in Fig. 21 will be omitted.

[0163] Each of the multiple pieces of teacher data DT may include, for example, a part of the connection state information 134a, the connection state information 134b, and the connection state information 134c.

[0164] 13, the model learning unit 116 generates a learning model MD by learning the multiple pieces of teacher data DT generated in the process of S72 (S73). After that, the model learning unit 116 stores the generated learning model MD in the information storage area 130, for example.

[0165] [Estimation process in the first embodiment] Next, the estimation process in the first embodiment will be described. Fig. 22 is a flow chart illustrating the estimation process in the first embodiment. Fig. 23 is a diagram illustrating the estimation process in the first embodiment.

[0166] 22, the information estimation unit 117 waits until the estimation timing (NO in S81), for example. The estimation timing may be the timing when the administrator inputs information to the information processing device 1 to estimate the quality information 132, for example.

[0167] Then, when the estimation timing arrives (YES in S81), the information estimation unit 117 inputs, for example, the communication state information 133 stored in the information storage area 130 and the connection state information 134 stored in the information storage area 130 into the learning model MD (S82).

[0168] Specifically, the information estimation unit 117 acquires, for example, communication state information 133 stored in the information storage area 130, that corresponds to the timing at which the quality information 132 is estimated. Also, the information estimation unit 117 acquires, for example, connection state information 134 stored in the information storage area 130, that corresponds to the timing at which the quality information 132 is estimated. Then, the information estimation unit 117 inputs, for example, the acquired communication state information 133 and connection state information 134 to the learning model MD.

[0169] More specifically, the information estimation unit 117 acquires, for example, the most recent communication state information 133 from among the communication state information 133 stored in the information storage area 130. Also, the information estimation unit 117 acquires, for example, the most recent connection state information 134 from among the connection state information 134 stored in the information storage area 130. Then, the information estimation unit 117 inputs, for example, the acquired communication state information 133 and connection state information 134 to the learning model MD.

[0170] After that, the information estimation unit 117 acquires, for example, the quality information 132 output from the learning model MD (S83).

[0171] Then, the information using unit 118 outputs the quality information 132 acquired in the process of S83 to an operation terminal (not shown) (S84).

[0172] In this case, the information using unit 118 controls each network device (not shown) included in the information processing system 10, for example, based on the quality information 132 acquired in the process of S83.

[0173] In this way, in the learning process, the information processing device 1 in this embodiment calculates first connection state information 134 indicating the connection state between the terminal device 4 and the multiple video distribution devices 2, for example, based on the acquisition status of the packets 131 transmitted from the video distribution device 2 to the terminal device 4 via the network NW. Then, the information processing device 1 generates a learning model MD by learning teacher data DT including, for example, the calculated first connection state information 134, first communication state information 133 indicating the communication state of the packets 131 in the network NW, and first quality information 132 indicating the quality related to the output of the packets 131 in the terminal device 4.

[0174] Furthermore, in the estimation process, for example, similarly to the learning process, the information processing device 1 calculates second connection state information 134 based on the acquisition status of packets 131 transmitted from the video distribution device 2 to the terminal device 4 via the network NW. Then, the information processing device 1 acquires second quality information 132 output from the learning model MD in response to input of the calculated second connection state information 134 and second communication state information 133 indicating the communication state of the packets 131 in the network NW. Then, the information processing device 1 outputs the acquired second quality information 132, for example.

[0175] As a result, the information processing device 1 in this embodiment can generate a learning model that can accurately estimate quality information 132 such as QoE even if the correlation between the timing at which the communication state information 133 changes in the network NW and the timing at which a change in the quality information 132 occurs in the terminal device 4 is not sufficiently strong.

[0176] That is, for example, when the playback of video data is interrupted, terminal device 4 may attempt to resume the playback of the video data by adjusting (hereinafter, simply referred to as image quality adjustment) the image quality of the video data distributed by video distribution device 2. Therefore, in terminal device 4, for example, even during a time period when traffic volume 133a is low, the playback may not be interrupted by adjusting the image quality of the video data.

[0177] However, for example, when a part of the packet 131 transmitted from the video distribution device 2 (for example, data included in a portion other than the header) is encrypted, the information processing device 1 cannot analyze the packet 131 transmitted from the video distribution device 2, and cannot determine the image quality of the video data being reproduced on the terminal device 4. Therefore, for example, even if the traffic volume 133a is decreasing, the information processing device 1 cannot determine whether the reproduction of the video is interrupted on the terminal device 4.

[0178] Therefore, the information processing device 1 in this embodiment makes the learning model MD learn the behavior when image quality is adjusted in the terminal device 4, for example, by making the learning model MD learn the teacher data DT including the connection state information 134.

[0179] This allows the information processing device 1 in the embodiment to generate a learning model MD capable of outputting quality information 132 such as QoE reflecting a prediction result as to whether or not image quality adjustment has been performed. Specifically, as shown in Fig. 25, for example, even when the traffic volume 133a is decreased, the information processing device 1 can generate a learning model MD capable of outputting quality information 132 such as QoE reflecting a prediction result as to whether or not image quality adjustment has been performed.

[0180] In the above example, the learning model MD is generated by using the teacher data DT including the quality information 132, the communication state information 133, and the connection state information 134 at the same timing, but the present invention is not limited to this. Specifically, the information processing device 1 may generate the learning model MD by using the teacher data DT including, for example, the communication state information 133 and the connection state information 134 at the same timing (hereinafter also referred to as the first timing) and the quality information 132 at the second timing later than the first timing.

[0181] This enables the information processing device 1 to, in the estimation process, estimate quality information 132 (i.e., future quality information 132) at a timing later than the timing corresponding to the communication state information 133 and connection state information 134 input to the learning model MD.

[0182] In addition, in the above example, a case has been described in which the information processing device 1 (packet acquisition unit 111) acquires the packet 131 transmitted from the capture device 3, and the information processing device 1 (communication analysis unit 113) generates communication status information 133 from the packet 131, but this is not limited to this.

[0183] Specifically, the communication state information 133 may be generated by, for example, the capture device 3 that captures the packet 131. Then, the information processing device 1 (information acquisition unit 112) may receive the communication state information 133 transmitted from the capture device 3, and store the received communication state information 133 in the information storage area 130, for example.

[0184] This eliminates the need for the information processing device 1 to have a storage area (information storage area 130) for storing the packet 131 transmitted from the capture device 3, for example, and makes it possible to reduce operation costs.

[0185] The above embodiment can be summarized as follows:

[0186] (Appendix 1) a connection status calculation unit that calculates first connection status information indicating a connection status between the terminal device and the plurality of application devices based on a capture status of a packet transmitted from at least one of the plurality of application devices to the terminal device via a network in a capture device; and a model learning unit that generates a learning model by learning teacher data including the calculated first connection state information, first communication state information indicating a communication state of the packet in the network, and first quality information indicating a quality related to an output of the packet in the terminal device. A network device comprising:

[0187] (Appendix 2) In Appendix 1: the packets constitute video data that is distributed from at least one of the plurality of application devices to the terminal device and played back on the terminal device; The first quality information is information indicating a Quality of Experience (QoE) for the video data in the terminal device. A network device comprising:

[0188] (Appendix 3) Appendix 1 further states: a packet acquisition unit that acquires the packet transmitted from the capture device; the connection state calculation unit calculates the first connection state information from an acquisition status of the packet in the packet acquisition unit; A network device comprising:

[0189] (Appendix 4) Appendix 3 further states: a communication analysis unit that calculates the first communication state information from an acquisition status of the packet in the packet acquisition unit; A network device comprising:

[0190] (Appendix 5) In Appendix 1: The capture device generates the first communication state information from an acquisition status of the packet at the capture device, and further an information acquisition unit that acquires the first communication state information transmitted from the capture device; A network device comprising:

[0191] (Appendix 6) In Appendix 1: the packets constitute video data that is distributed from at least one of the plurality of application devices to the terminal device and played back on the terminal device; The connection state calculation unit identifying, from the plurality of application devices, a first application device that first transmitted the packets constituting the video data to the terminal device, and a second application device that transmitted the packets constituting the video data to the terminal device at a predetermined timing; generating a similarity between the identified address information of the first application device and the address information of the second application device as at least one piece of the first connection state information; A network device comprising:

[0192] (Appendix 7) In Appendix 1: the predetermined timing is a timing when the packets constituting the video data were last transmitted from the second application device to the terminal device. A network device comprising:

[0193] (Appendix 8) In Appendix 1: the connection state calculation unit generates, as at least one piece of the first connection state information, a number of sessions in which the packets have been transmitted from any one of the plurality of application devices to the terminal device; A network device comprising:

[0194] (Appendix 9) In Appendix 8: the connection state calculation unit generates, as at least one piece of the first connection state information, a number of sessions among the sessions, each of which has an amount of data transmitted per unit time equal to or greater than a threshold. A network device comprising:

[0195] (Appendix 10) In Appendix 1: the connection state calculation unit generates, as at least one piece of the first connection state information, a first time period during which the packets were continuously transmitted in a session in which the packets were transmitted from any one of the plurality of application devices to the terminal device; A network device comprising:

[0196] (Appendix 11) In Appendix 10: when a plurality of the sessions exist, the connection state calculation unit generates, as at least one of the first connection state information, a second time calculated from the first time corresponding to each of the plurality of the sessions; A network device comprising:

[0197] (Appendix 12) In Appendix 1: the connection state calculation unit calculates second connection state information indicating a connection state between the terminal device and the plurality of application devices based on the acquisition status of another packet transmitted from the application device to the terminal device via the network, and further an information estimation unit that acquires second quality information output from the learning model in response to input of the calculated second connection state information and second communication state information indicating a communication state of the other packets in the network; and an information output unit that outputs the acquired second quality information. A network device comprising:

[0198] (Appendix 13) calculating first connection state information indicating a connection state between the terminal device and the plurality of application devices based on an acquisition status in a capture device of packets transmitted from the plurality of application devices to the terminal device via a network; generating a learning model by learning teacher data including the calculated first connection state information, first communication state information indicating a communication state of the packet in the network, and first quality information indicating a quality related to an output of the packet in the terminal device; A model learning method comprising:

[0199] (Appendix 14) Appendix 13 further states: calculating second connection state information indicating a connection state between the terminal device and the plurality of application devices based on the acquisition status of other packets transmitted from the plurality of application devices to the terminal device via the network; Acquire second quality information output from the learning model in response to input of the calculated second connection state information and second communication state information indicating a communication state of the other packets in the network; outputting the acquired second quality information; A model learning method comprising: [Explanation of symbols]

[0200] 1: Information processing device 2: Video distribution device 3: Capture device 4: Terminal device 10: Information Processing System 101: CPU 102: Memory 103: I / O interface 104: Storage device 105: Bus 130: Information storage area IN: Internet NW: Network

Claims

1. a connection status calculation unit that calculates first connection status information indicating a connection status between the terminal device and the plurality of application devices based on a capture status of a packet transmitted from at least one of the plurality of application devices to the terminal device via a network in a capture device; a model learning unit that generates a learning model by learning teacher data including the calculated first connection state information, first communication state information indicating a communication state of the packet in the network, and first quality information indicating a quality related to an output of the packet in the terminal device, A network device comprising:

2. In claim 1, the packets constitute video data that is distributed from at least one of the plurality of application devices to the terminal device and played back on the terminal device; The first quality information is information indicating a Quality of Experience (QoE) for the video data in the terminal device. A network device comprising:

3. In claim 1, the packets constitute video data that is distributed from at least one of the plurality of application devices to the terminal device and played back on the terminal device; The connection state calculation unit identifying, from the plurality of application devices, a first application device that first transmitted the packets constituting the video data to the terminal device, and a second application device that transmitted the packets constituting the video data to the terminal device at a predetermined timing; generating a similarity between the identified address information of the first application device and the identified address information of the second application device as at least one piece of the first connection state information; A network device comprising:

4. In claim 1, the connection state calculation unit generates, as at least one piece of the first connection state information, a number of sessions in which the packets have been transmitted from any one of the plurality of application apparatuses to the terminal device; A network device comprising:

5. In claim 4, the connection state calculation unit generates, as at least one piece of the first connection state information, a number of sessions among the sessions, each of which has an amount of data transmitted per unit time equal to or greater than a threshold; A network device comprising:

6. In claim 1, the connection state calculation unit generates, as at least one piece of the first connection state information, a first time period during which the packets are continuously transmitted in a session in which the packets are transmitted from any one of the plurality of application devices to the terminal device; A network device comprising:

7. In claim 6, the connection state calculation unit generates, when a plurality of the sessions exist, second times calculated from the first times corresponding to each of the plurality of the sessions as at least one of the first connection state information; A network device comprising:

8. In claim 1, the connection state calculation unit calculates second connection state information indicating a connection state between the terminal device and the plurality of application devices based on the acquisition status of another packet transmitted from the application device to the terminal device via the network, and further an information estimation unit that acquires second quality information output from the learning model in response to input of the calculated second connection state information and second communication state information indicating a communication state of the other packets in the network; and an information output unit that outputs the acquired second quality information. A network device comprising:

9. calculating first connection status information indicating a connection status between the terminal device and the plurality of application devices based on an acquisition status in a capture device of packets transmitted from the plurality of application devices to the terminal device via a network; generating a learning model by learning teacher data including the calculated first connection state information, first communication state information indicating a communication state of the packet in the network, and first quality information indicating a quality related to an output of the packet in the terminal device; A model learning method comprising:

10. In claim 9, further comprising: calculating second connection state information indicating a connection state between the terminal device and the plurality of application devices based on the acquisition status of other packets transmitted from the plurality of application devices to the terminal device via the network; acquiring second quality information output from the learning model in response to input of the calculated second connection state information and second communication state information indicating a communication state of the other packets in the network; outputting the acquired second quality information; A model learning method comprising:

Images