Parameter setting device and parameter setting method

Abstract

A parameter setting device for setting a parameter representing a weight between a first model constructed by a hierarchical Bayesian modeling method using observation data of a target cell and a second model constructed by the hierarchical Bayesian modeling method also using observation data of cells other than the target cell, the parameter setting device comprising: a division unit that divides traffic demand, which is included in observation data of cells including the target cell, into a plurality of groups; a first calculation unit that calculates a first index value indicating the significance level of the first model on the basis of information about the plurality of groups and the observation data; a second calculation unit that calculates a second index value indicating the significance level of the second model on the basis of the information about the plurality of groups and the observation data; a parameter calculation unit that calculates the parameter on the basis of the first index value and the second index value; and a setting unit that sets the parameter for a base station of the target cell.

Description

Parameter setting device and parameter setting method

[0001] The present disclosure relates to a parameter setting device and a parameter setting method.

[0002] To provide comfortable communication services to cellular network users, mobile communication operators aim to improve the quality of cellular networks, such as throughput, Radio Resource Control (RRC) connection completion rate, and handover (HO) success rate. Many operators improve quality by expanding base station cells and other facilities, but expanding facilities is expensive and difficult to achieve quality improvements in all areas. Therefore, in addition to expanding facilities, operators are also working to improve quality by adjusting equipment parameters. Operators at each operator predict the relationship between equipment parameters and each quality level based on their own know-how and adjust the settings of the equipment parameters to values ​​that can improve each quality level. However, because each quality level is determined by various factors related to the cellular network, it is difficult to grasp the relationship between equipment parameters and each quality level. For example, even if the traffic demand and the settings of the equipment parameters are the same, the quality level can vary depending on the cell's attributes (e.g., frequency band, bandwidth, etc.) and the cell's surrounding environment (e.g., buildings, terrain, etc.). Therefore, in order to appropriately adjust the equipment parameters, it is important to understand the relationship between the equipment parameters and the quality level for each cell.

[0003] Non-Patent Document 1 proposes a method of using hierarchical Bayesian modeling to utilize observation data from each cell and build a model for each cell (a model that estimates the quality of the corresponding cell from the observation data). By using hierarchical Bayesian modeling, a model for each cell can be built while utilizing observation data from other cells, making it possible to build a model that suppresses overfitting even for cells with a small number of observations. On the other hand, because observation data from other cells is utilized even for cells with a large number of observations and that do not require observation data from other cells, there is a problem in that model accuracy decreases for cells with a sufficient number of observations.

[0004] Non-Patent Document 2 proposes hierarchical Bayesian modeling that takes the number of observations into account in order to solve the problems of the method proposed in Non-Patent Document 1. The method proposed in Non-Patent Document 2 provides a parameter that determines whether to emphasize a model (common model) constructed using observation data from other cells or a model (single model) constructed using only the observation data from the cell itself, depending on the number of observations for each cell. As a result, for cells with a small number of observations, emphasizing the common model makes it possible to suppress a decrease in model accuracy due to overfitting. Furthermore, for cells with a large number of observations, emphasizing the single model makes it possible to suppress a decrease in model accuracy due to the influence of other cells and construct a model that fully reflects the characteristics of the cell itself.

[0005] Deniz Ustebay and Jie Chuai. "Hierarchical Bayesian Modeling for Wireless Cellular Networks," In Proceedings of the 2019 Workshop on Network Meets AI & ML (NetAI'19), pp. 76-82, Aug. 2019. Masahidero Kane and Megumi Takeshita, "Proposal of a Method for Building a Throughput Estimation Model for Wireless Base Stations Based on a Hierarchical Bayesian Model Taking into Account the Number of Observations," IEICE University, B-11-7, Sep. 2022.

[0006] In the method proposed in Non-Patent Document 2, a parameter (weight parameter) is set to determine whether a single model or a common model is to be emphasized depending on the number of observations of a cell. However, since setting this parameter requires manual adjustment, such as analysis of observation data, building models for a large number of cells requires enormous costs.

[0007] The present disclosure has been made in consideration of the above points, and aims to set parameters that represent weights between models.

[0008] A parameter setting device according to one aspect of the present disclosure is a parameter setting device that sets parameters representing weights between a first model constructed by a hierarchical Bayesian modeling technique using observation data of a target cell and a second model constructed by the hierarchical Bayesian modeling technique using observation data of cells other than the target cell, and includes: a division unit that divides traffic demands included in the observation data of each cell, including the target cell, into multiple groups; a first calculation unit that calculates a first index value representing how much importance to assign to the first model based on information about the multiple groups and the observation data; a second calculation unit that calculates a second index value representing how much importance to assign to the second model based on information about the multiple groups and the observation data; a parameter calculation unit that calculates the parameters based on the first index value and the second index value; and a setting unit that sets the parameters in a base station of the target cell.

[0009] A parameter representing the weight between models can be set.

[0010] It is a diagram showing an example of a hardware configuration of the parameter setting device according to the present embodiment. It is a diagram showing an example of a functional configuration of the parameter setting device according to the present embodiment. It is a flowchart showing an example of a parameter setting process according to the present embodiment.

[0011] An embodiment of the present invention will be described below with reference to the drawings. In the following embodiment, a parameter setting device 10 capable of setting a parameter (weight parameter) that determines whether to prioritize the common model or the single model proposed in Non-Patent Document 2 will be described. Here, both the common model and the single model are models (quality models) that estimate the quality (e.g., throughput, etc.) of a cell (base station) from observation data, and are constructed using a hierarchical Bayesian modeling technique that takes into account the number of observation data (number of observations) for each cell. A common model is a model constructed using observation data from other cells as well (in other words, a model constructed using observation data from all predetermined cells), and a single model is a model constructed using only observation data from a certain cell. The quality model may also be called, for example, a "performance model."

[0012] Hereinafter, the cell for which a parameter is to be set will be referred to as the "target cell." Furthermore, the parameter to be set in the target cell (more precisely, set in the base station of the target cell) is denoted by σ ０ The parameter setting device 10 according to this embodiment receives the observation data D ０ = {X ０ , Y ０} and the observation data D of the cell i other than the target cell ｉ = {X ｉ , Y ｉ} (i=1, 2, ..., C), where C is the total number of other cells. ０ = {x ０ （ｎ） |n=1,2,...,N ０} is the set of traffic demands x observed in the target cell, Y ０ = {y ０ （ｎ） |n=1,2,...,N ０} is the set of observed qualities y in the target cell. ０ is the number of observations in the target cell. Similarly, X ｉ = {x ｉ （ｎ） |n=1,2,...,N ｉ} and Y ｉ = {y ｉ（ｎ） |n=1,2,...,N ｉ} are the sets of traffic demands x and qualities y observed in cell i, respectively, and N ｉ is the number of observations in cell i, where x ｉ （ｎ） and y ｉ （ｎ） (i=0, 1, 2, . . . , C) are traffic demand x and quality y obtained at the same time.

[0013] The traffic demand x may be, for example, the number of active users or the resource block usage rate. Alternatively, multiple traffic demands such as the number of active users and the resource block usage rate may be used as the traffic demand x.

[0014] The quality y may be, for example, a throughput, an RRC connection completion rate, or an HO success rate. Furthermore, multiple qualities such as the throughput, the RRC connection completion rate, and the HO success rate may be used as the quality y.

[0015] In addition, observation data D ｉ The traffic demand x included in (i=0, 1, 2, ..., C) may be called, for example, an "observed value of traffic demand" or an "observed traffic demand." Similarly, the observed data D ｉ The quality y included in (i=0, 1, 2, . . . , C) may be called, for example, an "observed value of quality" or an "observed quality."

[0016] <Example of Hardware Configuration of Parameter Setting Device 10> An example of the hardware configuration of the parameter setting device 10 according to this embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram showing an example of the hardware configuration of the parameter setting device 10 according to this embodiment.

[0017] 1, a parameter setting device 10 according to this embodiment includes an input device 101, a display device 102, an external I / F 103, a communication I / F 104, a random access memory (RAM) 105, a read only memory (ROM) 106, an auxiliary storage device 107, and a processor 108. Each of these pieces of hardware is connected to each other via a bus 109 so as to be able to communicate with each other.

[0018] The input device 101 is, for example, a keyboard, a mouse, a touch panel, a physical button, etc. The display device 102 is, for example, a display, a display panel, etc. Note that the parameter setting device 10 does not necessarily have to include at least one of the input device 101 and the display device 102, for example.

[0019] The external I / F 103 is an interface with an external device such as a recording medium 103a. The parameter setting device 10 can read from and write to the recording medium 103a via the external I / F 103. Examples of the recording medium 103a include a flexible disk, a CD (Compact Disc), a DVD (Digital Versatile Disk), an SD memory card (Secure Digital memory card), and a USB (Universal Serial Bus) memory card.

[0020] The communication I / F 104 is an interface that allows the parameter setting device 10 to communicate with other devices and equipment. The RAM 105 is a volatile semiconductor memory (storage device) that temporarily stores programs and data. The ROM 106 is a non-volatile semiconductor memory (storage device) that can store programs and data even when the power is turned off. The auxiliary storage device 107 is a storage device (storage device) such as an HDD (Hard Disk Drive), an SSD (Solid State Drive), or a flash memory. The processor 108 is an arithmetic device such as a CPU (Central Processing Unit).

[0021] The parameter setting device 10 according to this embodiment has the hardware configuration shown in Fig. 1 and is therefore capable of implementing the parameter setting process described below. Note that the hardware configuration shown in Fig. 1 is merely an example, and the hardware configuration of the parameter setting device 10 is not limited to this. For example, the parameter setting device 10 may have multiple auxiliary storage devices 107 or multiple processors 108, may not have some of the hardware shown in the figure, or may have various hardware components other than the hardware shown in the figure.

[0022] <Example of Functional Configuration of Parameter Setting Device 10> An example of the functional configuration of the parameter setting device 10 according to this embodiment will be described with reference to Fig. 2. Fig. 2 is a diagram showing an example of the functional configuration of the parameter setting device 10 according to this embodiment.

[0023] 2, the parameter setting device 10 according to this embodiment includes an input unit 201, a division unit 202, a first uncertainty calculation unit 203, a second uncertainty calculation unit 204, a parameter calculation unit 205, and a parameter setting unit 206. Each of these units is realized by, for example, processing in which one or more programs installed in the parameter setting device 10 are executed by the processor 108 or the like.

[0024] The input unit 201 receives the given observation data D ｉ (i=0, 1, 2, ..., C) is input.

[0025] The division unit 202 divides the observation data D input by the input unit 201 into ｉ Traffic demand x included in ｉ （ｎ） into multiple groups.

[0026] The first uncertainty calculation unit 203 calculates the information on the plurality of groups divided by the division unit 202 and the observation data D input by the input unit 201. ｉ Based on the above, a first uncertainty is calculated as an index value representing how much importance is attached to a single model.

[0027] The second uncertainty calculation unit 204 calculates the information on the plurality of groups divided by the division unit 202 and the observation data D input by the input unit 201. ｉ Based on the above, a second uncertainty is calculated as an index value representing how much importance is attached to the common model.

[0028] The parameter calculation unit 205 calculates a parameter σ based on the first uncertainty calculated by the first uncertainty calculation unit 203 and the second uncertainty calculated by the second uncertainty calculation unit 204. ０ Calculate.

[0029] The parameter setting unit 206 sets the parameter σ calculated by the parameter calculation unit 205. ０ is set to the base station of the target cell.

[0030] <Parameter Setting Process> An example of the parameter setting process according to this embodiment will be described with reference to Fig. 3. Fig. 3 is a flowchart showing an example of the parameter setting process according to this embodiment.

[0031] The input unit 201 receives the given observation data D ｉ (i=0, 1, 2, . . . , C) is input (step S101).

[0032] The division unit 202 divides the observation data D input in step S101 into ｉ Traffic demand x included in ｉ （ｎ） Specifically, the dividing unit 202 divides the traffic demand x into a plurality of groups according to a predetermined interval Δ. ｉ （ｎ） is divided into groups at equal intervals. Here, Δ is, for example, the observed data D ｉ Traffic demand x included in ｉ （ｎ） The value may be set to a value that divides the number into 100 or a specified number, or may be set to an arbitrary value.

[0033] Hereinafter, each group obtained by the division in step S102 is referred to as G ｊ(j=1, 2, ..., G), where G is the number of groups. ｊ Traffic demand of the target cell x ０ （ｎ） ∈X ０ A set of groups that contains one or more ０ , the number of elements (i.e., the traffic demand of the target cell x ０ （ｎ） ∈X ０ The total number of groups that contain one or more ０ |, G ０ Group G included in ｊ Let the set of indices of ０ Furthermore, j∈J ０ Group G ｊ Regarding Group G ｊ Traffic demand x included in ０ （ｎ） ∈X ０ The number of M ０，ｊ In the following, for ease of understanding, we will refer to group G ｊ Traffic demand x included in ０ （ｎ） ∈X ０ to "x ０，ｊ （ｎ） " and x ０，ｊ （ｎ） The quality y corresponding to ０ （ｎ） ∈Y ０ to "y ０，ｊ （ｎ） " will also be expressed as ".

[0034] The first uncertainty calculation unit 203 calculates the group G obtained by the division in step S102. ｊ (j=1, 2, . . . , G) and the observation data D input in step S101. ｉ(i=0, 1, 2, ..., C) (step S103). The first uncertainty calculation unit 203 can calculate any index value as the first uncertainty as long as it is an index value that indicates how much importance is attached to a single model. For example, when an average model for traffic demand x is constructed as a single model, the first uncertainty calculation unit 203 calculates the first uncertainty based on the traffic demand x of the target cell. ０ （ｎ） ∈X ０ A set G of groups containing ０ Number of elements in |G ０ | and each group G ｊ traffic demand x for the target cell in (j=1, 2, ..., G) ０，ｊ （ｎ） ∈X ０ The quality y corresponding to ０，ｊ （ｎ） ∈Y ０ The variance of the expected value of σ ０，ｊ ２ and the first uncertainty U ０ Calculate.

[0035] However, U ０ The larger the value of σ, the greater the uncertainty. ０，ｊ ２ By the central limit theorem, the variance of the original population, σ ２ can be expressed as follows using

[0036] That is, j∈J ０ In contrast to M ０，ｊ The larger the number of observation data of the target cell, N ０ ), the variance is smaller and the first uncertainty U ０ The uncertainty of the original population variance σ ２ is precisely an infinite number of qualities ０，ｊ （ｎ） Therefore, instead, for example, the quality y ｉ （ｎ） (n=1,2,...,N ｉ , i = 0, 1, ..., C)２ the variance of the original population σ ２ Used as.

[0037] The second uncertainty calculation unit 204 calculates the group G obtained by the division in step S102. ｊ (j=1, 2, . . . , G) and the observation data D input in step S101. ｉ (i=0, 1, 2, ..., C) (step S104). The second uncertainty calculation unit 204 can calculate any index value as the second uncertainty as long as it is an index value that indicates how much importance is attached to the common model. For example, the second uncertainty calculation unit 204 calculates the second uncertainty based on the number of groups G and the number of groups G ｊ The distribution of the expected value of quality y corresponding to all traffic demands x in (j = 1, 2, ..., G) and its group G ｊ Traffic demand for the target cell at x ０，ｊ （ｎ） ∈X ０ The quality y corresponding to ０，ｊ （ｎ） ∈Y ０ The distribution that the expectation of follows and the variance σ of the mixture distribution ｊ ２ and the second uncertainty U １ Calculate.

[0038] where σ ｊ ２ Group G ｊ The distribution of the expected value of all qualities y in the group G ｊ Traffic demand for the target cell at x ０，ｊ （ｎ） ∈X ０ The quality y corresponding to ０，ｊ （ｎ） ∈Y ０ is the variance of a mixture distribution consisting of the distribution that the expectation of follows.

[0039] Group G ｊ Let μ be the mean of the distribution that the expected values ​​of all qualities y in １，ｊ , variance is σ １，ｊ ２ Similarly, the group Gｊ Traffic demand for the target cell at x ０，ｊ （ｎ） ∈X ０ The quality y corresponding to ０，ｊ （ｎ） ∈Y ０ Let μ be the mean of the distribution that the expected value of follows. ０，ｊ , variance is σ ０，ｊ ２ In this case, the variance σ ｊ ２ can be expressed as follows:

[0040] Also, Group G ｊ Let the number of all traffic demands x included in M １，ｊ Then, σ ０，ｊ ２ Similarly, σ １，ｊ ２ The variance of the original population σ ２ can be expressed as follows using

[0041] On the other hand, μ １，ｊ Group G ｊ The quality y corresponding to all traffic demands x in the ０，ｊ Group G ｊ Traffic demand for the target cell at x ０，ｊ （ｎ） ∈X ０ The quality y corresponding to ０，ｊ （ｎ） ∈Y ０ However, the traffic demand for the target cell x ０ （ｎ） ∈X ０ Group G does not include ｊ Regarding ｊ The quality obtained as the output when the average traffic demand x of the target cell is input to the single model may be used. ０ （ｎ） ∈X ０ Group G does not include ｊ σ with respect to ０，ｊ ２As the traffic demand for the target cell, x ０ （ｎ） ∈X ０ Group G includes ｊ σ with respect to ０，ｊ ２ The average value of may be used.

[0042] The parameter calculation unit 205 calculates the first uncertainty U ０ and the second uncertainty U calculated in step S104 above. １ and based on the parameter σ ０ At this time, the parameter calculation unit 205 calculates the parameter σ so that it becomes larger as the emphasis is placed on the single model and becomes smaller as the emphasis is placed on the common model. ０ For example, the parameter calculation unit 205 calculates the value of the parameter σ ０ Calculate.

[0043] This allows the first uncertainty U ０ is small, or the second uncertainty U １ The larger the uncertainty, the more importance is attached to a single model. ０ The larger the parameter σ, the more importance is placed on the common model. ０ However, U ０ and U １ is a real number greater than or equal to 0.

[0044] The parameter setting unit 206 sets the parameter σ calculated in step S105. ０ is set in the base station of the target cell (step S106). As a result, the target cell ０ It is possible to weight the single model and the common model according to

[0045] As described above, the parameter setting device 10 according to the present embodiment can automatically calculate and set weighting parameters from observation data that indicate whether to prioritize a single model or a common model constructed using the hierarchical Bayesian modeling technique, thereby reducing the cost required for parameter adjustment (e.g., the human and time costs of manually adjusting and setting parameters).

[0046] The present invention is not limited to the above-described specifically disclosed embodiments, and various modifications, changes, and combinations with known technologies are possible without departing from the scope of the claims.

[0047] 10 Parameter setting device 101 Input device 102 Display device 103 External I / F 103a Recording medium 104 Communication I / F 105 RAM 106 ROM 107 Auxiliary storage device 108 Processor 109 Bus 201 Input unit 202 Dividing unit 203 First uncertainty calculation unit 204 Second uncertainty calculation unit 205 Parameter calculation unit 206 Parameter setting unit

Claims

1. A parameter setting device that sets parameters that represent the weighting between a first model constructed by a hierarchical Bayesian modeling technique using observation data of a target cell and a second model constructed by the hierarchical Bayesian modeling technique using observation data of cells other than the target cell, the parameter setting device having: a division unit that divides traffic demands contained in the observation data of each cell, including the target cell, into multiple groups; a first calculation unit that calculates a first index value that represents how much importance is to be placed on the first model based on information about the multiple groups and the observation data; a second calculation unit that calculates a second index value that represents how much importance is to be placed on the second model based on information about the multiple groups and the observation data; a parameter calculation unit that calculates the parameters based on the first index value and the second index value; and a setting unit that sets the parameters in the base station of the target cell.

2. The parameter setting device of claim 1, wherein the first calculation unit calculates the first index value based on the number of elements in a set of target groups, which are groups among the plurality of groups that include traffic demand included in the observation data of the target cell, and the variance of the expected value of quality corresponding to the traffic demand included in the observation data of the target cell among the traffic demands within the target groups.

3. The parameter setting device described in claim 2, wherein the second calculation unit calculates the second index value based on the variance of a mixed distribution consisting of a distribution to which the expected values ​​of quality corresponding to all traffic demands in the target group follow and a distribution to which the expected values ​​of quality corresponding to traffic demands in the target group that are included in the observation data of the target cell follow.

4. A parameter setting method in which a parameter setting device sets parameters that represent weights between a first model constructed by a hierarchical Bayesian modeling technique using observation data of a target cell and a second model constructed by the hierarchical Bayesian modeling technique using observation data of cells other than the target cell, the parameter setting device executing: a division procedure that divides traffic demand contained in the observation data of each cell, including the target cell, into multiple groups; a first calculation procedure that calculates a first index value that represents how much importance is to be placed on the first model based on information about the multiple groups and the observation data; a second calculation procedure that calculates a second index value that represents how much importance is to be placed on the second model based on information about the multiple groups and the observation data; a parameter calculation procedure that calculates the parameters based on the first index value and the second index value; and a setting procedure that sets the parameters in the base station of the target cell.

Images