JP2022039617A - Time series prediction system, time series prediction method, and computer program - Google Patents

Abstract

To make a prediction even in a place where observation data that is the basis of the prediction cannot be obtained when the usage status of a specific radio frequency is predicted.SOLUTION: A time series prediction system includes an observation data acquisition unit that acquires an observation value observed in each observation area for first radio wave condition information of a first radio frequency and second radio wave condition information of a second radio frequency, an estimation model generation unit that generates an estimation model that outputs an estimated value of the first radio wave condition information from the observation value of the second radio wave condition information on the basis of the observation value of the first radio wave condition information and the observation value of the second radio wave condition information, a radio wave condition information estimation unit that acquires the estimated value of the first radio wave condition information of the estimated time of the estimated area from the observation value of the second radio wave condition information of the estimated time of the estimated area using the estimation model, and a time-series prediction unit that calculates the predicted value of the future first radio wave condition information of the estimated area on the basis of time-series data of the estimated value of the first radio wave condition information.SELECTED DRAWING: Figure 1

Description

The present invention relates to a time series prediction system, a time series prediction method and a computer program.

In recent years, it has been studied that a plurality of different wireless communication systems share the same frequency band. For example, it is considered that another wireless communication system (secondary operator system) secondarily uses the frequency band (shared frequency band) assigned to the existing wireless communication system (primary operator system). ing. In this frequency sharing between different systems, there are restrictions on the time and place where the secondary operator system can use the shared frequency band so as not to interfere with the operation of the primary operator system. Therefore, it is desirable to be able to predict the future usage of the shared frequency band of the primary operator system from the viewpoint of investment decision of the secondary operator system and formulation of utilization plan.

Patent Document 1 describes a technique for predicting future data of busy and idle time series data of a radio channel by using a stochastic neural network from the busy and idle time series data of the radio channel. ing.

Japanese Unexamined Patent Publication No. 2019-101792

However, in the technique described in Patent Document 1 described above, the time-series data of the busy state and the idle state of the radio channel can be observed in a place where the time-series data of the busy state and the idle state of the radio channel cannot be observed. Future data cannot be predicted.

The present invention has been made in consideration of such circumstances, and an object thereof is for a place where observation data on which the prediction is based cannot be obtained when predicting the usage status of a specific radio frequency. Is also to make predictions.

(1) In one aspect of the present invention, the first radio wave state information indicating the radio wave state of the first radio frequency and the second radio wave state information representing the radio wave state of the second radio wave different from the first radio frequency are provided. An observation data acquisition unit that acquires the observed values observed in each observation area, and an estimation model that outputs the estimated value of the first radio wave condition information from the observation value of the second radio wave condition information acquired from the observation area. , An estimation model generation unit generated based on the observed value of the first radio wave condition information and the observation value of the second radio wave condition information acquired by the observation data acquisition unit, and estimation of the estimation area of the first radio wave condition information. The radio wave condition obtained by using the estimation model from the observation value of the second radio wave condition information at the estimated time in the estimated area acquired by the observation data acquisition unit, the estimated value of the first radio wave condition information at the time. An information estimation unit, a time-series prediction unit that calculates a future predicted value of the first radio wave condition information in the estimated area based on the time-series data of the estimated value of the first radio wave condition information in the estimated area, and a time-series prediction unit. It is a time series prediction system equipped with.
(2) In one aspect of the present invention, the first radio wave state information indicating the radio wave state of the first radio frequency and the second radio wave state information representing the radio wave state of the second radio wave different from the first radio frequency are provided. The future first radio wave condition information of the observation area based on the observation data acquisition unit that acquires the observation value observed in each observation area and the time-series data of the observation value of the first radio wave condition information of the observation area. With a time-series prediction unit that calculates the predicted value of the second radio wave condition information in the observation area and calculates the predicted value of the second radio wave condition information in the future of the observation area based on the time-series data of the observation value of the second radio wave condition information in the observation area. , The estimation model that outputs the predicted value of the first radio wave condition information in the future from the predicted value of the second radio wave condition information in the future of the observation area is the first radio wave condition information calculated by the time series prediction unit. The time-series prediction of the estimation model generation unit generated based on the predicted value and the predicted value of the second radio wave state information, and the predicted value of the first radio wave state information in the future of the estimation area of the first radio wave state information. It is a time-series prediction system including a radio wave condition information prediction estimation unit obtained by using the estimation model from the future prediction value of the second radio wave condition information of the estimation area calculated by the unit.
(3) One aspect of the present invention is the time-series prediction system described above in which the estimation model is generated for each cluster area as a result of clustering a plurality of observation areas.

(4) In one aspect of the present invention, the first radio wave state information indicating the radio wave state of the first radio frequency and the second radio wave state information representing the radio wave state of the second radio wave different from the first radio frequency are provided. An observation data acquisition step for acquiring the observed values observed in each observation area, and an estimation model for outputting the estimated value of the first radio wave condition information from the observed value of the second radio wave condition information acquired from the observation area. , An estimation model generation step generated based on the observed value of the first radio wave condition information and the observed value of the second radio wave condition information acquired in the observation data acquisition step, and estimation of the estimation area of the first radio wave condition information. The radio wave condition obtained by using the estimation model from the observation value of the second radio wave condition information at the estimated time in the estimated area acquired in the observation data acquisition step, the estimated value of the first radio wave condition information at the time. An information estimation step, a time-series prediction step for calculating a future predicted value of the first radio wave condition information in the estimated area based on the time-series data of the estimated value of the first radio wave condition information in the estimated area, and a time-series prediction step. It is a time series prediction method including.
(5) In one aspect of the present invention, the first radio wave state information indicating the radio wave state of the first radio frequency and the second radio wave state information representing the radio wave state of the second radio wave different from the first radio frequency are provided. The future first radio wave condition information of the observation area based on the observation data acquisition step of acquiring the observation value observed in each observation area and the time series data of the observation value of the first radio wave condition information of the observation area. And the time-series prediction step to calculate the predicted value of the second radio wave condition information in the future of the observation area based on the time-series data of the observation value of the second radio wave condition information of the observation area. , The estimation model that outputs the predicted value of the first radio wave condition information in the future from the predicted value of the second radio wave condition information in the future of the observation area is the first radio wave condition information calculated in the time series prediction step. The time-series prediction of the estimation model generation step generated based on the predicted value and the predicted value of the second radio wave state information, and the predicted value of the first radio wave state information in the future of the estimation area of the first radio wave state information. It is a time series prediction method including a radio wave condition information prediction estimation step acquired by using the estimation model from the future prediction value of the second radio wave condition information of the estimation area calculated in the step.

(6) One aspect of the present invention is to provide a computer with first radio wave state information indicating a radio wave state of a first radio frequency and a second radio wave state indicating a radio wave state of a second radio frequency different from the first radio frequency. Information outputs the estimated value of the first radio wave condition information from the observation data acquisition step of acquiring the observation value observed in each observation area and the observation value of the second radio wave condition information acquired from the observation area. An estimation model generation step for generating an estimation model based on the observed value of the first radio wave condition information acquired in the observation data acquisition step and the observation value of the second radio wave condition information, and estimation of the first radio wave condition information. The estimated value of the first radio wave condition information of the estimated time of the area is acquired from the observed value of the second radio wave condition information of the estimated time of the estimated area acquired in the observation data acquisition step using the estimation model. Time-series prediction to calculate the predicted value of the future first radio wave condition information of the estimated area based on the time-series data of the radio wave condition information estimation step to be performed and the estimated value of the first radio wave condition information of the estimated area. It is a computer program for executing steps.
(7) In one aspect of the present invention, a computer is provided with first radio wave state information indicating a radio wave state of a first radio frequency and a second radio wave state indicating a radio wave state of a second radio frequency different from the first radio frequency. The information is the observation data acquisition step of acquiring the observation value observed in each observation area, and the future first of the observation area based on the time-series data of the observation value of the first radio wave condition information of the observation area. A time series that calculates the predicted value of the radio wave condition information and calculates the predicted value of the future second radio wave condition information of the observation area based on the time series data of the observation value of the second radio wave condition information of the observation area. The first radio wave calculated in the time series prediction step is an estimation model that outputs a prediction value of the first radio wave condition information in the future from the prediction step and the prediction value of the second radio wave condition information in the future of the observation area. The estimation model generation step generated based on the predicted value of the state information and the predicted value of the second radio wave state information, and the predicted value of the first radio wave state information in the future of the estimation area of the first radio wave state information are described. It is a computer program for executing the radio wave condition information prediction estimation step acquired by using the estimation model from the future prediction value of the second radio wave condition information of the estimation area calculated in the time series prediction step.

According to the present invention, when predicting the usage status of a specific radio frequency, it is possible to obtain the effect that the prediction can be performed even in a place where the observation data on which the prediction is based cannot be obtained.

It is a block diagram which shows the structural example of the time series prediction system which concerns on 1st Embodiment. It is explanatory drawing which shows the example of the observation area which concerns on embodiment. It is explanatory drawing for demonstrating the procedure of the time series prediction method which concerns on 1st Embodiment. It is a block diagram which shows the structural example of the time series prediction system which concerns on 2nd Embodiment. It is explanatory drawing for demonstrating the procedure of the time series prediction method which concerns on 2nd Embodiment. It is a block diagram which shows the structural example of the time series prediction system which concerns on 3rd Embodiment. It is explanatory drawing for demonstrating the procedure of the time series prediction method which concerns on 3rd Embodiment.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[First Embodiment]
FIG. 1 is a block diagram showing a configuration example of a time series prediction system according to the first embodiment. In FIG. 1, the time-series prediction system 1 includes an observation data acquisition unit 11, an estimation model generation unit 12, a radio wave state information estimation unit 13, and a time-series prediction unit 14.

Each function of the time-series prediction system 1 is that the time-series prediction system 1 is equipped with computer hardware such as a CPU (Central Processing Unit) and a memory, and the CPU executes a computer program stored in the memory. Is realized by. The time-series prediction system 1 may be configured by using a general-purpose computer device or may be configured as a dedicated hardware device. For example, the time series prediction system 1 may be configured by using a server computer connected to a communication network such as the Internet. Further, each function of the time series prediction system 1 may be realized by cloud computing. Further, the time-series prediction system 1 may be realized by a single computer, or may be realized by distributing the functions of the time-series prediction system 1 to a plurality of computers.

In the present embodiment, the frequency f1 which is the first radio frequency and the frequency f2 which is the second radio frequency are used as specific radio frequencies. The frequency f1 and the frequency f2 are different radio frequencies. It is preferable that the frequency f1 and the frequency f2 have similar radio wave propagation characteristics.

The observation data acquisition unit 11 has radio wave state information (first radio wave state information) indicating the radio wave state of the frequency f1 (first radio frequency) and frequency f2 (second radio frequency) different from the frequency f1 (first radio frequency). ) Radio wave condition information (second radio wave condition information) indicating the radio wave condition and the observed value observed in each observation area are acquired.

FIG. 2 shows an example of an observation area having a frequency f1 and an observation area having a frequency f2. FIG. 2 shows observation areas Pf1-1, Pf1-15, and Pf1-22 as examples of observation areas having a frequency f1 in 25 areas. In each observation area of frequency f1, a radio wave sensor or the like for observing radio wave condition information of frequency f1 is installed.

Further, FIG. 2 shows observation areas Pf2-4 and Pf2-15 as examples of observation areas having a frequency f2 in 25 areas. In each observation area of frequency f2, a radio wave sensor or the like for observing radio wave condition information of frequency f2 is installed. The frequency f2 is, for example, a radio frequency used by a wide area radio communication system, and is observed in all areas of FIG. 2, for example.

The radio wave state information is, for example, received power. In this case, a radio wave sensor for observing the received power of the frequency f1 is installed in each observation area of the frequency f1. Further, in each observation area of frequency f2, a radio wave sensor for observing the received power of frequency f2 is installed.

Areas P-1, P-15, and P-22 are observation areas of frequency f1 and observation areas of frequency f2. Area P-4 is an area (non-observation area) from which observation data of frequency f1 cannot be obtained, but is an observation area of frequency f2. In the non-observation area of frequency f1, a radio wave sensor or the like for observing radio wave condition information of frequency f1 is not installed. Therefore, the observation data acquisition unit 11 cannot acquire the observed value of the radio wave state information (for example, received power) of the frequency f1 in the non-observation area of the frequency f1 such as the area P-4.

The observation data acquisition unit 11 is the observation value A-Pf1-1, A-Pf1 of the radio wave condition information of the frequency f1 observed in each observation area Pf1-1, Pf1-15, Pf1-22, ... Of the frequency f1. -15, A-Pf1-22, ... And the radio wave condition information of the frequency f2 observed in each observation area Pf2-1, Pf2-4, Pf2-15, Pf2-22, ... of the frequency f2. The observed values A-Pf2-1, A-Pf2-4, A-Pf2-15, A-Pf2-22, ... Are acquired.

The estimation model generation unit 12 generates the estimation model C. The estimation model C is a model that outputs an estimated value of the radio wave condition information of the frequency f1 from the observed value of the radio wave condition information of the frequency f2 acquired from the observation area of the frequency f2. In the estimation model generation unit 12, the observed value B of the radio wave condition information of the frequency f1 and the observed value B of the radio wave condition information of the frequency f2 observed in the area which is the observation area of the frequency f1 and the observation area of the frequency f2 are used. It is input from the observation data acquisition unit 11. The estimation model generation unit 12 generates an estimation model C based on the observation value B of the radio wave condition information of the frequency f1 and the observation value B of the radio wave condition information of the frequency f2 input from the observation data acquisition unit 11.

As an example of the method of generating the estimation model C, the estimation model generation unit 12 generates the estimation model C by using the machine learning model. The machine learning model is, for example, a neural network. The estimation model generation unit 12 performs machine learning of a machine learning model (for example, a neural network) by using the observed value B of the radio wave state information of the frequency f1 and the observed value B of the radio wave state information of the frequency f2. This machine learning is to learn the relationship of the fluctuation pattern of the radio wave state information between the frequency f1 and the frequency f2. By this machine learning, the estimation model C is configured to output the time-series data of the estimated value of the radio wave state information of the frequency f1 from the time-series data of the observed value of the radio wave state information of the frequency f2.

According to the estimation model C, for example, the area P-4 which is the non-observation area of the frequency f1 and the observation area of the frequency f2 in FIG. 2, the radio wave condition information of the frequency f2 acquired from the area P-4. From the observed value, the estimated value of the radio wave condition information of the frequency f1 of the area P-4 can be obtained.

The radio wave condition information estimation unit 13 uses the estimation model C generated by the estimation model generation unit 12 from the observation value B of the frequency f2 acquired by the observation data acquisition unit 11 to use the non-observation area (estimation area) of the frequency f1. The estimated value D of the radio wave state information of the frequency f1 of is acquired. More specifically, the radio wave state information estimation unit 13 has acquired the estimation area of frequency f1 which is the non-observation area of frequency f1 and the observation area of frequency f2. From the observed value B of the frequency f2 at the estimated time of the above, the estimated value D of the radio wave condition information of the frequency f1 of the estimated time of the estimated area is acquired by using the estimation model C. For example, the radio wave condition information estimation unit 13 has a frequency of the estimated time t acquired from the area P-4 with respect to the area P-4 which is the non-observation area of the frequency f1 and the observation area of the frequency f2 in FIG. From the observed value B of the radio wave condition information of f2, the estimated value D of the radio wave condition information of the frequency f1 at the estimated time t of the area P-4 is acquired by using the estimation model C.

The time-series prediction unit 14 calculates the predicted value E of the future radio wave condition information of the frequency f1 based on the time-series data of the radio wave condition information of the frequency f1. Further, the time-series prediction unit 14 calculates the predicted value E of the future radio wave state information of the frequency f2 based on the time-series data of the radio wave state information of the frequency f2.

The time-series prediction unit 14 has the frequency f1 of the area P-1 acquired by the observation data acquisition unit 11 with respect to the area P-1 of FIG. 2, which is the observation area of the frequency f1 and the observation area of the frequency f2, for example. Based on the time-series data of the observed value B of the radio wave condition information, the predicted value E of the future radio wave condition information of the frequency f1 of the area P-1 is calculated. Further, the time-series prediction unit 14 indicates the area P-1 based on the time-series data of the observation value B of the radio wave condition information of the frequency f2 of the area P-1 acquired by the observation data acquisition unit 11. The predicted value E of the future radio wave condition information of the frequency f2 of P-1 is calculated.

The time-series prediction unit 14 is a non-observation area having a frequency f1 and an observation area having a frequency f2, for example, an area P-4 (estimation) acquired by the radio wave condition information estimation unit 13 with respect to the area P-4 in FIG. Based on the time series data of the estimated value D of the radio wave condition information of the frequency f1 of the area), the predicted value E of the future radio wave condition information of the frequency f1 of the area P-4 (estimated area) is calculated. Further, the time-series prediction unit 14 indicates the area P-4 based on the time-series data of the observation value B of the radio wave condition information of the frequency f2 of the area P-4 acquired by the observation data acquisition unit 11. The predicted value E of the future radio wave condition information of the frequency f2 of P-4 is calculated.

The time-series prediction unit 14 uses, for example, an LSTM (Long Short Term Memory) or a GRU (Gated Recurrent Unit) as a time-series prediction method.

Next, a time series prediction method according to the present embodiment will be described with reference to FIG. FIG. 3 is an explanatory diagram for explaining the procedure of the time series prediction method according to the first embodiment. Hereinafter, the observation areas Pf1-1, Pf1-15, Pf1-22, ... Of the frequency f1 in FIG. 2, the observation areas Pf2-4, Pf2-15, ... Of the frequency f2, and the non-observation areas of the frequency f1 ( Estimated area) P-4 will be described as an example. Further, here, the received power will be described as an example of the radio wave condition information.

(Step S1) The observation data acquisition unit 11 has each observation area of frequency f1 with respect to the areas P-1, P-15, P-22, ... Which are the observation areas of frequency f1 and the observation areas of frequency f2. Observation values of radio wave condition information (received power) of frequency f1 observed at Pf1-1, Pf1-15, Pf1-22, ... A-Pf1-1, A-Pf1-15, A-Pf1-22, ... and the observed values A-Pf2-1, A of the radio wave state information (received power) of the frequency f2 observed in each observation area Pf2-1, Pf2-15, Pf2-22, ... -Pf2-15, A-Pf2-22, ... Are acquired at regular intervals. As a result, the observed value B of the radio wave condition information (received power) of the frequency f1 in the areas P-1, P-15, P-22, ... Which is the observation area of the frequency f1 and the observation area of the frequency f2. The time-series data and the time-series data of the observed value B of the radio wave state information (received power) of the frequency f2 are acquired.

(Step S2) Step S2 is a time-series prediction process for the areas P-1, P-15, P-22, ... Which are the observation areas of the frequency f1 and the observation areas of the frequency f2. The time-series prediction unit 14 has each observation area Pf1-1, Pf1-15, Pf1-22, ... Acquired by the observation data acquisition unit 11 for each observation area Pf1-1, Pf1-15, Pf1-22, ... -22, ... Frequency f1 of each observation area Pf1-1, Pf1-15, Pf1-22, ... Based on the time series data of the observation value B of the radio wave state information (received power) of the frequency f1. The predicted value E of the future radio wave condition information (received power) of is calculated. Further, the time-series prediction unit 14 has each observation area Pf2-1, Pf2-15 acquired by the observation data acquisition unit 11 for each observation area Pf2-1, Pf2-15, Pf2-22, ... Of the frequency f2. , Pf2-22, ... Based on the time-series data of the observed value B of the radio wave state information (received power) of the frequency f2, each observation area Pf2-1, Pf2-15, Pf2-22, ... The predicted value E of the future radio wave condition information (received power) of the frequency f2 is calculated.

(Step S10) The estimation model generation unit 12 has radio wave condition information of the frequency f1 of the areas P-1, P-15, P-22, ... Which is the observation area of the frequency f1 and the observation area of the frequency f2. Based on the time-series data of the observed value B of the received power) and the time-series data of the observed value B of the radio wave state information (received power) of the frequency f2, the non-observed area (estimated area) P-4 of the frequency f1 is set. Generate an estimation model C of the included model generation target area. The time-series data used to generate the estimation model C may be time-series data of all observation areas or may be time-series data of some observation areas.

(Step S20) The observation data acquisition unit 11 receives radio wave condition information (reception) of the frequency f2 observed in the area P-4 (observation area Pf2-4) which is the non-observation area of the frequency f1 and the observation area of the frequency f2. The observed value A-Pf2-4 of (electricity) is acquired at a fixed cycle. As a result, the time-series data of the observed value B of the radio wave state information (received power) of the frequency f2 of the area P-4 which is the non-observed area of the frequency f1 and the observed area of the frequency f2 is acquired.

(Step S21) The radio wave condition information estimation unit 13 is generated in step S10 from the time series data of the observed value B of the radio wave condition information (received power) of the frequency f2 of the non-observation area (estimated area) P-4 of the frequency f1. Using the estimated model C, the time series data of the estimated value D of the radio wave state information (received power) of the frequency f1 of the estimated area P-4 is acquired.

The estimation model C may be updated as needed. In this case, the estimation model C before the update is continuously used until the estimation model C is newly updated. For example, when the radio wave propagation characteristics of the model generation target area change by a certain amount or more, the estimation model C of the model generation target area is updated.

Further, when the radio wave propagation characteristics of the model generation target area change in a specific period or more, the estimation model C of the model generation target area for the specific period may be generated. For example, when the radio wave propagation characteristics of the model generation target area change more than a certain amount between weekdays and weekends, the estimation model C of the model generation target area for weekdays and the estimation model C of the model generation target area for weekends are generated, respectively. May be done.

(Step S22) Step S22 is a time-series prediction process for the area P-4 which is the non-observation area (estimated area) of the frequency f1 and the observation area of the frequency f2. The time-series prediction unit 14 refers to the radio wave condition information (received power) of the frequency f1 of the area P-4 acquired by the radio wave condition information estimation unit 13 with respect to the area P-4 which is the non-observation area (estimated area) of the frequency f1. ), The predicted value E of the future radio wave condition information (received power) of the frequency f1 of the area P-4 is calculated based on the time series data of the estimated value D. Further, the time-series prediction unit 14 observes the radio wave condition information (received power) of the frequency f2 of the area P-4 acquired by the observation data acquisition unit 11 with respect to the area P-4 which is the observation area of the frequency f2. Based on the time series data of B, the predicted value E of the future radio wave condition information (received power) of the frequency f2 of the area P-4 is calculated.

The received power of the frequency f1 is information indicating the usage status of the frequency f1 and is used as information indicating whether or not the radio channel of the frequency f1 is used. As a method of determining whether or not to use a wireless channel, for example, when the received power of the frequency f1 in a certain area is equal to or higher than a predetermined threshold, it is determined that the wireless channel of the frequency f1 in the area is used, and when not, it is determined. It may be determined that the radio channel of frequency f1 in the area is not used. Therefore, the predicted value E of the future radio wave condition information (received power) of the frequency f1 in each area is the information for predicting the future use of the radio channel of the frequency f1 in each area.

According to this embodiment, the future radio wave condition information (received power) of the frequency f1 is applied not only to the observation area where the observation data of the frequency f1 can be obtained but also to the non-observation area where the observation data of the frequency f1 cannot be obtained. Predicted value E can be obtained. As a result, when predicting the usage status of the frequency f1, it is possible to obtain the effect that the prediction can be performed even in a place (non-observation area of the frequency f1) where the observation data that is the basis of the prediction cannot be obtained.

[Second Embodiment]
FIG. 4 is a block diagram showing a configuration example of the time series prediction system according to the second embodiment. In FIG. 4, the parts corresponding to the parts of FIG. 1 are designated by the same reference numerals, and the description thereof will be omitted.

In the time-series prediction system 1a shown in FIG. 4, first, the time-series prediction unit 14 determines the predicted value E of the future radio wave condition information of the frequency f1 from the observed value B of the radio wave condition information of the frequency f1 for each observation area. It is calculated, and the predicted value E of the future radio wave condition information of the frequency f2 is calculated from the observed value B of the radio wave condition information of the frequency f2. Next, the estimation model generation unit 12a generates an estimation model Ca based on the predicted value E of the future radio wave condition information of the frequency f1 and the predicted value E of the future radio wave condition information of the frequency f2. This estimation model Ca is a model that outputs the predicted value Ea of the future radio wave condition information of the frequency f1 of the estimated area from the predicted value E of the future radio wave condition information of the frequency f2 in the estimated area of the frequency f1. Next, the radio wave condition information prediction estimation unit 30 predicts the future radio wave condition information of the frequency f2 of the area with respect to the area which is the non-observation area (estimation area) of the frequency f1 and the observation area of the frequency f2. From E, the estimated model Ca is used to obtain the predicted value Ea of the future radio wave condition information of the frequency f1 in the area.

The estimation model generation unit 12a generates an estimation model Ca. The estimation model Ca is a model that outputs the predicted value of the future radio wave condition information of the frequency f1 from the predicted value of the future radio wave condition information of the observation area of the frequency f2. In the estimation model generation unit 12a, the predicted value E of the radio wave condition information of the frequency f1 and the predicted value E of the radio wave condition information of the frequency f2 in the observation area of the frequency f1 and the observation area of the frequency f2 are time-series predicted. It is input from the unit 14. The estimation model generation unit 12a generates an estimation model Ca based on the predicted value E of the radio wave condition information of the frequency f1 and the predicted value E of the radio wave condition information of the frequency f2 input from the time series prediction unit 14.

The radio wave condition information prediction estimation unit 30 has a future frequency f2 of the area calculated by the time series prediction unit 14 with respect to an area that is a non-observation area (estimation area) of the frequency f1 and an observation area of the frequency f2. From the predicted value E of the radio wave condition information, the estimated model Ca is used to acquire the predicted value Ea of the future radio wave condition information of the frequency f1 in the area. For example, the radio wave condition information prediction estimation unit 30 is the area P-4 calculated by the time series prediction unit 14 with respect to the area P-4 which is the non-observation area of the frequency f1 and the observation area of the frequency f2 in FIG. From the predicted value E of the future radio wave condition information of the frequency f2 of, the estimation model Ca is used to acquire the predicted value Ea of the future radio wave condition information of the frequency f1 of the area P-4.

Next, a time series prediction method according to the present embodiment will be described with reference to FIG. FIG. 5 is an explanatory diagram for explaining the procedure of the time series prediction method according to the second embodiment. In FIG. 5, the parts corresponding to each step in FIG. 3 are designated by the same reference numerals, and the description thereof will be omitted.

Here, the observation areas Pf1-1, Pf1-15, Pf1-22, ... Of the frequency f1 in FIG. 2, the observation areas Pf2-4, Pf2-15, ... Of the frequency f2, and the non-observation area of the frequency f1. (Estimated area) P-4 will be described as an example. Further, here, the received power will be described as an example of the radio wave condition information.

Steps S1 and S2 are the same as those in the first embodiment described above.

(Step S10a) The estimation model generation unit 12a is the future radio wave state of the frequency f1 of the areas P-1, P-15, P-22, ... Which is the observation area of the frequency f1 and the observation area of the frequency f2. An estimation model Ca is generated based on the predicted value E of information (received power) and the predicted value E of future radio wave condition information (received power) of frequency f2. The predicted value E used to generate the estimation model Ca may be the predicted value E of all observation areas or the predicted value E of some observation areas.

Steps S20 and S22 are the same as those in the first embodiment described above. However, in step S22 of the present embodiment, the future radio wave condition of the frequency f2 of the area P-4 is relative to the area P-4 which is the non-observation area (estimated area) of the frequency f1 and the observation area of the frequency f2. Only the predicted value E of the information (received power) is calculated.

(Step S30) The radio wave condition information prediction estimation unit 30 is generated in step S10a from the predicted value E of the future radio wave condition information (received power) of the frequency f2 of the non-observation area (estimated area) P-4 of the frequency f1. Using the estimated model Ca, the predicted value Ea of the future radio wave condition information (received power) of the frequency f1 of the estimated area P-4 is acquired.

According to the present embodiment, as in the first embodiment described above, the frequency f1 is applied not only to the observation area where the observation data of the frequency f1 can be obtained but also to the non-observation area where the observation data of the frequency f1 cannot be obtained. It is possible to obtain the predicted value Ea of the future radio wave condition information (received power) of. As a result, when predicting the usage status of the frequency f1, it is possible to obtain the effect that the prediction can be performed even in a place (non-observation area of the frequency f1) where the observation data that is the basis of the prediction cannot be obtained.

[Third Embodiment]
FIG. 6 is a block diagram showing a configuration example of the time series prediction system according to the third embodiment. In FIG. 6, the parts corresponding to the parts of FIG. 1 are designated by the same reference numerals, and the description thereof will be omitted. In the time series prediction system 1b shown in FIG. 6, an area clustering unit 20 is provided.

The area clustering unit 20 clusters a plurality of areas as the model generation target area of the target for generating the estimation model C. The area clustering unit 20 notifies the estimation model generation unit 12b of the cluster area G as a result of the clustering as a model generation target area. The estimation model generation unit 12b generates each estimation model C for each cluster area G. Therefore, the estimation model C is generated for each cluster area as a result of clustering a plurality of areas.

As a method of clustering a plurality of areas as model generation target areas, for example, clustering may be performed based on the distance between the areas, the population density in the area, the density of buildings in the area, and the like. For example, areas within a certain distance between areas are clustered into the same cluster area. For example, areas with similar population densities in an area are clustered into the same cluster area. For example, areas with similar density of buildings in the area are clustered into the same cluster area. According to these clustering methods, it is possible to cluster areas having similar radio wave propagation characteristics into the same cluster area. Thereby, the same estimation model C can be applied to each area belonging to the same cluster area.

Next, a time-series prediction method according to the present embodiment will be described with reference to FIG. 7. FIG. 7 is an explanatory diagram for explaining the procedure of the time series prediction method according to the third embodiment.

In the example of FIG. 7, the frequency f1 is a radio frequency (shared frequency) shared by a plurality of wireless communication systems. There are an observed area and a non-observed area for the shared frequency f1. Further, the frequency f2 is a radio frequency (LTE frequency) used by an LTE (Long Term Evolution) system which is an example of a wide area wireless communication system. The LTE frequency f2 is observed in all areas illustrated in FIG. Here, the area may be referred to as a mesh.

(Step S50) The area clustering unit 20 clusters a predetermined plurality of areas as clustering targets. In the example of FIG. 7, 25 meshes are clustering targets. Further, in the example of FIG. 7, the area clustering unit 20 has similar geographical features based on the geographical information recognized by the image recognition process from the image of the aerial photograph in which 25 meshes are captured from the sky. Cluster the meshes into the same cluster area. For example, meshes having similar density of buildings in the mesh are clustered in the same cluster area. In the example of FIG. 7, as a result of clustering, three cluster areas “first cluster”, “second cluster”, and “third cluster” are generated from 25 meshes.

(Step S51) The estimation model generation unit 12b generates an estimation model C with the cluster area “first cluster” to which the prediction target mesh (Prediction Mesh), which is the estimation area of the shared frequency f1, belongs as the model generation target area. In the generation of the estimation model C, the estimation model generation unit 12b is a mesh (reference mesh (reference mesh)) which is an observation area of the shared frequency f1 and an observation area of the LTE frequency f2 among the meshes belonging to the cluster area “first cluster”. Reference Mesh))) observation data is extracted. The extracted observation data has an observation value B of the radio wave condition information of the shared frequency f1 and an observation value B of the radio wave condition information of the LTE frequency f2.

Next, the estimation model generation unit 12 performs machine learning of a machine learning model (for example, a neural network) using the extracted observation data. In this machine learning, the relationship of the fluctuation pattern of the radio wave state information between the shared frequency f1 and the LTE frequency f2 is learned. By this machine learning, the estimation model C is configured to output the time series data of the estimated value of the radio wave condition information of the shared frequency f1 from the time series data of the observed value of the radio wave condition information of the LTE frequency f2.

(Step S52) The radio wave condition information estimation unit 13 uses the estimation model C generated in step S51 from the time series data of the observed value B of the radio wave condition information of the LTE frequency f2 of the prediction target mesh, and the prediction target. The time series data of the estimated value D of the radio wave state information of the shared frequency f1 of the mesh is acquired.

(Step S53) The time-series prediction unit 14 refers to the prediction target mesh based on the time-series data of the estimated value D of the radio wave condition information of the shared frequency f1 of the prediction target mesh acquired by the radio wave condition information estimation unit 13. , The predicted value E of the future radio wave condition information of the shared frequency f1 of the predicted target mesh is calculated. Further, the time-series prediction unit 14 predicts the prediction target mesh based on the time-series data of the observation value B of the radio wave state information of the LTE frequency f2 of the prediction target mesh acquired by the observation data acquisition unit 11. The predicted value E of the future radio wave condition information of the LTE frequency f2 of the target mesh is calculated.

The area clustering unit 20 may be applied to the time series prediction system 1a (FIG. 4) of the second embodiment described above.

In the time series prediction systems 1, 1a, 1b according to each of the above-described embodiments, the frequency band (shared frequency band) assigned to the existing wireless communication system (primary operator system) is changed to another wireless communication system (2). When the secondary operator system) uses it secondarily, it can be applied to a system that predicts the future usage status of the shared frequency band (frequency f1) of the primary operator system.

In the above-described embodiment, the received power is taken as an example of the radio wave state information, but the radio wave state information is not limited to the received power. For example, the radio wave condition information may be the number of pedestrians. The greater the number of pedestrians present in an area, the greater the likelihood of affecting the radio wave conditions in that area. Further, the number of pedestrians is information indicating the usage status of wireless communication, and it is considered that the larger the number of pedestrians existing in the area, the larger the number of wireless communication users in the area. From this, it is possible to contribute to the prediction of the usage status of the frequency f1 by using the number of pedestrians as the radio wave condition information and predicting the future number of pedestrians in each area.

Although the embodiments of the present invention have been described in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and includes design changes and the like within a range that does not deviate from the gist of the present invention.

Further, a computer program for realizing the function of the time series prediction system described above may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read by the computer system and executed. good. The "computer system" here may include hardware such as an OS and peripheral devices.
The "computer-readable recording medium" is a flexible disk, a magneto-optical disk, a ROM, a writable non-volatile memory such as a flash memory, a portable medium such as a DVD (Digital Versatile Disc), and a built-in computer system. A storage device such as a hard disk.

Further, the "computer-readable recording medium" is a volatile memory inside a computer system that serves as a server or client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line (for example, DRAM (Dynamic)). It also includes those that hold the program for a certain period of time, such as Random Access Memory)).
Further, the program may be transmitted from a computer system in which this program is stored in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting a program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.
Further, the above program may be for realizing a part of the above-mentioned functions. Further, a so-called difference file (difference program) may be used, which can realize the above-mentioned function in combination with a program already recorded in the computer system.

1,1a, 1b ... time series prediction system, 11 ... observation data acquisition unit, 12,12a, 12b ... estimation model generation unit, 13 ... radio wave condition information estimation unit, 14 ... time series prediction unit, 20 ... area clustering unit, 30 ... Radio wave condition information prediction estimation unit

Claims (7)

The first radio wave condition information indicating the radio wave condition of the first radio frequency and the second radio wave condition information indicating the radio wave condition of the second radio frequency different from the first radio frequency are observed values observed in each observation area. The observation data acquisition unit to be acquired and
An estimation model that outputs an estimated value of the first radio wave condition information from the observed value of the second radio wave condition information acquired from the observation area is obtained from the observed value of the first radio wave condition information acquired by the observation data acquisition unit. And the estimation model generation unit generated based on the observed value of the second radio wave state information,
The estimated value of the first radio wave condition information of the estimated time of the estimated area of the first radio wave condition information is obtained from the observed value of the second radio wave condition information of the estimated time of the estimated area acquired by the observation data acquisition unit. The radio wave condition information estimation unit acquired using the estimation model, and
A time-series prediction unit that calculates a future predicted value of the first radio wave condition information in the estimated area based on the time-series data of the estimated value of the first radio wave condition information in the estimated area.
Time series forecasting system. The first radio wave condition information indicating the radio wave condition of the first radio frequency and the second radio wave condition information indicating the radio wave condition of the second radio frequency different from the first radio frequency are observed values observed in each observation area. The observation data acquisition unit to be acquired and
The predicted value of the future first radio wave condition information of the observation area is calculated based on the time series data of the observation value of the first radio wave condition information of the observation area, and the second radio wave condition information of the observation area is calculated. A time-series prediction unit that calculates the predicted value of the second radio wave condition information in the future of the observation area based on the time-series data of the observed value, and
The prediction of the first radio wave condition information calculated by the time series prediction unit is an estimation model that outputs the predicted value of the first radio wave condition information in the future from the predicted value of the second radio wave condition information in the future of the observation area. An estimation model generator generated based on the value and the predicted value of the second radio wave state information,
The estimated model of the future predicted value of the first radio wave condition information in the estimated area of the first radio wave condition information from the predicted value of the future second radio wave condition information of the estimated area calculated by the time series prediction unit. Radio condition information prediction estimation unit to be acquired using
Time series forecasting system. The estimation model is generated for each cluster area as a result of clustering multiple observation areas.
The time series prediction system according to any one of claims 1 or 2. The first radio wave condition information indicating the radio wave condition of the first radio frequency and the second radio wave condition information indicating the radio wave condition of the second radio frequency different from the first radio frequency are observed values observed in each observation area. The observation data acquisition step to be acquired and
An estimation model that outputs an estimated value of the first radio wave condition information from the observed value of the second radio wave condition information acquired from the observation area is obtained as an observed value of the first radio wave condition information acquired in the observation data acquisition step. And the estimation model generation step to be generated based on the observed value of the second radio wave state information, and
The estimated value of the first radio wave condition information of the estimated time of the estimated area of the first radio wave condition information is obtained from the observed value of the second radio wave condition information of the estimated time of the estimated area acquired in the observation data acquisition step. The radio wave condition information estimation step acquired using the estimation model, and
A time-series prediction step for calculating a future predicted value of the first radio wave condition information in the estimated area based on the time-series data of the estimated value of the first radio wave condition information in the estimated area.
Time series forecasting method including. The first radio wave condition information indicating the radio wave condition of the first radio frequency and the second radio wave condition information indicating the radio wave condition of the second radio frequency different from the first radio frequency are observed values observed in each observation area. The observation data acquisition step to be acquired and
The predicted value of the future first radio wave condition information of the observation area is calculated based on the time series data of the observation value of the first radio wave condition information of the observation area, and the second radio wave condition information of the observation area is calculated. A time-series prediction step for calculating the predicted value of the second radio wave condition information in the future of the observation area based on the time-series data of the observed value, and
Prediction of the first radio wave condition information calculated in the time series prediction step by an estimation model that outputs the predicted value of the first radio wave condition information in the future from the predicted value of the second radio wave condition information in the future of the observation area. An estimation model generation step to be generated based on the value and the predicted value of the second radio wave state information, and
The estimated model of the future predicted value of the first radio wave condition information in the estimated area of the first radio wave condition information from the predicted value of the future second radio wave condition information of the estimated area calculated in the time series prediction step. The radio wave condition information prediction estimation step to be acquired using
Time series forecasting method including. On the computer
The first radio wave condition information indicating the radio wave condition of the first radio frequency and the second radio wave condition information indicating the radio wave condition of the second radio frequency different from the first radio frequency are observed values observed in each observation area. The observation data acquisition step to be acquired and
An estimation model that outputs an estimated value of the first radio wave condition information from the observed value of the second radio wave condition information acquired from the observation area is obtained as an observed value of the first radio wave condition information acquired in the observation data acquisition step. And the estimation model generation step to be generated based on the observed value of the second radio wave state information, and
The estimated value of the first radio wave condition information of the estimated time of the estimated area of the first radio wave condition information is obtained from the observed value of the second radio wave condition information of the estimated time of the estimated area acquired in the observation data acquisition step. The radio wave condition information estimation step acquired using the estimation model, and
A time-series prediction step for calculating a future predicted value of the first radio wave condition information in the estimated area based on the time-series data of the estimated value of the first radio wave condition information in the estimated area.
A computer program to run. On the computer
The first radio wave condition information indicating the radio wave condition of the first radio frequency and the second radio wave condition information indicating the radio wave condition of the second radio frequency different from the first radio frequency are observed values observed in each observation area. The observation data acquisition step to be acquired and
The predicted value of the future first radio wave condition information of the observation area is calculated based on the time series data of the observation value of the first radio wave condition information of the observation area, and the second radio wave condition information of the observation area is calculated. A time-series prediction step for calculating the predicted value of the second radio wave condition information in the future of the observation area based on the time-series data of the observed value, and
Prediction of the first radio wave condition information calculated in the time series prediction step by an estimation model that outputs the predicted value of the first radio wave condition information in the future from the predicted value of the second radio wave condition information in the future of the observation area. An estimation model generation step to be generated based on the value and the predicted value of the second radio wave state information, and
The estimated model of the future predicted value of the first radio wave condition information in the estimated area of the first radio wave condition information from the predicted value of the future second radio wave condition information of the estimated area calculated in the time series prediction step. The radio wave condition information prediction estimation step to be acquired using
A computer program to run.

Images