JP2018132802A - Learning apparatus, attribute estimation apparatus, learning method, attribute estimation method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate attribute information on a family based on information indicating a usage power amount of the family.SOLUTION: A learning apparatus includes: an acquisition unit for acquiring information indicating each of power usage amounts of a plurality of families and attribute information on the families of which the power usage amounts have been acquired; a learning unit for machine-learning the information indicating each of the power usage amounts of the plurality of families acquired by the acquisition unit and associating feature amounts of the power usage amounts acquired by machine leaning with the attribute information on the families of which the power usage amounts have been acquired; and a storage unit for storing the feature amounts of the power usage amounts associated by the learning unit and the attribute information on the families of which the power usage amount have been acquired in association with each other.SELECTED DRAWING: Figure 3

Description

  Embodiments described herein relate generally to a learning device, an attribute estimation device, a learning method, an attribute estimation method, and a program.

Consider developing a business that provides new services and providing it to customers. When developing a business, it is necessary to consider the contents of the service in accordance with the needs of the consumers based on the attribute information of the consumers such as the family structure of the consumers such as customers, lifestyles, and lifestyles. Since attribute information such as lifestyle is often different for each consumer, it is difficult to correctly grasp them by data analysis or the like.
Therefore, it is performed to grasp the attribute information of the consumer by using the power usage data of the consumer who is relatively directly connected to the life of the consumer. Specifically, hypothesis verification is performed on power usage data based on empirical rules each time, and consumers are typified by numerical analysis techniques such as clustering.

  As a technique for analyzing data, a technique is known in which a process item is specified as a factor that causes variation in the distribution of characteristic values for a product having a spatial spread (see, for example, Patent Document 1). According to this technology, based on the characteristic distribution data of a plurality of products, a plurality of components representing the characteristics of fluctuations in the characteristic distribution data are extracted by the pattern classification method, and for each of the characteristic distribution data, A weighting factor that contributes to the plurality of components representing the characteristic distribution data is calculated. Then, based on the calculated weighting coefficient, at least one component to be analyzed is selected from the plurality of components. Then, based on the weighting coefficient relating to the selected component and the manufacturing data acquired for each process item, a process item as a variation occurrence factor in the characteristic distribution data is extracted.

JP 2012-63928 A

The technology described above has the following problems.
When examining the contents of a service in accordance with the needs of consumers, it is necessary to repeat or redo various analyzes according to the needs of consumers.
Further, when hypothesis verification is performed on power usage data based on an empirical rule each time, it is necessary to make a hypothesis and verify the hypothesis. Hypothesis planning depends heavily on the experience and skills of analysts. In addition, applications necessary for providing services basically need to be developed individually.
The present invention has been made in view of the above points, and an object of the present invention is to provide home attribute information based on information indicating the power consumption of the home.

One aspect of the present invention is an acquisition unit that acquires information indicating power usage of each of a plurality of households and home attribute information from which the power usage is obtained, and a plurality of households acquired by the acquisition unit. Machine learning information indicating each power usage amount, and a learning unit associating the characteristic amount of the power usage amount obtained by the machine learning and the attribute information of the home from which the power usage amount was obtained, The learning apparatus includes: a storage unit that stores the characteristic amount of the power usage amount associated with the learning unit and the attribute information of the home from which the power usage amount is obtained in association with each other.
In the learning device according to an aspect of the present invention, the information indicating the power usage amount acquired by the acquisition unit represents the magnitude of the power usage amount by a color difference.
In the learning device according to an aspect of the present invention, the information indicating the power usage amount acquired by the acquisition unit further represents the concept of time by a color difference.
In the learning device according to one aspect of the present invention, the information indicating the power usage acquired by the acquisition unit includes power usage and information indicating an environmental condition measured in the home from which the power usage was obtained. It is expressed by the difference in color.
In the learning device according to one aspect of the present invention, among the information indicating the power usage of each of the plurality of homes acquired by the acquisition unit, the usage of power from each of the power usages of homes with similar attribute information One or a plurality of usage-specific power usage indicating another usage is acquired, and based on the acquired one or a plurality of usage-specific power usage, a creation unit that newly creates information indicating the power usage is provided. The learning unit machine learns information indicating the newly generated power usage amount by the creation unit, and the power usage amount obtained by machine learning information indicating the newly created power usage amount Are associated with the similar attribute information.
One aspect of the present invention provides an acquisition unit that acquires information indicating power consumption of a home whose attribute information is unknown, and machine learning of the information indicating the power usage acquired by the acquisition unit. A feature amount of the power usage amount and the power usage amount obtained by machine learning of attribute information related to the feature amount, information indicating the power usage amount of each of a plurality of households is obtained. It is an attribute estimation apparatus provided with the estimation part which estimates from the information linked | related with the obtained household attribute information.
In the attribute estimation device according to one aspect of the present invention, the information indicating the power usage amount of the home whose attribute information acquired by the acquisition unit is unknown is a representation of the magnitude of the power usage amount by a color difference.
In the attribute estimation device according to one aspect of the present invention, the information indicating the power usage amount of the home whose attribute information acquired by the acquisition unit is unknown further represents the concept of time with a color difference.
In the attribute estimation device according to one aspect of the present invention, the information indicating the power usage amount of a household whose attribute information acquired by the acquisition unit is unknown is measured in the household where the power usage amount and the power usage amount are obtained. The information indicating the environmental status is expressed by the difference in color.
In the learning method of one aspect of the present invention, a step of acquiring information indicating power usage of each of a plurality of homes and home attribute information from which the power usage is obtained, and the plurality of acquired in the acquiring step A step of machine learning information indicating the amount of power used in each household, and a step of associating the characteristic amount of the power usage obtained by the machine learning with the attribute information of the home from which the power usage is obtained And a step of associating and storing the characteristic amount of the power usage associated in the associating step and the attribute information of the home from which the power usage was obtained.
In the attribute estimation method according to one aspect of the present invention, acquiring information indicating a household power usage whose attribute information is unknown, and machine learning the information indicating the power usage acquired in the acquiring step The feature amount of the power usage obtained by machine learning of the information indicating the power usage of each of a plurality of households, and the attribute information related to the feature amount obtained from the step of obtaining the feature amount of the power usage by And an estimation step based on information associating the attribute information of the home from which the amount of power used is obtained.
One embodiment of the present invention includes a step of acquiring information indicating the power usage amount of each of a plurality of homes and home attribute information from which the power usage amount is obtained in a computer included in the learning device, and the acquiring step. The machine learning of the information indicating the obtained power usage of each of the plurality of homes, the characteristic amount of the power usage obtained by the machine learning, and the home attribute information obtained from the power usage , And a step of associating and storing the characteristic amount of the power usage associated in the associating step and the attribute information of the home from which the power usage was obtained.
According to one aspect of the present invention, a computer included in an attribute estimation apparatus includes a step of acquiring information indicating a household power usage whose attribute information is unknown, and information indicating the power usage acquired in the acquiring step. Obtaining the feature amount of the power usage amount by machine learning, and attribute information related to the feature amount, the power usage obtained by machine learning information indicating the power usage amount of each of a plurality of households This is a program for executing an estimation step based on information associating the feature amount of the quantity with the attribute information of the home from which the power consumption is obtained.

  According to the embodiment of the present invention, home attribute information can be provided based on information indicating the amount of power used at home.

It is a figure which shows an example (the 1) of the information which shows electric power consumption. It is a figure which shows an example (the 2) of the information which shows electric power consumption. It is a figure which shows an example of the hardware constitutions of the learning apparatus which concerns on embodiment. It is a figure which shows an example of an attribute information estimation table. It is a functional block diagram of the learning device concerning an embodiment. It is a figure which shows an example of the heat map of the information which shows the electric power usage of each of a some household. It is a flowchart which shows an example of operation | movement of a learning apparatus. It is a functional block diagram of the learning device concerning a modification. It is a figure which shows the creation example of the information which shows electric power consumption. It is a flowchart which shows an example of operation | movement of the learning apparatus which concerns on a modification. It is a figure which shows an example of the hardware constitutions of the attribute estimation apparatus which concerns on embodiment. It is a functional block diagram of the attribute estimation apparatus which concerns on embodiment. It is a flowchart which shows an example of operation | movement of an attribute estimation apparatus. It is a figure which shows an example of operation | movement of the attribute estimation system which concerns on embodiment. It is a figure which shows an example (the 1) of an estimation precision of the attribute information which the attribute estimation apparatus which concerns on embodiment estimates. It is a figure which shows an example (the 2) of the estimation accuracy of the attribute information which the attribute estimation apparatus which concerns on embodiment estimates. It is a figure which shows an example (the 3) of the estimation precision of the attribute information which the attribute estimation apparatus which concerns on embodiment estimates. It is a figure which shows an example of the heat map of the information which shows the electric power usage of each of a some household. It is a figure which shows an example (the 4) of the estimation precision of the attribute information which the attribute estimation apparatus which concerns on embodiment estimates.

Next, modes for carrying out the present invention will be described with reference to the drawings. Embodiment described below is only an example and embodiment to which this invention is applied is not restricted to the following embodiment.
Note that components having the same function are denoted by the same reference symbols throughout the drawings for describing the embodiments, and the repetitive description will be omitted.

(First embodiment)
(Learning device)
A learning apparatus according to an embodiment will be described. The learning device according to the embodiment acquires information indicating the power usage amount of each of a plurality of homes. Here, the information indicating the amount of power used represents the amount of power used, such as a heat map, in different colors.
Furthermore, the learning device according to the embodiment acquires home attribute information from which information indicating the power usage of each of the acquired homes is obtained. The learning device according to the embodiment machine learns information indicating the power usage of each of a plurality of homes, and features of the power usage obtained by machine learning and home attribute information from which the power usage is obtained Associate with. The learning device according to the embodiment stores the feature amount of the power usage amount and the home attribute information from which the power usage amount is obtained in association with each other.

FIG. 1 is a diagram illustrating an example (part 1) of information indicating power usage. In FIG. 1, the horizontal axis indicates time, and the vertical axis indicates date and time. FIG. 1 shows a heat map of information indicating the amount of power used acquired from a home energy management system (HEMS) installed in each of a plurality of buildings. In the example shown in FIG. 1, the amount of power usage is graphically represented in time series.
According to FIG. 1, it can be seen that the power consumption increases between 6: 00-18: 00 and after 18:00. Further, it can be seen that the amount of power used is small during the Golden Week period displayed as “GW”. As shown in FIG. 1, the amount of power usage can be visualized by making the amount of power usage graphically in time series. For this reason, life dynamics can be grasped intuitively.

FIG. 2 is a diagram illustrating an example (part 2) of information indicating power usage. In FIG. 2, the horizontal axis indicates the outside air temperature, and the vertical axis indicates the power consumption.
According to FIG. 2, it can be seen that the power consumption increases as the outside air temperature decreases, while the power consumption increases as the outside air temperature increases. By referring to the information indicating the power consumption shown in FIG. 2, it is possible to estimate the performance of an air conditioner such as a heater, the heat insulation of a house, and the like.

(Configuration of learning device)
FIG. 3 is a diagram illustrating an example of a hardware configuration of the learning device according to the embodiment.
The learning apparatus 100 includes a CPU (Central Processing Unit) 102, a ROM (Read Only Memory) 104, a RAM (Random Access Memory) 106, a nonvolatile memory 108, a communication I / F 110, an operation unit 112, and a display. A section 114 and a bus 116 are provided.
The CPU 102 reads a program 1082 and the like for controlling the operation of the learning device 100 from the nonvolatile memory 108, expands the program in the RAM 106, and executes the program. The non-volatile memory 108 includes a flash memory, an HDD (Hard Disc Drive), an SSD (Solid State Drive), an SD (Secure Digital) card, and the like. The nonvolatile memory 108 stores a program 1082 executed by the CPU 102, power usage information 1084, attribute information 1086, and an attribute information estimation table 1088. The power usage information 1084 is information indicating the power usage of each of a plurality of homes, and represents the magnitude of the power usage, such as a heat map, with a color difference. The attribute information 1086 is information indicating home attributes such as a wake-up time, the number of family members, a dwelling form, the presence of elderly people, the presence of infants, the presence of pets, absence time, and vacation acquisition information. Information indicating the power usage of each of a plurality of homes included in the power usage information 1084 and information indicating the home attributes included in the attribute information 1086 are associated with each other. The attribute information estimation table 1088 will be described later.

The communication I / F 110 includes communication modules such as an NFC (Near Field Radio Communication) module and a wireless LAN module. Further, the communication I / F 110 may be configured by a USB (Universal Serial Bus). The communication I / F 110 performs communication with an external device by NFC or wireless LAN. The communication I / F 110 is connected to a smart meter (not shown) and an attribute estimation device 300 described later.
The operation unit 112 is an input device that receives a user operation, and includes a pointing device such as a touch panel, buttons, a dial, a touch sensor, a touch pad, and the like.
The display unit 114 is configured by, for example, a liquid crystal display or the like, and displays information indicating the power usage amount, home attribute information from which the power usage amount is obtained, and the like.

(Attribute information estimation table)
FIG. 4 is a diagram illustrating an example of the attribute information estimation table. The attribute information estimation table 1088 is data in a table format in which the feature amount of the power usage amount and the attribute information are associated with each other. The characteristic amount of the power usage is obtained by machine learning such as deep learning for information indicating the power usage of each of a plurality of households. The attribute information is attribute information of a household from which power usage is obtained. In the example illustrated in FIG. 4, the characteristic amount “AAA” of the power usage amount and the home attribute information “aaa” from which the power usage amount is obtained are associated with each other.

(Functional configuration of learning device)
FIG. 5 is a functional block diagram of the learning device according to the embodiment.
The learning device 100 functions as an acquisition unit 152, a learning unit 154, and an output unit 156 when the CPU 102 executes a program 1082 expanded from the nonvolatile memory 108 onto the RAM 106.

(Each functional configuration of the learning device)
Each functional configuration of the learning apparatus 100 will be described in detail with reference to FIGS. 3 and 5. In the following, in describing each function of the learning device 100, the relationship with main hardware for realizing each functional block of the learning device 100 shown in FIG. 5 will also be described.
The acquisition unit 152 is realized by a command from the CPU 102 and the communication I / F 110. The acquisition part 152 acquires the information which shows the electric power consumption which the smart meter installed in each of several households transmitted. Furthermore, the acquisition unit 152 acquires household attribute information from which information indicating the power usage is obtained. When the acquisition unit 152 acquires the information indicating the power usage of each of the plurality of homes and the attribute information of the home from which the power usage is obtained, the information indicating the power usage of each of the plurality of homes is stored in the nonvolatile memory. It outputs to the power usage information 1084 of 108, and outputs the attribute information of the home from which the power usage was obtained to the attribute information 1086 of the nonvolatile memory 108.

The learning unit 154 is realized by a command from the CPU 102. The learning unit 154 acquires information indicating the power usage of each of the plurality of homes from the power usage information 1084 of the nonvolatile memory 108, and acquires the home attribute information from which the power usage is obtained from the attribute information 1086. . When the learning unit 154 acquires information indicating the power usage of each of the plurality of homes and the attribute information of the home from which the power usage is obtained, the learning unit 154 performs information such as deep learning on the power usage of each of the plurality of homes. Perform machine learning.
The learning unit 154 sends the feature amount of the power usage obtained by deep learning of the information indicating the power usage of each of the plurality of homes and the attribute information of the home from which the power usage is obtained to the output unit 156. Output.

  The output unit 156 is realized by a command from the CPU 102. When the output unit 156 acquires the feature amount of the power usage amount output from the learning unit 154 and the home attribute information from which the power usage amount is obtained, the feature amount and power usage amount of the acquired power usage amount are obtained. The attribute information is stored in the attribute information estimation table 1088 of the nonvolatile memory 108 in association with the household attribute information.

(Example of information indicating power consumption)
FIG. 6 is a diagram illustrating an example of a heat map of information indicating the power usage of each of a plurality of households. Information indicating the power usage of each of the plurality of homes is represented by a heat map. The heat map includes a heat map 10 and a heat map 20.
The heat map 10 is information indicating the power usage amount of each of a plurality of households from January to June. The heat map 20 is information indicating the power usage amount of each of a plurality of households from July to December. In FIG. 6, the horizontal axis indicates time 0:00 to 24:00, and the vertical axis indicates the month. That is, in the example of the heat map of the information indicating the power usage shown in FIG. 6, the magnitude of the power usage is represented by the difference in color, and the concept of time is represented by the difference in color.

(Operation of learning device)
FIG. 7 is a flowchart illustrating an example of the operation of the learning device.
(Step S702) The acquisition unit 152 of the learning device 100 acquires information indicating the power usage amount transmitted by the smart meter installed in each of a plurality of homes. When acquiring the information indicating the power usage of each of the plurality of homes, the acquiring unit 152 outputs the information indicating the power usage to the power usage information 1084 of the nonvolatile memory 108. The power usage information 1084 of the nonvolatile memory 108 stores information indicating the power usage of each of the plurality of homes output from the acquisition unit 152.
(Step S704) When the acquisition unit 152 of the learning device 100 acquires information indicating the power usage of each of the plurality of homes, the acquisition unit 152 acquires the home attribute information from which the information indicating the power usage is obtained. When acquiring the attribute information from which the information indicating the power usage is obtained, the acquiring unit 152 outputs the attribute information to the attribute information 1086 of the nonvolatile memory 108. The attribute information 1086 of the nonvolatile memory 108 stores the attribute information output from the acquisition unit 152.

(Step S706) The learning unit 154 of the learning device 100 acquires information indicating the power usage of each of the plurality of homes from the power usage information 1084 of the nonvolatile memory 108, and the power usage is obtained from the attribute information 1086. Get home attribute information. When the learning unit 154 acquires the information indicating the power usage of each of the plurality of homes and the attribute information obtained from the power usage, the learning unit 154 performs deep learning on the information indicating the power usage of each of the plurality of homes. Do.
(Step S708) When the learning unit 154 of the learning device 100 performs deep learning on the information indicating the power usage of each of a plurality of homes, the feature amount and power of the power usage obtained by performing the deep learning Associate it with the attribute information of the household from which the usage amount was obtained. When the learning unit 154 associates the feature amount of the power usage amount with the home attribute information from which the power usage amount is obtained, the learning unit 154 associates the feature amount of the power usage amount with the home attribute information from which the power usage amount is obtained. The data is output to the output unit 156.
(Step S <b> 710) When the output unit 156 of the learning device 100 acquires the information that associates the feature value of the power usage amount output by the learning unit 154 with the home attribute information from which the power usage amount is obtained, the power usage amount Are associated with the attribute information of the home from which the power consumption is obtained and output to the attribute information estimation table 1088 of the nonvolatile memory 108. The attribute information estimation table 1088 of the nonvolatile memory 108 stores the feature value of the power usage amount output from the output unit 156 and the home attribute information from which the power usage amount is obtained in association with each other.
In the flowchart shown in FIG. 7, the order of step S702 and step S704 may be switched.

  In the above-described embodiment, the case where the learning apparatus 100 acquires information indicating the power usage amount transmitted by the smart meter installed in each of a plurality of homes has been described, but the present invention is not limited to this example. For example, when the information indicating the power usage output from the smart meter is not a heat map, the information indicating the power usage output from the smart meter may be output to a conversion device that converts the information into a heat map. When the conversion device acquires information indicating the power usage amount output from the smart meter, the conversion device converts the acquired information indicating the power usage amount into a heat map. The conversion device may output the information indicating the heat map to the learning device 100 when the information indicating the power consumption is converted into the heat map.

In the above-described embodiment, the case where the learning apparatus 100 includes the nonvolatile memory 108 in which the attribute information estimation table 1088 is stored has been described, but the present invention is not limited to this example. For example, the attribute information estimation table 1088 may be stored on the cloud. In this case, the learning apparatus 100 stores the attribute amount estimation table 1088 on the cloud in association with the feature amount of the power usage amount and the home attribute information from which the power usage amount is obtained.
In the above-described embodiment, the case where the learning apparatus 100 acquires information indicating the power usage amount transmitted by the smart meter installed in each of a plurality of homes has been described, but the present invention is not limited to this example. For example, the learning apparatus 100 may acquire information indicating the power usage amount that represents the power usage amount and the information indicating the environmental state measured in the home from which the power usage amount is obtained with a color difference. Good. Here, the information indicating the environmental condition measured at home includes information indicating the outside air temperature, information indicating the temperature of the living room, information indicating a temperature difference between the living room and the outside air, information indicating the humidity, and the like. Specifically, the amount of power used and information indicating the environmental state measured at the home from which the amount of power used is obtained are represented by a heat map. Then, the learning apparatus 100 may acquire home attribute information from which information indicating the power usage of each of the acquired homes is obtained. The learning device performs machine learning on the information indicating the power usage of each of a plurality of households, and associates the characteristic amount of the power usage obtained by the machine learning with the attribute information of the home from which the power usage is obtained. It may be. The learning device may store the feature amount of the power usage amount and the attribute information of the home from which the power usage amount is obtained in association with each other.

  According to the learning device 100 according to the present embodiment, the feature amount of the power usage obtained by deep learning the information indicating the power usage in the heat map format is associated with the attribute information obtained from the power usage. Remember. Because the heat map is “visualized”, it is possible to intuitively grasp the life dynamics. In addition, by associating the feature value of the power usage amount obtained by deep learning of the information indicating the power usage amount in the heat map format with the attribute information of the home from which the power usage amount was obtained, the feature amount of the power usage amount Therefore, it is possible to estimate attribute information such as owned facilities, housing performance, and life dynamics with high accuracy. By performing machine learning, features that are common among information indicating power usage are automatically extracted. Therefore, compared to the case where an analyst analyzes information indicating power usage, it is at a deeper level. Analysis becomes possible. In addition, by performing machine learning, any kind of attribute information can be estimated by a similar method.

(Modification)
A learning apparatus according to a modification will be described. The learning apparatus 200 according to the modified example has the same attribute information from which the power usage is obtained among the information indicating the power usage of each of the plurality of homes stored in the power usage information 1084 of the nonvolatile memory 108. Or the information which shows the electric power consumption of the similar household is extracted. The learning device 200 uses the power usage amount of the air conditioner, the power usage amount of the water heater, the power usage amount of the cooking area, and the like from each of the information indicating the power usage amount of the home where the extracted attribute information is the same or similar. Acquire the power consumption of the air conditioner, the power consumption of the water heater, and the power consumption other than the power consumption of the cooking area (other power consumption). Hereinafter, each of the power consumption of the air conditioner, the power consumption of the water heater, the power consumption of the cooking area, and the other power consumption is indicated by the usage amount of the power according to the usage. Call. That is, the learning device 200 acquires one or a plurality of usage-specific power usage amounts from each of the pieces of information indicating household power usage amounts with the same or similar extracted attribute information.

The learning apparatus 200 newly creates information indicating the power usage amount from the acquired one or more usage-specific power usage amounts. The learning device 200 outputs information indicating the newly created power usage amount to the power usage amount information 1084 of the nonvolatile memory 108. Furthermore, the learning device 200 stores similar household attribute information in the attribute information 1086 of the nonvolatile memory 108 in association with the newly created information indicating the power consumption.
The learning device 200 deep learns information indicating the newly created power consumption. When the learning apparatus 200 performs deep learning on the information indicating the newly generated power usage amount, the non-volatile memory 108 associates the feature amount of the power usage amount obtained by the deep learning with the similar household attribute information. Is stored in the attribute information estimation table 1088.

As an example of the hardware configuration of the learning apparatus 200 according to the modification, FIG. 2 can be applied.
<Functional configuration of learning device>
FIG. 8 is a functional block diagram of a learning device according to a modification.
The learning apparatus 200 functions as an acquisition unit 252, a learning unit 254, an output unit 256, and a creation unit 258 when the CPU 102 executes a program 1082 expanded from the nonvolatile memory 108 onto the RAM 106.

<Each functional configuration of the learning device>
Each functional configuration of the learning apparatus 200 will be described in detail with reference to FIGS. 2 and 8. In the following, in describing each function of the learning apparatus 200, a relationship with main hardware for realizing each functional block of the learning apparatus 200 shown in FIG. 8 will also be described.
The acquisition unit 252, the learning unit 254, and the output unit 256 can apply the acquisition unit 152, the learning unit 254, and the output unit 256 of the learning device 100, respectively.

The creation unit 268 is realized by a command from the CPU 102 and the operation unit 112. The creation unit 268 functions when the operation unit 112 is operated to increase information indicating the power consumption. When the operation unit 112 is operated to increase the information indicating the power usage, the creation unit 268 performs the power usage of each of the plurality of homes stored in the power usage information 1084 of the nonvolatile memory 108. Among the information indicating the information indicating the power consumption of the home with the same or similar attribute information. Specifically, the creation unit 268 extracts information indicating the power usage of each of a plurality of households having the same or similar household numbers, living forms, floor areas, building ages, and the like.
The creation unit 268 acquires one or more usage-specific power usage amounts from each piece of information indicating the extracted power usage amounts. Specifically, the creation unit 268 includes, from each of the information indicating the power usage, a pattern of information indicating the power usage of the air conditioner, a pattern of information indicating the power usage of the water heater, and the power of the cooking area The information pattern indicating the usage amount and the information pattern indicating the other power usage amount are acquired.

The creation unit 268 newly creates information indicating the power usage of households with similar attributes by combining the acquired one or more usage-specific power usage in a permutation combination.
The creation unit 268 stores information indicating the newly created power usage amount in the power usage amount information 1084 of the nonvolatile memory 108. Furthermore, the creation unit 268 stores similar household attribute information in the attribute information 1086 of the nonvolatile memory 108 in association with the newly created information indicating the power usage.

(Example of creating a heat map)
As described above, the creation unit 268 acquires one or a plurality of usage-specific power usage amounts from each piece of information indicating household power usage amounts having the same or similar attribute information. Then, the creation unit 268 newly creates information indicating the power usage of households with similar attributes by combining the acquired one or more usage-specific power usage with permutation combinations.
FIG. 9 is a diagram illustrating a creation example of information indicating the power usage amount. In the example shown in FIG. 9, it is assumed that there are two households (A, B) having the same or similar attribute information. The creation unit 268 acquires one or a plurality of usage-specific power usage amounts from the information A indicating the power usage amount. In addition, the creation unit 268 acquires one or a plurality of usage-specific power usage amounts from the information B indicating the power usage amount.
The creation unit 268 uses one or more usage-specific power usage amounts acquired from the information A indicating the power usage amount and one or more usage-specific power usage amounts acquired from the information B indicating the power usage amount. By combining and combining, information ABA indicating power usage, information AAB indicating power usage, information ABB indicating power usage, information BAB indicating power usage, information BAA indicating power usage and power usage Information BBA is newly created. That is, 2 (other power consumption) × 2 (power consumption of the air conditioner) × 2 (power consumption of the water heater) × 2 (power consumption of the kitchen) = 8, and the creation unit 268 Information indicating six power consumptions can be newly created.

(Operation of learning device)
FIG. 10 is a flowchart illustrating an example of the operation of the learning device according to the modification. FIG. 10 mainly shows the operation of the creation unit 268 of the learning device 200.
(Step S1002) The creation unit 268 of the learning device 200 has obtained the power usage amount among the information indicating the power usage amounts of the plurality of homes stored in the power usage amount information 1084 of the nonvolatile memory 108. Information indicating the amount of power usage with the same or similar household attribute information is extracted.
(Step S1004) The creation unit 268 of the learning device 200 acquires information indicating the power usage of the air conditioner from each of the information indicating the power usage of the home with the same or similar attribute information.
(Step S1006) The creation unit 268 of the learning device 200 acquires information indicating the power usage amount of the water heater from each of the information indicating the power usage amount of the home having the same or similar attribute information.

(Step S1008) The creation unit 268 of the learning device 200 acquires information indicating the power usage amount of the kitchen from each of the information indicating the power usage amount of the home having the same or similar attribute information.
(Step S1010) The creation unit 268 of the learning device 200 acquires information indicating the other power usage from each of the information indicating the power usage of the home having the same or similar attribute information.
(Step S1012) The creation unit 268 of the learning device 200 acquires the information indicating the acquired power usage of the air conditioner, the information indicating the power usage of the water heater, the information indicating the power usage of the cooking area, and the other power usage. Is combined with permutation information to create information indicating power consumption.
(Step S1014) The creation unit 268 of the learning device 200 outputs information indicating the newly created power usage amount to the power usage amount information 1084 of the nonvolatile memory 108. Furthermore, the learning device 200 stores similar household attribute information in the attribute information 1086 of the nonvolatile memory 108 in association with the newly created information indicating the power consumption.

In the above-described modification, from each piece of information indicating household power usage with the same or similar attribute information, the power usage of the air conditioner, the power usage of the water heater, the power usage of the kitchen, and other However, the present invention is not limited to this example. For example, from each piece of information indicating household power usage with the same or similar attribute information, the power usage of the air conditioner other than the power usage of the kitchen, the power usage of the water heater, and other power usage May be acquired, or the power usage amount of the living room or the power usage amount of the entrance may be acquired.
In the above-described modification, the case where it is assumed that there are two households (A, B) having the same or similar attribute information has been described, but the present invention is not limited to this example. For example, the present invention can be applied to a case where there are three homes (A, B, C) having the same or similar attribute information. In this case, the creation unit 268 obtains three power usage amounts by use (power consumption amount of the air conditioner, power consumption amount of the water heater, and other power usage amounts) from the information A indicating the power usage amount. In addition, the creation unit 268 acquires three power usage amounts for each application (power usage amount of the air conditioner, power usage amount of the water heater, and other power usage amounts) from the information B indicating the power usage amount. In addition, the creation unit 268 acquires three power usage amounts for each application (the power usage amount of the air conditioner, the power usage amount of the water heater, and other power usage amounts) from the information C indicating the power usage amount.
The creation unit 268 indicates the three power usage amounts by usage acquired from the information A indicating the power usage amount, the three power usage amounts by usage acquired from the information B indicating the power usage amount, and the power usage amount. By combining the three usage-specific power usage amounts acquired from the information C in combination with the brute force, information indicating 27 power usage amounts is newly created. That is, 3 (power consumption of the air conditioner) × 3 (power consumption of the water heater) × 3 (other power usage) = 27, and the creation unit 268 displays information indicating the 24 power usages. Newly created.

  According to the learning device 200 according to the modified example, when an operation for increasing the information indicating the power usage is performed, the power is newly generated from each of the information indicating the power usage of the home having the same or similar attribute information. Information indicating the amount used can be created. That is, even when the number of information indicating the power usage obtained from the smart meter is small, the information indicating the power usage can be increased. Since the information indicating the power usage amount can be increased, the number of feature amounts of the power usage amount can also be increased. For this reason, it is possible to increase the number of attribute information that can be estimated from the feature amount of the power consumption.

(Second Embodiment)
(Attribute estimation device)
An attribute estimation apparatus according to an embodiment will be described. The attribute estimation apparatus 300 according to the embodiment acquires information indicating the power usage amount of a home whose attribute information is unknown. Here, the information indicating the power usage amount of the home whose attribute information is unknown represents the magnitude of the power usage amount, such as a heat map, by a color difference.
Furthermore, the attribute estimation device 300 according to the embodiment performs machine learning on information indicating the power usage amount of a home whose acquired attribute information is unknown, thereby obtaining a feature amount of the power usage amount of the home whose attribute information is unknown. Ask. The attribute estimation device 300 according to the embodiment acquires information that associates the feature amount of the power usage amount with the home attribute information from which the power usage amount is obtained from the learning device 100 described above. The attribute estimation apparatus 300 according to the embodiment uses a power consumption amount of a home whose attribute information is unknown based on information obtained by associating the acquired feature value of the power usage amount with the attribute information of the home from which the power usage amount is obtained. Attribute information related to the feature amount is estimated.

(Configuration of attribute estimation device)
FIG. 11 is a diagram illustrating an example of a hardware configuration of the attribute estimation apparatus according to the embodiment.
The attribute estimation apparatus 300 includes a CPU 302, a ROM 304, a RAM 306, a nonvolatile memory 308, a communication I / F 310, an operation unit 312, a display unit 314, and a bus 316.
The CPU 302 reads out a program 3082 and the like for controlling the operation of the attribute estimation apparatus 300 from the nonvolatile memory 308, expands it in the RAM 306, and executes it. The nonvolatile memory 308 is configured by a flash memory, an HDD, an SSD, an SD card, or the like. The nonvolatile memory 308 stores a program 3082 executed by the CPU 302 and an attribute information estimation table 3088. The attribute information estimation table 3088 can be applied to the attribute information estimation table 1088 described above.
The communication I / F 310 includes a communication module such as an NFC module or a wireless LAN module. Further, the communication I / F 310 may be configured by USB. The communication I / F 310 performs communication with an external device by NFC or wireless LAN. The communication I / F 310 is connected to a smart meter (not shown) and an attribute estimation device 300 described later.
The operation unit 312 is an input device that receives a user operation, and includes a pointing device such as a touch panel, buttons, a dial, a touch sensor, and a touch pad.
The display unit 314 is configured by, for example, a liquid crystal display and displays home attribute information and the like from which power consumption is obtained.

(Functional configuration of attribute estimation device)
FIG. 12 is a functional block diagram of the attribute estimation apparatus according to the embodiment.
The attribute estimation apparatus 300 functions as an acquisition unit 352, a learning unit 354, an estimation unit 356, and an output unit 358 when the CPU 302 executes a program 3082 expanded from the nonvolatile memory 308 onto the RAM 306.

(Each functional configuration of the attribute estimation device)
Each functional configuration of the attribute estimation apparatus 300 will be described in detail with reference to FIGS. 11 and 12. In the following, in describing each function of the attribute estimation apparatus 300, the relationship with main hardware for realizing each functional block of the attribute estimation apparatus 300 shown in FIG. 12 will also be described.
The acquisition unit 352 is realized by a command from the CPU 302 and the communication I / F 310. The acquisition unit 352 acquires information indicating the power usage of a home whose attribute information is unknown. Here, the information indicating the power usage amount of the home whose attribute information is unknown represents the magnitude of the power usage amount, such as a heat map, by a color difference. When the acquisition unit 352 acquires information indicating the power usage of a household whose attribute information is unknown, the acquisition unit 352 outputs information indicating the power usage to the learning unit 354.
The learning unit 354 is realized by a command from the CPU 302. When the learning unit 354 acquires information indicating the power usage amount of the home whose attribute information output by the acquisition unit 352 is unknown, deep learning is performed on the information indicating the power usage amount of the home whose attribute information is unknown. I do. When the learning unit 354 performs deep learning on the information indicating the power usage of a home whose attribute information is unknown, the learning unit 354 outputs a feature value of the power usage obtained by performing the deep learning to the estimation unit 356. .

The estimation unit 356 acquires the attribute information estimation table 1088 in which the feature amount of the power usage amount and the attribute information of the power usage amount are associated from the learning device 100 via the communication I / F 310. Upon obtaining the attribute information estimation table 1088, the estimation unit 356 stores the attribute information estimation table 1088 in the attribute information estimation table 3088 of the nonvolatile memory 308.
When the estimation unit 356 obtains the feature amount of the household power consumption whose attribute information output from the learning unit 354 is unknown, the estimation unit 356 refers to the attribute information estimation table 3088 and attributes related to the acquired feature amount of the power usage amount Estimate information. When estimating the attribute information, the estimation unit 356 outputs the attribute information estimation result to the output unit 358.
The output unit 358 is realized by a command from the CPU 302. When the output unit 358 acquires the attribute information estimation result output by the estimation unit 356, the output unit 358 outputs the acquired attribute information estimation result to the display unit 314.

(Operation of learning device)
FIG. 13 is a flowchart illustrating an example of the operation of the attribute estimation apparatus. In FIG. 13, the attribute estimation apparatus 300 acquires the attribute information estimation table 1088 in which the characteristic amount of power usage and the attribute information of power usage are associated from the learning apparatus 100, and the attribute information estimation table 1088 is acquired from the attribute information. The operation after storing in the estimation table 3088 is shown.
(Step S1302) The acquisition unit 352 of the attribute estimation apparatus 300 acquires information indicating the power usage amount of a home whose attribute information is unknown. When the acquisition unit 352 acquires information indicating the power usage of a home whose attribute information is unknown, the acquisition unit 352 outputs information indicating the power usage of a home whose attribute information is unknown to the learning unit 354.
(Step S1304) When the learning unit 354 of the attribute estimation device 300 acquires information indicating the power consumption of the home whose attribute information output by the acquisition unit 352 is unknown, the learning unit 354 calculates the power usage of the home whose attribute information is unknown. Deep learning is performed on the information shown. When the learning unit 354 performs deep learning on the information indicating the power usage of a home whose attribute information is unknown, the learning unit 354 outputs a feature value of the power usage obtained by performing the deep learning to the estimation unit 356. .

(Step S1306) When the estimation unit 356 of the attribute estimation apparatus 300 acquires the feature value of the household power consumption whose attribute information output by the learning unit 354 is unknown, the attribute calculation unit 300 refers to the attribute information estimation table 3088 and acquires the acquired attribute. The attribute information associated with the feature amount of the household power consumption whose information is unknown is estimated. If the estimation unit 356 estimates attribute information related to the feature amount of the household power consumption whose attribute information is unknown, the estimation unit 356 outputs the attribute information estimation result to the output unit 358.
(Step S1308) When the output unit 358 of the attribute estimation apparatus 300 acquires the attribute information estimation result output by the estimation unit 356, the output unit 358 outputs the acquired attribute information estimation result to the display unit 314.

  In the embodiment described above, the attribute estimation device 300 acquires the attribute information estimation table 1088 in which the feature amount of power usage and the attribute information of power usage are associated from the learning device 100, and the acquired attribute information estimation table 1088. Is described in the attribute information estimation table 3088, but is not limited to this example. For example, the attribute estimation apparatus 300 may inquire the learning apparatus 100 about attribute information related to the feature quantity of the household power consumption whose attribute information is unknown. In response to the inquiry from the learning device 100, the attribute estimation device 300 notifies the learning device 100 of attribute information associated with the feature amount of the household power usage whose attribute information is unknown.

  According to the attribute estimation device 300 according to the embodiment, based on the attribute information estimation table 1088 acquired from the learning device 100, the attribute information related to the feature amount of the household power consumption whose attribute information is unknown is estimated. Can do. Since it is possible to acquire household attribute information from which an unknown amount of power is used, it is possible to help develop a business that provides a new service based on the acquired attribute information.

(Third embodiment)
An attribute estimation system according to an embodiment will be described. The attribute estimation system according to the embodiment includes a learning device 100 according to the first embodiment and an attribute estimation device 300 according to the second embodiment.
As described above, the learning device 100 according to the embodiment acquires information indicating the power usage of each of a plurality of homes and home attribute information from which information indicating the power usage is obtained. The learning device 100 performs machine learning on the information indicating the power usage, and obtains a feature amount of the power usage. The learning device 100 associates the feature amount of the power usage amount with the home attribute information from which the information indicating the power usage amount is obtained. The learning apparatus 100 stores the feature amount of the power usage amount and the attribute information in association with each other.

  In addition, as described above, the attribute estimation apparatus 300 according to the embodiment acquires information indicating the power usage amount of a home whose attribute information is unknown. The attribute estimation apparatus 300 obtains a feature value of the power consumption of a home whose attribute information is unknown by machine learning information indicating the power consumption of a home whose acquired attribute information is unknown. The attribute estimation device 300 according to the embodiment acquires information that associates the feature amount of the power usage amount with the attribute information from the learning device 100 described above. The attribute estimation device 300 according to the embodiment obtains attribute information related to the feature amount of the household power usage amount whose attribute information is unknown based on the information that associates the acquired feature amount of the power usage amount with the attribute information. presume.

(Operation of attribute estimation system)
FIG. 14 is a diagram illustrating an operation of the attribute estimation system according to the embodiment. In FIG. 14, the attribute estimation apparatus 300 acquires from the learning apparatus 100 an attribute information estimation table 1088 that associates a feature value of power usage and attribute information of power usage, and uses the attribute information estimation table 1088 as attribute information estimation. The operation after storing in the table 3088 is shown.
(Step S1402) The acquisition unit 152 of the learning device 100 acquires information indicating the power usage of each of the plurality of homes from the smart meter.
(Step S1404) When the acquisition unit 152 of the learning device 100 acquires information indicating the power usage of each of a plurality of homes, the acquisition unit 152 acquires the home attribute information from which the information indicating the acquired power usage is obtained.
(Step S1406) The learning unit 154 of the learning device 100 learns the information indicating the power usage amount of each of the plurality of homes acquired by the acquisition unit 152 by performing deep learning. When the learning unit 154 learns the information indicating the power usage of each of the plurality of homes by performing deep learning, the feature amount and the power usage of the power usage obtained by performing the deep learning are obtained. The home attribute information is stored in association with each other.

(Step S1408) The acquisition unit 352 of the attribute estimation apparatus 300 acquires power usage information whose attribute information is unknown.
(Step S1410) The learning unit 354 of the attribute estimation apparatus 300 learns the information indicating the power usage information whose attribute information acquired by the acquisition unit 352 is unknown by performing deep learning.
(Step S1412) The estimation unit 356 of the attribute estimation apparatus 300 acquires attribute information associated with the feature amount of the power usage obtained by performing deep learning from the attribute information estimation table 3088. The estimation unit 356 of the attribute estimation device 300 outputs the acquired attribute information to the output unit 358. When the output unit 358 acquires the attribute information output by the estimation unit 356, the output unit 358 displays the attribute information on the display unit 314. In the example shown in FIG. 14, wake-up time, family structure, presence / absence of elderly people, absence time, and vacation acquisition time are displayed as attribute information.

The attribute estimation device 300 obtained the correct answer rate of the estimated attribute information by taking a questionnaire in the home where the power consumption was obtained.
(Estimation accuracy (1))
FIG. 15 is a diagram illustrating an example (part 1) of estimation accuracy of attribute information estimated by the attribute estimation apparatus 300 according to the embodiment. FIG. 15 shows an estimation example of the presence or absence of an elderly person. In FIG. 15, the horizontal axis is Epoch, the left vertical axis is Loss, and the right vertical axis is Accuracy. According to FIG. 15, it can be seen that the loss decreases and the accuracy approaches a constant value as the Epoch increases. According to FIG. 15, it can be seen that an accuracy of about 70% is obtained.

(Estimation accuracy (2))
FIG. 16 is a diagram illustrating an example (part 2) of the estimation accuracy of attribute information estimated by the attribute estimation apparatus 300 according to the embodiment. FIG. 16 shows an estimation example of whether or not he is at home in the daytime. In FIG. 16, the horizontal axis is Epoch, the left vertical axis is Loss, and the right vertical axis is Accuracy. According to FIG. 16, it can be seen that the loss decreases and the accuracy approaches a constant value as the Epoch increases. According to FIG. 16, it can be seen that an accuracy of about 80% is obtained.

(Estimation accuracy (3))
FIG. 17 is a diagram illustrating an example (part 3) of the estimation accuracy of attribute information estimated by the attribute estimation apparatus 300 according to the embodiment. FIG. 17 shows an estimation example of whether or not a pet is kept. In FIG. 17, the horizontal axis is Epoch, the left vertical axis is Loss, and the right vertical axis is Accuracy. According to FIG. 17, it can be seen that the loss decreases and the accuracy approaches a constant value as the Epoch increases. According to FIG. 17, it can be seen that an accuracy of about 60% is obtained.

  In the embodiment described above, the case where the attribute estimation apparatus 300 acquires the attribute information estimation table 1088 from the learning apparatus 100 has been described, but the present invention is not limited to this example. For example, the attribute estimation device 300 may inquire the learning device 200 about attribute information associated with the feature amount of the power usage information whose attribute information is unknown. Then, in response to the inquiry from the attribute estimation device 300, the learning device 100 may transmit attribute information associated with the feature amount of the power usage information whose attribute information is unknown to the attribute estimation device 300. .

In the above-described embodiment, the case where the learning apparatus 100 and the attribute estimation apparatus 300 are different apparatuses has been described, but the present invention is not limited to this example. For example, the learning device 100 and the attribute estimation device 300 may be realized by a single device.
In the above-described embodiment, the case where the attribute estimation system includes the learning device 100 and the attribute estimation device 300 has been described. However, the present invention is not limited to this example. For example, the attribute estimation system may include the learning device 200 and the attribute estimation device 300. In this case, the learning device 200 and the attribute estimation device 300 may be realized by a single device.

According to the attribute estimation system according to the embodiment, home attribute information from which information indicating the power usage of each of a plurality of homes and information indicating the power usage is obtained is acquired. The attribute estimation system performs machine learning on information indicating the power usage, and obtains a feature amount of the power usage. The attribute estimation system stores the characteristic amount of the power usage amount and the attribute information from which the information indicating the power usage amount is obtained in association with each other.
Furthermore, according to the attribute estimation system, by acquiring information indicating the power usage of a home whose attribute information is unknown, and deep learning information indicating the power usage of a home whose acquired attribute information is unknown, A feature amount of power consumption of a household whose attribute information is unknown is obtained. The attribute estimation system uses attribute information related to a feature amount of household power usage whose attribute information is unknown from information that associates the feature amount of power usage with the attribute information from which information indicating the power usage is obtained. Is estimated. By configuring in this way, the attribute estimation system according to the embodiment can acquire household attribute information from which an unknown amount of power usage has been obtained, so a business that provides a new service according to the acquired attribute information Can help with development.

<Modification>
The attribute estimation system according to the modification can apply the third embodiment described above. The attribute estimation system according to the modification applies a color heat map to information indicating the power usage. In the attribute system according to the modified example, the concept of time is included in the heat map in addition to the power consumption. By configuring in this way, the magnitude of power usage and the concept of time can be represented by different colors.
FIG. 18 is an example of information indicating power usage used by the attribute estimation system according to the modification. In the example illustrated in FIG. 18, the information indicating the power usage amount is represented by 256 gradations of RGB for each pixel constituting the information indicating the power usage amount. Specifically, the electric energy is shown in red, the time (1 o'clock to 24:00) is shown in green, and the date (1 / 1-12 / 31) is shown in blue. In FIG. 18, the magnitude of power usage and the concept of time are graphically represented.

(Estimation accuracy (4))
FIG. 19 is a diagram illustrating an example (part 4) of estimation accuracy of attribute information estimated by the attribute estimation apparatus 300 according to the embodiment.
FIG. 19 shows an example of estimating the presence or absence of an elderly person. In FIG. 19, the horizontal axis is Epoch, the left vertical axis is Loss, and the right vertical axis is Accuracy. According to FIG. 19, it can be seen that the loss decreases and the accuracy approaches a constant value as the Epoch increases. According to FIG. 15, it can be seen that an accuracy of about 80% is obtained.
It can be seen that the estimation accuracy is improved by about 10% compared to the estimation accuracy shown in (Estimation accuracy (part 1)).

According to the modification described above, the case where the color heat map is applied to the attribute estimation system has been described, but the present invention is not limited to this example. For example, the color heat map may be applied to the learning device 100 according to the first embodiment, may be applied to the learning device 200 according to the modification, or the attribute estimation device 300 according to the second embodiment. You may apply to.
According to the modified example described above, the case where the color heat map includes the concept of time in addition to the amount of power used has been described, but the present invention is not limited to this example. For example, you may make it include external temperature data, gas usage-amount data, and date information.
According to the attribute estimation system according to the modified example, by applying a color heat map to the information indicating the power usage amount, the estimation accuracy of the attribute information of the information indicating the household power usage amount whose attribute information is unknown is improved. be able to.

  The embodiments and the modifications thereof have been described above. However, these embodiments and the modifications thereof are presented as examples, and are not intended to limit the scope of the invention. These embodiments and modifications thereof can be implemented in various other forms, and various omissions, replacements, changes, and combinations can be made without departing from the spirit of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and at the same time, are included in the invention described in the claims and the equivalents thereof.

Note that the learning device and the attribute estimation device described above may be realized by a computer. In that case, a program for realizing the function of each functional block is recorded on a computer-readable recording medium. The program recorded on the recording medium may be read by a computer system and executed by the CPU. The “computer system” here includes hardware such as an OS (Operating System) and peripheral devices.
The “computer-readable recording medium” refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, and a CD-ROM. The “computer-readable recording medium” includes a storage device such as a hard disk built in the computer system.

Furthermore, the “computer-readable recording medium” may include a medium that dynamically holds a program for a short time. What holds the program dynamically for a short time is, for example, a communication line when the program is transmitted via a network such as the Internet or a communication line such as a telephone line.
In addition, the “computer-readable recording medium” may include a medium that holds a program for a certain period of time, such as a volatile memory inside a computer system that serves as a server or a client. The program may be for realizing a part of the functions described above. Further, the program may be a program that can realize the above-described functions in combination with a program already recorded in the computer system. The program may be realized using a programmable logic device. The programmable logic device is, for example, an FPGA (Field Programmable Gate Array).

The learning device and the attribute estimation device described above have a computer inside. Each process of the learning device and the attribute estimation device described above is stored in a computer-readable recording medium in the form of a program. The computer reads and executes this program, so that the above processing is performed. Is called.
Here, the computer-readable recording medium means a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like. Alternatively, the computer program may be distributed to the computer via a communication line, and the computer that has received the distribution may execute the program.
The program may be for realizing a part of the functions described above.
Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, what is called a difference file (difference program) may be sufficient.

DESCRIPTION OF SYMBOLS 100, 200 ... Learning apparatus, 102 ... CPU, 104 ... ROM, 106 ... RAM, 108 ... Non-volatile memory, 110 ... Communication I / F, 112 ... Operation part, 114 ... Display part, 116 ... Bus, 152 ... Acquisition part 154: Learning unit, 156 ... Output unit, 252 ... Acquisition unit, 254 ... Learning unit, 256 ... Output unit, 258 ... Creation unit, 300 ... Attribute estimation device, 302 ... CPU, 304 ... ROM, 306 ... RAM, 308 ... Non-volatile memory, 310 ... Communication I / F, 312 ... Operation part, 314 ... Display part, 316 ... Bus, 352 ... Acquisition part, 354 ... Learning part, 356 ... Estimation part, 358 ... Output part, 1082 ... Program, 1084: Power usage information, 1086: Attribute information, 1088 ... Attribute information estimation table, 3082 ... Program, 3088 ... Attribute information estimation table

Claims (13)

An acquisition unit that acquires information indicating the power usage of each of a plurality of homes and home attribute information from which the power usage is obtained;
Machine learning the information indicating the power usage of each of the plurality of households acquired by the acquisition unit, and the characteristic amount of the power usage obtained by the machine learning and the home from which the power usage is obtained A learning unit that associates the attribute information with
A learning device comprising: a storage unit that stores the characteristic amount of the power usage amount associated with the learning unit and the attribute information of the home from which the power usage amount is obtained in association with each other.   The learning apparatus according to claim 1, wherein the information indicating the power usage amount acquired by the acquisition unit is a representation of the magnitude of the power usage amount by a color difference.   The learning apparatus according to claim 2, wherein the information indicating the power usage amount acquired by the acquisition unit further represents the concept of time by a color difference.   The information indicating the power usage amount acquired by the acquisition unit is a color difference between the power usage amount and information indicating the environmental situation measured in the home from which the power usage amount was obtained. The learning device according to claim 1. Of the information indicating the power usage of each of the plurality of homes acquired by the acquisition unit, the power usage by usage indicating the usage by power usage from each of the power usage of homes with similar attribute information A creation unit that newly obtains information indicating the power usage based on one or more of the obtained power usage by usage,
The learning unit machine learns information indicating the newly generated power usage amount by the creation unit, and the power usage amount obtained by machine learning information indicating the newly created power usage amount The learning apparatus according to claim 1, wherein the feature amount is associated with the similar attribute information. An acquisition unit for acquiring information indicating power consumption of a household whose attribute information is unknown;
A feature amount of the power usage amount is obtained by machine learning information indicating the power usage amount acquired by the acquisition unit, and attribute information related to the feature amount indicates the power usage amount of each of a plurality of homes. An attribute estimation apparatus comprising: an estimation unit configured to estimate from information in which a feature amount of the power usage obtained by machine learning of information and attribute information of a home from which the power usage is obtained are associated.   The attribute estimation apparatus according to claim 6, wherein the information indicating the power usage amount of a home whose attribute information acquired by the acquisition unit is unknown is a representation of the magnitude of the power usage amount by a color difference.   The attribute estimation device according to claim 7, wherein the information indicating the household power consumption for which the attribute information acquired by the acquisition unit is unknown further represents a concept of time by a color difference.   The information indicating the power usage amount of the home whose attribute information acquired by the acquisition unit is unknown includes the power usage amount and information indicating the environmental condition measured in the home from which the power usage amount is obtained. The attribute estimation apparatus according to claim 6, wherein the attribute estimation apparatus is represented by a difference. Obtaining information indicating the power usage of each of the plurality of homes and the attribute information of the home from which the power usage was obtained;
Machine learning information indicating the power usage of each of the plurality of homes acquired in the acquiring step;
Associating the feature quantity of the power usage obtained by the machine learning with the attribute information of the home from which the power usage was obtained;
A step of associating and storing the characteristic amount of the power usage associated in the associating step and the attribute information of the home from which the power usage was obtained. Obtaining information indicating the power consumption of a household whose attribute information is unknown;
Obtaining a feature quantity of the power usage by machine learning information indicating the power usage obtained in the obtaining step;
The attribute information relating to the feature amount, the feature amount of the power usage amount obtained by machine learning of the information indicating the power usage amount of each of a plurality of households, and the home attribute information obtained from the power usage amount An attribute estimation method comprising: estimating from information associated with. In the computer with the learning device,
Obtaining information indicating the power usage of each of the plurality of homes and the attribute information of the home from which the power usage was obtained;
Machine learning information indicating the power usage of each of the plurality of homes acquired in the acquiring step;
Associating the feature quantity of the power usage obtained by the machine learning with the attribute information of the home from which the power usage was obtained;
A program for executing the step of associating and storing the characteristic amount of the power usage associated in the associating step and the attribute information of the home from which the power usage was obtained. In the computer with the attribute estimation device,
Obtaining information indicating the power consumption of a household whose attribute information is unknown;
Obtaining a feature quantity of the power usage by machine learning information indicating the power usage obtained in the obtaining step;
The attribute information relating to the feature amount, the feature amount of the power usage amount obtained by machine learning of the information indicating the power usage amount of each of a plurality of households, and the home attribute information obtained from the power usage amount A program that executes an estimation step based on information associated with.