Information processing method, program, and information processor

Abstract

To provide an information processing method and the like for supporting a user to interpret which event in the past affects the future.SOLUTION: In an information processing method, a computer executes processing of: acquiring time series data f(t), g(t), and h(t) concerning a plurality of respective items; recording a plurality of sets of an objective data vector yn constituted by data of each of items at a specific time tn and an explanatory data vector xn constituted by data of each of the items at a plurality of times before the specific time tn in association with each other; calculating an interpretation matrix (A dagger) which is a vector product of an explanatory matrix X in which the plurality of sets of explanatory data vectors xn are arranged and a generalized inverse matrix of an objective matrix Y in which the objective data vectors yn are arranged in an order corresponding to the explanatory data vectors xn; and outputting the interpretation matrix (A dagger).SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an information processing method, a program, and an information processing apparatus.

Background Art

[0002] A machine learning model generated by machine learning is a black box, and it is difficult for a user to interpret the behavior. XAI (Explainable Artificial Intelligence) technology has been proposed to show a reasonable basis for the results output by the machine learning model (Patent Document 1, Non-Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Non-Patent Documents

[0004]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] By the way, a phenomenon in which the reason for the result cannot be easily interpreted by humans occurs not only in the output results of machine learning models. Events occurring in the real world are the results of various events that have occurred in the past. However, in the real world, since various events influence each other, it is difficult to know which past event has influenced the event that will occur in the future.

[0006] On one side, it aims to provide an information processing method or the like that supports a user in interpreting which past events will affect the future.

Means for Solving the Problem

[0007] The information processing method acquires time-series data for each of a plurality of items, and associates and records a plurality of sets of a target database vector constituted by the data of each of the items at a specific time and an explanatory database vector constituted by the data of each of the items at a plurality of times before the specific time. The computer executes a process of calculating an interpretation matrix that is a vector product of an explanatory matrix in which a plurality of sets of the explanatory database vectors are arranged and a generalized inverse matrix of a target matrix in which the target database vectors are arranged in an order corresponding to the explanatory database vectors, and outputting the interpretation matrix.

Effect of the Invention

[0008] On one side, it is possible to provide an information processing method or the like that supports a user in interpreting which past events will affect the future.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Mode for Carrying Out the Invention

[0010] [Embodiment 1] A machine learning model that receives input of explanatory data and outputs target data is generated using various machine learning algorithms. When using the generated machine learning model, it is difficult for humans to interpret the judgment process from the input of explanatory data to the output of target data.

[0011] However, when applying the machine learning model to decision-making in the real world, it is important for humans to be able to interpret the judgment process of the machine learning model. For example, when the output target data seems to deviate greatly from human common sense, if humans can appropriately interpret the judgment process of the machine learning model, they can also appropriately judge how to handle the target data and the machine learning model.

[0012] Regarding the target data output from a machine learning model, the technology for explaining the reason for the output is called XAI (Explainable AI). Non-Patent Document 1 discloses AIME (Approximate Inverse Model Explanations), which is a type of XAI. AIME is an information processing method that supports users so that they can interpret the behavior of various machine learning models, including black box models whose generation algorithms and the like are unknown, from various viewpoints.

[0013] By the way, as described above, the events occurring in the real world are affected by various events that have occurred in the past. Since various events in the real world influence each other, it is difficult to surely predict the future from the data on past events. That is, the real world can be regarded as a kind of black box model that uses past events as explanatory data and future events as target data. Here, the explanatory data is time series data indicating the time series changes of past events.

[0014] In the present embodiment, by regarding the real world as a black box model, AIME disclosed in Cited Document 1 is applied to the interpretation of the real world. That is, in the present embodiment, an information processing method for supporting a user regarding the interpretation of events occurring in the real world is provided.

[0015] FIG. 1 is an explanatory diagram for explaining the outline of the interpretation matrix A†. The interpretation matrix A† is a matrix proposed in Non-Patent Document 1 and can be used for the interpretation of a black box model that receives an input of explanatory data and outputs target data. However, in the present embodiment, the point that the explanatory data is time series data is different from the matter disclosed in Non-Patent Document 1.

[0016] Various events occurring in the real world can be represented by time-series data that is a function of time. Therefore, the real world, which is a combination of multiple events, can be represented by multi-dimensional time-series data. In FIG. 1, the real world is schematically represented by three functions, f(t), g(t), and h(t), which show time-series data. t represents time. f(t), g(t), and h(t) for each item are examples of time-series data.

[0017] The horizontal axis of the graph shown at the top of FIG. 1 is time. The unit of the horizontal axis is any value related to time, such as year, month, day, hour, minute, or second. The vertical axis of the graph is the value related to each event such as f(t), g(t), and h(t). The unit of the vertical axis varies according to each event.

[0018] Functions such as f(t) are time-series data showing changes in items indicating physical quantities such as temperature, speed, flow rate, or power generation. Functions such as f(t) may be time-series data showing changes in items indicating artificial indicators such as exchange rates or stock prices. Functions such as f(t) may be data obtained by replacing the time-series changes of items showing non-numerical data such as "east, west, south, north" with the time-series changes of numerical data such as "1, 2, 3, 4".

[0019] Functions such as f(t) are discretized at every common time interval Δt. For example, when there are differences in the data measurement intervals of sensors, etc., it is corrected to data at the Δt interval by known methods such as interpolation or averaging.

[0020] The state of the real world at a specific time tn can be represented by the values of the functions at time tn, f(tn), g(tn), h(tn). In the following description, the data obtained by collecting the values of each function at time tn is described as "the current data at tn".

[0021] In the real world, it is common for the state of the recent past to have a stronger influence on the current state than the state of the distant past. Past data within the range that significantly affects the state at time tn is referred to as "past data of tn". In FIG. 1, the past data of tn is shown enclosed by a dashed line.

[0022] By arranging the individual data that constitutes the "past data of tn" in a column based on the rules described later, an explanatory database vector xn can be generated. By arranging the individual data that constitutes the "current data of tn" in a column based on the rules described later, a target database vector yn can be generated. The real world can be interpreted as a black box model that accepts the input of the explanatory database vector xn and outputs the target database vector yn.

[0023] An explanatory matrix X is generated based on the explanatory database vectors x for a plurality of times t. A target matrix Y is generated based on the target database vectors y for a plurality of times t. An interpretation matrix A† shown in equation (1) is calculated based on the explanatory matrix X and the target matrix Y. Details of the explanatory matrix X, the target matrix Y, and the interpretation matrix A† will be described later. X = A†Y ‥‥‥ (1)

[0024] FIG. 2 is an explanatory diagram for explaining the explanatory database vector x and the target database vector y. The record layout of the observed data DB33 shown on the left side of FIG. 2 is explained. The observed data DB33 is a database that records the results of observing events that occurred in the real world.

[0025] The observed data DB33 has a time field and an observed data field. The observed data field has fields related to various events that occur in the real world, such as the f(t) field, the g(t) field, and the h(t) field. In the following description, when there is no need to particularly distinguish which event the field relates to, it may be described as an event field. In FIG. 2 and subsequent descriptions, the case of using three event fields will be described as an example.

[0026] The time field records the time. Each event field records the state of each event at the time recorded in the time field. The observation data DB33 has one record for one time.

[0027] By arranging the observation data recorded in the record corresponding to time t2 in a column in the same order as the arrangement of the observation data fields, a data vector D2 representing the state of the real world at time t2 can be created. Similarly, a data vector representing the state of the real world at time tn is denoted as data vector Dn. n is an integer. The data vector Dn is a three-element vector. The time tn is an example of a specific time in the present embodiment.

[0028] Continuing the explanation by taking as an example the case where the event at time t10 is affected by the events at four times from time t6 to t9. That is, the real world is a black box model that receives the explanatory data vector x10 representing the events at four times from time t6 to t9 and outputs the explanatory data vector x10 representing the event at time t10.

[0029] The explanatory data vector x10 for the real world at time t10 is a vector in which the data vectors D6 to D9 for each of t6 to t9 are arranged in a column in chronological order. The target data vector y10 for the real world at time t10 is the same as the data vector D10 at t10. Therefore, the explanatory data vector x10 is a twelve-element vector, and the target data vector y10 is a three-element vector.

[0030] Note that the current data and the past data do not have to be continuous. For example, the past data at time t10 may be the data from time t3 to t8. In this case, the explanatory data vector x10 is an eighteen-element vector in which six vectors from the data vector D3 to the data vector D8 are arranged in a column in chronological order.

[0031] Figure 3 is an explanatory diagram for explaining the explanatory matrix X, the target matrix Y, and the interpretation matrix A†. The explanatory matrix X is a matrix in which explanatory data vectors xn for N time points are arranged in order of time series in the row direction. The explanatory matrix X is a two-dimensional matrix of 12 rows and N columns.

[0032] The target matrix Y is a matrix in which target data vectors yn corresponding to the respective explanatory data vectors xn are arranged in the row direction in the same order as the explanatory data vectors xn, that is, in time series order. The target matrix Y is a two-dimensional matrix of 3 rows and N columns.

[0033] Note that in order to execute the subsequent processes, the target data vectors yn need to be linearly independent. That is, the vector product of the target matrix Y and the transposed matrix of the target matrix Y needs to be a regular matrix.

[0034] Regarding the target matrix Y, Y† which is the Moore-Penrose generalized inverse matrix is calculated. In the following explanation, Y† may be described as the target inverse matrix Y†. The target inverse matrix Y† is calculated by Equation (2). The target inverse matrix Y† is a matrix of N rows and 3 columns. Y† = Y T (YY T ) -1 ‥‥‥ (2)

[0035] The interpretation matrix A† is the vector product of the explanatory matrix X and the target inverse matrix Y†. The equation for calculating the interpretation matrix A† is shown in Equation (3). The interpretation matrix A† is a two-dimensional matrix of 12 rows and 3 columns. A† = XY† ‥‥‥ (3)

[0036] The element in the a-th row and b-th column of the interpretation matrix A† indicates the influence of the a-th element of the explanatory data vector xn on the b-th element of the target data vector yn. Specifically, in the interpretation matrix A† shown in Figure 3, the element AD1_2 in the first row and second column is the influence of the first element of the explanatory data vector xn, that is, the value f(t - 4) of the function f at the most past time (t - 4) among the past data, on the second element g(t) of the target data vector yn at an arbitrary time t.

[0037] By examining the interpretation matrix A†, an expert can interpret the cause of the realization of the target data vector yn in the real world. By examining the interpretation matrix A†, an expert can determine what measures should be taken to realize a desirable target data vector yn. In addition, in order to support such interpretation and determination, it is desirable that the interpretation matrix A† be displayed not only as a numerical matrix but also by various graphs and the like.

[0038] For reference, the outline of the formula transformation for deriving Equation (3) from Equations (1) and (2) is shown below. First, multiply both sides of Equation (1) by the transpose matrix of the target matrix Y from the right to obtain Equation (4). XY T = A†YY T ‥‥‥ (4)

[0039] As described above, since the vector product of the target matrix Y and the transpose matrix of the target matrix Y is a regular matrix, the inverse matrix can be calculated. Multiply both sides of Equation (4) by this inverse matrix from the right to obtain Equation (5). XY T (YY T ) -1 = A†(YY T )(YY T ) -1 = A† ‥‥‥ (5)

[0040] After swapping the left and right sides of Equation (5) and then substituting Equation (2) into the right side, Equation (6) is obtained. Equation (2) is derived from both ends of Equation (6). A† = XY T (YY T ) -1 = XY† ‥‥‥ (6)

[0041] Figure 4 is an explanatory diagram for explaining the explanatory matrix X, the target matrix Y, and the interpretation matrix A†. Using Figure 4, the case where the number of elements of the database vector D and the like are generalized will be described. As shown within the frame of the preconditions, the number of elements of the database vector D is represented by p, the number of database vectors D used for past data is represented by q, and the number of times used for the explanatory matrix X and the target matrix Y is represented by N.

[0042] The number of elements of the explanatory data vector xn is p×q. The explanatory matrix X is a two-dimensional matrix with p×q rows and N columns. The number of elements of the target data vector yn is the same p as that of the data vector D. The target matrix Y is a two-dimensional matrix with p rows and N columns. The target inverse matrix Y† is a two-dimensional matrix with N rows and p columns. The interpretation matrix A† is a two-dimensional matrix with p×q rows and p columns.

[0043] FIG. 5 is an explanatory diagram for explaining the configuration of the information processing apparatus 10. The information processing apparatus 10 includes a control unit 11, a main memory device 12, an auxiliary storage device 13, a communication unit 14, a display unit 15, an input unit 16, a reading unit 19, and a bus.

[0044] The control unit 11 is an arithmetic control device that executes the program of the present embodiment. One or more CPUs (Central Processing Units), GPUs (Graphics Processing Units), TPUs (Tensor Processing Units), or multi-core CPUs, etc. are used for the control unit 11. The control unit 11 is connected to each hardware part constituting the information processing apparatus 10 via a bus.

[0045] The main memory device 12 is a storage device such as SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), or flash memory. Information necessary during the processing performed by the control unit 11 and the program being executed by the control unit 11 are temporarily stored in the main memory device 12.

[0046] The auxiliary storage device 13 is a storage device such as SRAM, flash memory, hard disk, or magnetic tape. The observation data DB33, the program to be executed by the control unit 11, and various data necessary for the execution of the program are stored in the auxiliary storage device 13. The observation data DB33 may be stored in an external storage device connected via a network.

[0047] The communication unit 14 is an interface that performs communication between the information processing apparatus 10 and a network. The display unit 15 is, for example, a liquid crystal display device or an organic EL (Electro Luminescence) display device. The input unit 16 is an input device such as, for example, a keyboard, a mouse, a trackball, or a microphone.

[0048] The portable recording medium 96 is, for example, a USB (Universal Serial Bus) memory, a CD-ROM (Compact Disc Read only memory), a magneto-optical disk medium, other optical disk media, or an SD memory card, etc. The portable recording medium 96 stores a program 97 for realizing AIME.

[0049] The reading unit 19 is an interface such as, for example, a USB connector, a CD-ROM drive, or an SD memory reader, etc., that can connect the portable recording medium 96. The semiconductor memory 98 stores the program 97 and is a memory that can be installed inside the information processing apparatus 10.

[0050] The information processing apparatus 10 is a general-purpose personal computer, a tablet, a mainframe computer, a virtual machine operating on a mainframe computer, or a quantum computer. The information processing apparatus 10 may be configured by a plurality of personal computers that perform distributed processing, or hardware such as a mainframe computer. The information processing apparatus 10 may be configured by a cloud computing system. The information processing apparatus 10 may be configured by a plurality of personal computers that operate in cooperation, or hardware such as a mainframe computer.

[0051] The program 97 is recorded on the portable recording medium 96. The control unit 11 reads the program 97 via the reading unit 19 and stores it in the auxiliary storage device 13. Further, the control unit 11 may read out the program 97 stored in the semiconductor memory 98. Furthermore, the control unit 11 may download the program 97 from another server computer (not shown) connected via the communication unit 14 and a network (not shown) and store it in the auxiliary storage device 13.

[0052] Program 97 is installed as a control program for the information processing apparatus 10, loaded into the main storage device 12, and executed. The program 97 of the present embodiment is an example of a program product.

[0053] FIG. 6 is a flowchart for explaining the processing flow of a program for calculating the interpretation matrix A†. The control unit 11 acquires the preconditions described with reference to FIG. 4, that is, the number p of elements of the data vector D, the sum q of the data vectors D used for past data, and the number N of times used for the explanatory matrix X and the target matrix Y (step S501).

[0054] The preconditions may include an instruction regarding the time in the observation data DB33 at which the creation of the explanatory matrix X and the target matrix Y starts. In the following description, the case where the creation of the explanatory matrix X and the target matrix Y starts from the latest time recorded in the observation data DB33 will be described as an example.

[0055] The preconditions may include an instruction regarding the interval between current data and past data. In the following description, the case where the past data and the current data are continuous as illustrated in FIG. 2 will be described as an example.

[0056] The preconditions are input, for example, by the user via the input unit 16. A file recording the preconditions may be stored in the auxiliary storage device 13 or an external storage device connected via a network.

[0057] The control unit 11 sets the current data described with reference to FIG. 2 (step S502). For example, in the first step S502, the control unit 11 sets the latest data recorded in the observation data DB33 as the current data. In the second and subsequent steps S502, the control unit 11 sets data that is one or a predetermined number of data points older than the data set in the previous step S502 as the current data.

[0058] The control unit 11 acquires past data composed of a predetermined number of data vectors D recorded in the past from the observation data DB33 than the current data (step S503). The control unit 11 generates an explanatory data vector xn based on the acquired past data (step S504).

[0059] The control unit 11 acquires the current data from the observation data DB33 (step S505). The control unit 11 generates a target data vector yn based on the acquired current data (step S506). The control unit 11 associates the generated explanatory data vector xn and the target data vector yn and records them in the auxiliary storage device 13 or the main storage device 12 (step S507).

[0060] The control unit 11 determines whether or not the generation of the N explanatory data vectors xn and the target data vectors yn acquired in step S501 has been completed (step S508). If it is determined that the generation has not been completed (NO in step S508), the control unit 11 returns to step S502.

[0061] If it is determined that the generation has been completed (YES in step S508), the control unit 11 generates an explanatory matrix X based on the data recorded in step S507 (step S509). The control unit 11 generates a target matrix Y based on the data recorded in step S508 (step S510). The control unit 11 calculates a target inverse matrix Y† which is the Moore-Penrose generalized inverse matrix based on the target matrix Y (step S511). The control unit 11 calculates an interpretation matrix A† which is the vector product of the explanatory matrix X and the target inverse matrix Y† (step S512). The control unit 11 ends the process.

[0062] The generated interpretation matrix A† is charted, and an outline of a method for interpreting the behavior of the existing model 21 will be described below.

[0063] [Embodiment 2] This embodiment relates to a method of utilizing the interpretation matrix A† for weather interpretation. Regarding the parts common to Embodiment 1, the description will be omitted.

[0064] In weather forecasts announced by the Japan Meteorological Agency and others, a large amount of observation data, simulations using supercomputers, and machine learning models are utilized. However, even currently, the accuracy of weather forecasts is not 100 percent. For example, regarding locally occurring severe weather phenomena such as guerrilla heavy rain and tornadoes, forecasts that enable appropriate responses such as evacuation may not be made, and damage may occur.

[0065] By utilizing the interpretation matrix A† described in Embodiment 1, it is possible to provide information regarding the cause of a weather for which appropriate forecasting has been difficult. If the cause can be estimated, the Japan Meteorological Agency and others can take measures to improve the forecasting accuracy, such as enhancing measuring instruments related to the cause or improving the accuracy of simulations.

[0066] FIG. 7 is an explanatory diagram for explaining the record layout of the observation data DB33 of Embodiment 2. The observation data DB33 of the present embodiment is a database that accumulates observation data regarding weather. The observation data DB33 includes a date field, an average atmospheric pressure field, an average temperature field, a maximum temperature field, a minimum temperature field, an average humidity field, a sunshine duration field, a precipitation amount field, and the like.

[0067] The date is recorded in the date field. The average atmospheric pressure for a day is recorded in the average atmospheric pressure field. The average temperature for a day is recorded in the average temperature field. The maximum temperature for a day is recorded in the maximum temperature field. The minimum temperature for a day is recorded in the minimum temperature field. The average humidity for a day is recorded in the average humidity field. The time during which direct sunlight irradiates the instrument during a day is recorded in the sunshine duration field. The precipitation amount is recorded in the precipitation amount field.

[0068] The observation data DB33 may include a field for recording the time when peak values such as the maximum temperature and the minimum temperature are observed. The observation data DB33 may also include fields for recording data observed at a plurality of surrounding points respectively.

[0069] FIG. 8 is an explanatory diagram for explaining the outline of the second embodiment. In the preparation stage, using the observation data DB33, the interpretation matrix A† of the present embodiment is calculated by the procedure described in the first embodiment. The interpretation matrix A† is calculated, for example, for each region. The interpretation matrix A† may be calculated for each season.

[0070] The utilization stage of the calculated interpretation matrix A† will be described. The Japan Meteorological Agency or the like issues, for example, a weather forecast for the next day based on the latest observation data. Hereinafter, the case where the weather forecast is different from the actual weather, that is, the case where the weather forecast is off, will be described.

[0071] The announced weather forecast is converted into a forecast target data vector yp by applying the forecast content to each observation item. According to equation (7), a forecast explanation data vector xp corresponding to the forecast target data vector yp is calculated. xp = A†yp ‥‥‥ (7)

[0072] The Japan Meteorological Agency or the like continues to accumulate observation data. The observation data on the day corresponding to the weather forecast is converted into an actually measured target data vector ya. According to equation (8), an actually measured explanation data vector xa corresponding to the actually measured target data vector ya is calculated. xa = A†ya ‥‥‥ (8)

[0073] When the weather forecast is off, there is a difference between the forecast target data vector yp and the actually measured target data vector ya. Each element constituting the forecast explanation data vector xp calculated by equation (7) is compared with each element constituting the actually measured explanation data vector xa calculated by equation (8), and the element with a large difference is presumed to be the cause of the difference between the forecast target data vector yp and the actually measured target data vector ya.

[0074] For elements with large differences, experts can conduct a detailed examination and, for example, implement measures such as improving the observation accuracy or increasing the number of observation items, so as to expect to improve the accuracy of future weather forecasts. In any case of various meteorological conditions, for elements with small differences, since the impact on the accuracy of weather forecasts is small, it may be possible to reduce the cost of weather forecasts, for example, by suspending observations.

[0075] FIG. 9 is a flowchart for explaining the processing flow of the program according to Embodiment 2. In the following description, the case of utilizing the interpretation matrix A† using the information processing apparatus 10 of Embodiment 1 described with reference to FIG. 5 will be taken as an example for explanation. The program of FIG. 9 may be executed by hardware different from the information processing apparatus 10 used for calculating the interpretation matrix A†.

[0076] The control unit 11 acquires the interpretation matrix A† calculated in the preparation stage (step S530). The control unit 11 acquires the published weather forecast (step S531). The control unit 11 creates a forecast target data vector yp based on the acquired weather forecast (step S532). The control unit 11 calculates a forecast explanation data vector xp by equation (7) (step S533).

[0077] The control unit 11 acquires the observation data of the day corresponding to the weather forecast (step S534). The control unit 11 creates an actual measurement target data vector ya based on the observation data (step S535). The control unit 11 calculates an actual measurement explanation data vector xa based on equation (8) (step S536).

[0078] The control unit 11 compares each element constituting the forecast explanation data vector xp with each element constituting the actual measurement explanation data vector xa, and extracts elements with large differences (step S537). The control unit 11 displays the information regarding the elements with large differences on the display unit 15 (step S538). The control unit 11 ends the processing.

[0079] According to the present embodiment, it is possible to provide an information processing method and the like that assist experts in terms of the possibility of improving the accuracy of weather forecasts and the possibility of reducing the cost of weather forecasts.

[0080] The application target of the interpretation matrix A† is not limited to weather forecasts. The interpretation matrix A† can be applied to assist in understanding various time-series phenomena such as complex phenomena caused by many human behaviors such as traffic congestion prediction, or complex physical phenomena, etc., where it is difficult to intuitively grasp the causal relationship or correlation.

[0081] [Embodiment 3] This embodiment relates to a method of utilizing the interpretation matrix A† for improving air conditioning equipment. Regarding the parts common to Embodiment 1, the description will be omitted.

[0082] Air conditioning equipment is installed in many facilities used by people. In large-scale facilities such as shopping malls, high-rise buildings, or stadiums, it is common to centrally control the air conditioning equipment based on the room temperature detected by sensors arranged in various places.

[0083] However, in actual facilities, the expected air conditioning effect cannot be obtained, and there may be places that are too hot or too cold. In a facility with a complex structure, if the air conditioning equipment is operated to cool a place that is too hot, a place that is far away may become too cold over time. In this embodiment, the interpretation matrix A† is utilized to improve the air conditioning equipment in such facilities.

[0084] FIG. 10 is an explanatory diagram for explaining the record layout of the observation data DB33 in Embodiment 3. The observation data DB33 of this embodiment is a database that accumulates the temperatures observed at various places in the facility. The observation data DB33 has a date and time field and a temperature field. The temperature field has a first location field, a second location field, and the like.

[0085] The measurement date and time are recorded in the date and time field. The temperature measured at each location is recorded in each sub-field of the temperature field. Note that the observation data DB33 may have a field for recording humidity, dew point, etc. in addition to the field for recording temperature.

[0086] FIG. 11 is an explanatory diagram for explaining the outline of Embodiment 3. First, temperature sensors are installed at a plurality of locations to measure time-series data of temperature. The measurement results are stored in the observation data DB33 described with reference to FIG. 10.

[0087] Using the observation data DB33, the interpretation matrix A† of the present embodiment is calculated by the procedure described in Embodiment 1. A target data vector yn indicating the target temperature distribution is created. The created target data vector yn is, for example, a target data vector yn in which all elements are set to the target temperature. For example, a target data vector yn in which different temperatures are set for elements corresponding to corridors and elements corresponding to rooms may be created.

[0088] According to Equation (9), an explanatory data vector xn corresponding to the target data vector yn is calculated. xn = A†yn ‥‥‥ (9)

[0089] The calculated explanatory data vector xn indicates how to control the temperature at which position in order to realize the target data vector. An expert in air conditioning equipment examines the explanatory data vector xn and adjusts the air conditioning equipment, ventilation equipment, etc. The expert checks whether the target has been achieved based on the time-series data of the temperature measured after the adjustment. If the target has not been achieved, the above procedure is repeated based on the newly measured data.

[0090] FIG. 12 is a flowchart for explaining the processing flow of the program according to Embodiment 3. In the following description, a case where the interpretation matrix A† is utilized using the information processing apparatus 10 of Embodiment 1 described with reference to FIG. 5 will be described as an example. The program in FIG. 12 may be executed on hardware different from the information processing apparatus 10 used for calculating the interpretation matrix A†.

[0091] The control unit 11 acquires the interpretation matrix A† calculated in advance by the program described in Embodiment 1 (step S550). The control unit 11 acquires the target data vector yn created by the user (step S551). The control unit 11 calculates the explanatory data vector xn according to equation (9) (step S552). The control unit 11 displays the calculated explanatory data vector xn (step S553). The control unit 11 ends the process.

[0092] According to the present embodiment, an information processing method and the like for assisting experts in improving air conditioning equipment can be provided.

[0093] The application target of the interpretation matrix A† is not limited to the improvement of air conditioning equipment. The interpretation matrix A† can be utilized for various applications such as, for example, the improvement of audio equipment and the improvement of manufacturing equipment.

[0094] The program is an example of a program product. The computer program can be deployed to be executed on a single computer, or placed at one site, or distributed over a plurality of sites and interconnected by a communication network and executed on a plurality of computers.

[0095] The technical features (constituent elements) described in each embodiment can be combined with each other, and by combining them, new technical features can be formed. The embodiments disclosed herein should be considered illustrative in all respects and not restrictive. The scope of the present invention is shown not by the above description but by the claims, and it is intended that all modifications within the meaning and scope equivalent to the claims be included.

[0096] The independent claims and dependent claims recited in the claims can be combined with each other in any combination regardless of the citation form. Further, the claims use a form (multi-claim form) of reciting a claim that cites two or more other claims, but it is not limited thereto. A form of reciting a multi-claim (multi-multi-claim) that cites at least one multi-claim may be used.

Description of Reference Numerals

[0097] 10 Information processing apparatus 11 Control unit 12 Main memory device 13 Auxiliary storage device 14 Communication unit 15 Display unit 16 Input unit 19 Reading unit 96 Portable recording medium 97 Program 98 Semiconductor memory

Claims

1. Obtain time series data for each of a number of items, a plurality of sets of an objective data vector constituted by the data of each of the items at a specific time and an explanation data vector constituted by the data of each of the items at a plurality of times prior to the specific time are recorded in association with each other; Calculate an interpretation matrix which is a vector product of an explanation matrix in which a plurality of sets of the explanation data vectors are arranged and a generalized inverse matrix of a target matrix in which the target data vectors are arranged in an order corresponding to the explanation data vectors; Output the interpretation matrix An information processing method in which processing is performed by a computer.

2. The explanation data vector is a sequence of the items arranged in the same order as the target data vector, and arranged in chronological order. The information processing method according to claim 1 .

3. The explanatory data vector is composed of data at a plurality of consecutive times. The information processing method according to claim 1 .

4. The explanation data vector includes a portion of the object data vector.

4. The information processing method according to claim 1.

5. Obtain time series data for each of a number of items, a plurality of sets of an objective data vector constituted by the data of each of the items at a specific time and an explanation data vector constituted by the data of each of the items at a plurality of times prior to the specific time are recorded in association with each other; Calculate an interpretation matrix which is a vector product of an explanation matrix in which a plurality of sets of the explanation data vectors are arranged and a generalized inverse matrix of a target matrix in which the target data vectors are arranged in an order corresponding to the explanation data vectors; Output the interpretation matrix A program that causes a computer to carry out processing.

6. An information processing device including a control unit, The control unit is Obtain time series data for each of a number of items, a plurality of sets of an objective data vector constituted by the data of each of the items at a specific time and an explanation data vector constituted by the data of each of the items at a plurality of times prior to the specific time are recorded in association with each other; Calculate an interpretation matrix which is a vector product of an explanation matrix in which a plurality of sets of the explanation data vectors are arranged and a generalized inverse matrix of a target matrix in which the target data vectors are arranged in an order corresponding to the explanation data vectors; Output the interpretation matrix Information processing device.

Images