JP2011028339A - Information processor, information processing system and information processing method - Google Patents

Abstract

A causal relationship and a conditional probability between events corresponding to a user are calculated, and a Bayesian network is easily constructed.
An information processing apparatus includes, for each of a plurality of nodes corresponding to a plurality of events representing a user's state, a prior probability indicating a probability that an event corresponding to the node occurs, and an event corresponding to the node occurred. The conditional probability indicating the probability that an event corresponding to another node will occur under the condition is received and stored in advance associated with each combination of the node and the other node, based on the prior information, Graph data indicating a parent-child relationship between a plurality of nodes and a conditional probability table indicating a conditional probability that a child node event occurs corresponding to the parent node event are generated and stored for each parent-child relationship.
[Selection] Figure 1

Description

  The present invention relates to a technique for constructing a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing a user's state and a causal relationship between the plurality of events.

  2. Description of the Related Art In recent years, research has been conducted on information systems that estimate a user's situation and mood based on data acquired by various sensors and provide information services according to the estimated status to the user. Such an information system includes, for example, data such as a user's body temperature and blood pressure acquired by a vital sensor, and user position data acquired by a GPS (Global Positioning System) function provided in a mobile phone terminal held by the user. Based on the user's schedule data, etc., an information service corresponding to the user's state is provided to the user terminal to support the user's behavior.

  In such an information system, it is conceivable to use a Bayesian network when determining an advanced state that cannot be acquired by a sensor among information services according to the state of the user and related user states. A Bayesian network is a model of probabilistic reasoning in which the relationship between events is abstracted and represented by a directed graph structure. In a Bayesian network, the causal relationship between an event and another event is quantified and expressed by a conditional probability. This predicts the likelihood and likelihood of a complex and uncertain event, and determines the information service and advanced state provided to the user. Patent Document 1 or Patent Document 2 proposes an information system that performs estimation using a Bayesian network. The relationship between events and conditional probabilities in such a Bayesian network is calculated based on data acquired in advance, or based on user characteristics by a system administrator with Bayesian network expertise. The value calculated in this way is used.

JP 2007-279887 A JP 2008-269215 A

  However, since the number of conditional probability parameters indicating the causal relationship between events in a Bayesian network increases explosively in combination with the number of events, a Bayesian network including a large number of events that can withstand practical use is constructed. For this purpose, a huge amount of data is required. It is difficult to obtain all such data corresponding to each of a large number of users in advance. On the other hand, when a Bayesian network is constructed by a system administrator with specialized knowledge, the system administrator must set up events and parameters corresponding to all of a large number of users, which requires considerable effort and time. It becomes. Therefore, it is conceivable that a user who receives an action support service by an information system designs and constructs a Bayesian network corresponding to the user, but in order to extract a combination that has a direct causal relationship between events, A panoramic knowledge of all the events included in the Bayesian network is required. For example, if the combination of a plurality of events having a direct causal relationship with a certain event is not comprehensively understood, the conditional probability between the plurality of events cannot be set.

  The present invention has been made in view of such a situation. An information processing apparatus, an information processing system, and an information processing system that easily calculate a causal relationship and a conditional probability between events corresponding to a user and easily construct a Bayesian network. Provide a method.

  In order to solve the above-described problem, the present invention provides an information processing apparatus that generates a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing a user's state and a causal relationship between the plurality of events. A prior probability storage unit that stores a prior probability indicating a probability that an event corresponding to the node occurs for each node, and another node on condition that the event corresponding to the node has occurred Conditional probability indicating the probability that an event corresponding to the occurrence will occur in association with each combination of the node and other nodes, and a conditional probability storage unit, and graph data indicating a parent-child relationship between a plurality of nodes And a conditional probability table storage unit for storing a conditional probability table indicating a conditional probability that an event of a child node occurs corresponding to the event of the parent node for each parent-child relationship Receiving the prior information including the prior probability and the conditional probability, storing the prior probability included in the received prior information in the prior probability storage unit, and storing the conditional probability included in the received prior information as the conditional probability A conditional probability table is generated by generating a conditional probability table based on the prior information acquisition unit to be stored in the unit, the prior probability stored in the prior probability storage unit, and the conditional probability stored in the conditional probability storage unit. And a conditional probability table generation unit stored in the table storage unit.

  According to the present invention, the information processing apparatus described above generates graph data based on the prior probability stored in the prior probability storage unit and the conditional probability stored in the conditional probability storage unit, and stores the graph data. A graph generation unit to be stored in the unit is provided.

  Further, according to the present invention, the graph generation unit described above calculates a mutual information amount indicating a correlation of a node pair with another node for each of a plurality of nodes based on the prior probability and the conditional probability, and calculates the calculated mutual information. From a node pair having a relatively high amount of information in descending order, a link that associates a link that associates a node pair with each other as long as a cyclic route does not occur, and a node that is associated with a plurality of links by the link selector as a central node, A direction determining unit that sets the direction of the link that associates a node whose mutual information amount is below a predetermined threshold and the center node among the plurality of nodes adjacent to the center node to the center node direction, Graph data is generated based on the link and the direction of the link.

  Further, the present invention relates to a node in which the prior information is not received by the prior information acquisition unit among the plurality of nodes stored in the prior probability storage unit or the conditional probability storage unit. A prior information complementing unit that generates the prior information to be generated based on the prior information received by the prior information acquisition unit, and stores the generated prior information in the prior probability storage unit or the conditional probability storage unit. .

  In addition, the present invention is characterized in that the conditional probability table generation unit described above generates a conditional probability table using a naive Bayes method.

  Further, the present invention provides an observation value storage unit that stores an observation value indicating whether an event has occurred, an observation value storage unit that receives the observation value, and the received observation value is an observation value storage unit. Based on the observation value receiving unit to be stored in the observation value, the observation value stored in the observation value storage unit, the conditional probability table stored in the conditional probability table storage unit, and the graph data stored in the graph storage unit An estimation unit that extracts another event corresponding to the event indicated by the observation value received by the observation value reception unit.

  In addition, the present invention provides an information service corresponding to a user's state based on a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing the user's state and a causal relationship between the plurality of events. A prior probability storage unit in which a prior probability indicating a probability that an event corresponding to the node occurs is stored for each node, and an event corresponding to the node has occurred As a conditional probability storage unit in which a conditional probability indicating a probability of occurrence of an event corresponding to another node is stored in association with each combination of the node and the other node, and a parent-child relationship between a plurality of nodes A graph storage unit storing graph data indicating a conditional probability table indicating a conditional probability that an event of a child node occurs corresponding to an event of the parent node is stored for each parent-child relationship A conditional probability table storage unit that receives the prior information including the prior probability and the conditional probability, and stores the prior probability included in the received prior information in the prior probability storage unit and is included in the received prior information Based on the prior information acquisition unit that stores the conditional probability in the conditional probability storage unit, the prior probability stored in the prior probability storage unit, and the conditional probability stored in the conditional probability storage unit A conditional probability table generating unit that generates a table and stores it in the conditional probability table storage unit, an observation value receiving unit that receives observation values, and stores the received observation values in the observation value storage unit, and an observation value storage unit The event indicated by the observation received by the observation receiver based on the observation stored in the conditional probability table stored in the conditional probability table storage and the graph data stored in the graph storage An estimator that extracts other events corresponding to The information processing device provided, the user terminal that transmits the prior information input from the user to the information processing device, the sensor terminal that generates the observation value corresponding to the event and transmits it to the information processing device, and the information processing device extract And an application server that transmits an information service corresponding to the other event to the user terminal.

  In addition, the present invention provides an information service corresponding to a user's state based on a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing the user's state and a causal relationship between the plurality of events. A prior probability storage unit in which a prior probability indicating a probability that an event corresponding to the node occurs is stored for each node, and an event corresponding to the node has occurred As a conditional probability storage unit in which a conditional probability indicating a probability of occurrence of an event corresponding to another node is stored in association with each combination of the node and the other node, and a parent-child relationship between a plurality of nodes A graph storage unit storing graph data indicating a conditional probability table indicating a conditional probability that an event of a child node occurs corresponding to an event of the parent node is stored for each parent-child relationship The conditional probability table storage unit, the prior information including the prior probability and the conditional probability input from the user is received, the prior probability included in the input prior information is stored in the prior probability storage unit, and received. Based on the prior information acquisition unit that stores the conditional probability included in the prior information in the conditional probability storage unit, the prior probability stored in the prior probability storage unit, and the conditional probability stored in the conditional probability storage unit A conditional probability table generating unit that generates a conditional probability table and stores the conditional probability table in the conditional probability table storage unit, an observation value receiving unit that receives the observation values and stores the received observation values in the observation value storage unit; The observation value reception unit receives the observation value stored in the observation value storage unit, the conditional probability table stored in the conditional probability table storage unit, and the graph data stored in the graph storage unit. Other corresponding to the event indicated by the observation A sensor that extracts an elephant and includes an estimation unit that receives an information service from an application server that provides an information service corresponding to the extracted event, and generates an observation value corresponding to the event and transmits the observation value to the user terminal A terminal and an application server that transmits an information service corresponding to another event extracted by the user terminal to the user terminal.

  In addition, the present invention provides an information service corresponding to a user's state based on a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing the user's state and a causal relationship between the plurality of events. A prior probability storage unit in which a prior probability indicating a probability that an event corresponding to the node occurs is stored for each node, and an event corresponding to the node has occurred As a conditional probability storage unit in which a conditional probability indicating a probability of occurrence of an event corresponding to another node is stored in association with each combination of the node and the other node, and a parent-child relationship between a plurality of nodes A graph storage unit storing graph data indicating a conditional probability table indicating a conditional probability that an event of a child node occurs corresponding to an event of the parent node is stored for each parent-child relationship A conditional probability table storage unit that receives the prior information including the prior probability and the conditional probability, and stores the prior probability included in the received prior information in the prior probability storage unit and is included in the received prior information Based on the prior information acquisition unit that stores the conditional probability in the conditional probability storage unit, the prior probability stored in the prior probability storage unit, and the conditional probability stored in the conditional probability storage unit A conditional probability table generating unit that generates a table and stores it in the conditional probability table storage unit, an observation value receiving unit that receives observation values, and stores the received observation values in the observation value storage unit, and an observation value storage unit The event indicated by the observation received by the observation receiver based on the observation stored in the conditional probability table stored in the conditional probability table storage and the graph data stored in the graph storage Extract other events corresponding to the An application server comprising: an estimation unit that transmits an information service corresponding to the user terminal; a user terminal that transmits advance information input from the user to the application server; and an observation value corresponding to the event, And a sensor terminal that transmits to the server.

  Further, the present invention generates a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing a user's state and a causal relationship between the plurality of events, and an event corresponding to the node for each node. A prior probability storage unit in which a prior probability indicating the probability of occurrence of an event is associated and stored, and a conditional probability indicating the probability of an event corresponding to another node occurring on the condition that an event corresponding to the node has occurred A conditional probability storage unit that is stored in association with each combination of a node and another node, a graph storage unit that stores graph data indicating a parent-child relationship between a plurality of nodes, and a parent-child relationship A conditional probability table storage unit that stores a conditional probability table that indicates a conditional probability that an event of a child node occurs in response to an event of a parent node, and an information processing method for an information processing device comprising: Prior information including prior probabilities and conditional probabilities is received, prior probabilities included in the received prior information are stored in a prior probability storage unit, and conditional probabilities included in the received prior information are conditional probability storage units And generating a conditional probability table on the basis of the prior probability stored in the prior probability storage unit and the conditional probability stored in the conditional probability storage unit. And a step of storing.

  As described above, according to the present invention, the information processing apparatus includes, for each of a plurality of nodes corresponding to a plurality of events representing a user's state, a prior probability indicating a probability that an event corresponding to the node occurs, and a node Receives and stores a priori information in which a conditional probability indicating the probability that an event corresponding to another node will occur on the condition that an event corresponding to the occurrence of the event corresponds to each combination of the node and the other node Based on prior information, graph data indicating a parent-child relationship between a plurality of nodes, and a conditional probability table indicating a conditional probability that an event of a child node occurs corresponding to an event of the parent node for each parent-child relationship Are generated and stored, so that it is possible to easily construct a Bayesian network by calculating a causal relationship between events and a conditional probability based on prior information input from the user.

It is a block diagram which shows the structural example of the information processing system by one Embodiment of this invention. It is a figure which shows the example of the node defined by one Embodiment of this invention. It is a block diagram which shows the example of data of the prior probability table by one Embodiment of this invention. It is a block diagram which shows the example of data of the conditional probability table by one Embodiment of this invention. It is a block diagram which shows the example of data of the prior probability table complemented by one Embodiment of this invention. It is a block diagram which shows the example of data of the conditional probability table complemented by one Embodiment of this invention. It is a figure which shows the example of the link between the nodes defined by one Embodiment of this invention. It is a figure which shows the example of the link direction between the nodes defined by one Embodiment of this invention. It is a block diagram which shows the example of data of the graph produced | generated by one Embodiment of this invention. It is a block diagram which shows the example of the conditional probability produced | generated by one Embodiment of this invention. It is a block diagram which shows the example of CPT produced | generated by one Embodiment of this invention. It is a block diagram which shows the example of data of the observation value by one Embodiment of this invention. It is a sequence diagram which shows the operation example of the information processing system by one Embodiment of this invention. It is a flowchart figure which shows the operation example of the link selection process by one Embodiment of this invention. It is a flowchart figure which shows the operation example of the link direction determination process by one Embodiment of this invention. It is a block diagram which shows the structural example of the information processing system by one Embodiment of this invention. It is a block diagram which shows the structural example of the information processing system by one Embodiment of this invention.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram illustrating a configuration of an information processing system 1 according to the present embodiment. The information processing system 1 includes a plurality of user terminals 400 (user terminals 400-1, user terminals 400-2, user terminals 400-3,...), An information processing apparatus 10, and a plurality of application servers 500 (application servers). 500-1, application server 500-2, application server 500-3,...) And a plurality of sensor terminals 600 (sensor terminal 600-1, sensor terminal 600-2, sensor terminal 600-3,...). And.

  The user terminal 400 is a computer terminal that includes an input unit and a transmission unit, and transmits prior information input from the user to the information processing apparatus 10. The prior information is information in which an event is associated with a probability that the event occurs. The probability includes a prior probability indicating a probability that a corresponding event occurs when there is no condition, and a conditional probability indicating a probability that another event occurs when a certain event occurs. For example, the prior information includes a prior probability (for example, 0.15) corresponding to the event “home”, and a conditional probability (for example, the event “I want to watch a video” occurs when the event “home” occurs) , 0.6).

  The information processing apparatus 10 is based on prior information transmitted from the user terminal 400 and includes information on a plurality of nodes corresponding to a plurality of events representing a user state and causal relationships between the plurality of events. Is generated and stored. The information processing apparatus 10 receives data such as user positions and schedules acquired by the plurality of sensor terminals 600, and responds to the event of the user state based on the received data and the Bayesian network corresponding to the user. Determine the event. In addition, the information processing apparatus 10 transmits information recommending to the user terminal 400 an information service by the application server 500 that provides an information service corresponding to the determined event. The information processing apparatus 10 includes a probability network creation unit 100, a Bayesian network storage unit 200, and an estimation unit 300. The probability network creation unit 100 includes a prior information acquisition unit 110, a prior information holding unit 120, a graph generation unit 130, and a CPT generation unit 140.

The prior information acquisition unit 110 receives the prior information including the prior probability and the conditional probability from the user terminal 400, stores the prior probability included in the received prior information in the prior probability storage unit 121, and receives the received prior information. The conditional probability included in the information is stored in the conditional probability storage unit 122.
The prior information holding unit 120 includes a prior probability storage unit 121, a conditional probability storage unit 122, and a prior information complement unit 123.

  The prior probability storage unit 121 stores, for each node of the Bayesian network, a prior probability indicating the probability that an event corresponding to the node occurs in association with each other. Here, the prior probability storage unit 121 stores a prior probability table including a plurality of prior probabilities included in the prior information transmitted from the user terminal 400. Here, a Bayesian network node is commonly defined for all users as a pre-configuration by the system administrator. FIG. 2 is a diagram illustrating an example of nodes defined in the Bayesian network. For example, nodes indicating events such as “I want to listen to music”, “I want to play a game”, “Home”, and the like are defined. Each event is associated with one of two values, for example, occurrence (for example, “1”) or non-occurrence (for example, “0”). Further, identification information (node ID) for identifying the node is given to these nodes. Although the node ID may be information such as a numerical value, in the present embodiment, a description will be given assuming that a character string such as “I want to hear music” or “Home” is the node ID as it is.

  FIG. 3 is a diagram illustrating a data example of prior probabilities associated with the nodes defined as described above. As shown in the figure, for example, in the prior probability table, the prior probability “0.15” having the event “home” as a node and the prior probability “0.2” having the event “I want to play a game” as a node are shown. Are stored in association with each other. In addition to this example, prior probabilities corresponding to a plurality of nodes may be stored in the prior probability table. Here, in each storage unit described in the present embodiment, information on prior probabilities for each user may be stored in association with identification information for identifying each of a plurality of users. For simplicity, information corresponding to one user will be described.

  In the conditional probability storage unit 122, a conditional probability indicating a probability that an event corresponding to another node occurs on the condition that an event corresponding to the node of the Bayesian network has occurred is a combination of the node and the other node. It is stored in association with each other. Here, the conditional probability storage unit 122 stores a plurality of conditional probabilities included in the prior information transmitted from the user terminal 400 as a conditional probability table. FIG. 4 is a diagram illustrating a data example of the conditional probability table stored in the conditional probability storage unit 122. In the conditional probability table, a conditional probability that is a probability that an event corresponding to a child node occurs is stored in association with a parent node value indicating whether or not an event corresponding to the parent node has occurred. Has been. For example, as shown in the figure, when the parent node value of the parent node “home” is “1” and the parent node value has occurred, the conditional probability that the child node “I want to watch a movie” occurs is “0.6”. Are stored in association with each other.

  The prior information complementing unit 123 corresponds to a node stored in the prior probability storage unit 121 and corresponding to a node whose value is not included in the prior information transmitted from the user terminal 400 and no prior probability is stored. Is estimated, calculated, and stored in the prior probability storage unit 121. Here, the prior information complementing unit 123 performs a complementing process based on Bayes' theorem. For example, in the Bayes' theorem, if events X and Y exist, the prior probability P (Y) of event Y and the conditional probability P (X | Y) that event X occurs when event Y occurs The prior probability P (X) of event X can be expressed by the quotient divided by the conditional probability P (Y | X) that event Y occurs when event X occurs (P (X)) = P (Y) · P (X | Y) / P (Y | X)).

Here, as shown in FIG. 3, a case is considered where the prior probability of the node “I want to watch a movie” is unknown in the prior probability storage unit 121. Here, the prior probability corresponding to the node “I want to watch a movie” is calculated based on the prior probability corresponding to the node “home” stored in the prior probability table and the node “home” stored in the conditional probability table by Bayes' theorem. ”And the node“ I want to watch a movie ”. In other words, the prior probability (P (I want to watch a moving image)) of the node “I want to watch a moving image” can be calculated by the following formula by Bayes' theorem.
P (I want to watch a movie) = P (Home) / P (I want to watch a movie | Home) / P (Home)

  Here, the prior probability of P (home) is “0.15”, the conditional probability of P (watching a movie | home) is “0.6”, and the condition of P (home | watching like a movie) Since the attached probability is “0.9”, the prior information complementing unit 123 indicates the value of P (seeing a moving image) according to Bayes' theorem (0.15 · 0.6 / 0.9 =) “0”. .1 "is calculated. As described above, the prior information complementing unit 123 calculates an unknown prior probability based on a known conditional probability using Bayes' theorem. The prior information complementing unit 123 stores the calculated prior probability in the prior probability storage unit 121. FIG. 5 is a diagram illustrating a data example of the prior probability table in which the prior probability calculated by the prior information complementing unit 123 is complemented and stored in this manner. Similarly, the prior information complementing unit 123 calculates an unknown conditional probability based on a known prior probability using Bayes' theorem. FIG. 6 is a diagram illustrating a data example of the conditional probability table in which the conditional probability calculated by the prior information complementing unit 123 using the Bayes' theorem is supplemented.

The graph generation unit 130 generates a Bayesian network using the correlation of each node pair. The graph generation unit 130 includes a link selection unit 131 and a direction determination unit 132.
The link selection unit 131 uses a MWST (Maximum Weight Spanning Tree) method based on the prior probability table stored in the prior probability storage unit 121 and the conditional probability stored in the conditional probability storage unit 122. As long as a cyclic route does not generate a link that associates a node pair in descending order from a node pair with a relatively high calculated mutual information amount, calculating the mutual information amount indicating the correlation of the node pair with other nodes for each node. Select the link to be associated. Here, the mutual information (I (X, Y)) indicating the correlation between the node X and the node Y is calculated by the following equation (1).

  For example, when the mutual information amount between the node “home” and the node “working day” is calculated by the above formula (1) based on the data shown in FIGS. 3 to 6, it becomes “0”. In this case, there is no correlation between “home” and “working day”, that is, there is no causal relationship. As will be described later, the link selection unit 131 performs link selection processing using such mutual information of each node pair.

  By such link selection processing, the mutual information such as the node “home” and the node “I want to watch a video”, the node “working day” and the node “I want to watch a video” are calculated, and indicate between nodes having high correlation A link is selected. FIG. 7 is a diagram illustrating an example of a link between nodes selected with respect to FIG. Here, for example, links may be generated between all nodes having a correlation value greater than 0, or links may be generated between nodes having a correlation value equal to or greater than a predetermined threshold.

  The direction determination unit 132 performs link direction determination processing that indicates the direction of the link selected by the link selection unit 131. Here, the direction determination unit 132 sets a node associated with a plurality of links by the link selection unit 131 as a central node, and the mutual information amount falls below a predetermined threshold among a plurality of nodes adjacent to the central node. The direction of the link that associates the node with the central node is the central node direction. For example, the direction determining unit 132 generates a node group in which the mutual information amount is “0” based on the mutual information amount between the nodes for which the links are generated. For example, the node “I want to watch a movie” is connected to both the node “home” and the node “working day”, and the correlation between the node “home” and the node “working day” is 0. The link between them is the link from "Home" and "Working day" to "Watch the video". FIG. 8 is a diagram illustrating the relationship of nodes whose link directions have been determined by the direction determining unit 132. The direction determining unit 132 generates graph data in which the corresponding parent node and child node are associated with each other based on the link selected between the nodes and the link direction in this manner, and stores the graph data in the graph storage unit 210. FIG. 9 is a diagram illustrating an example of graph data generated by the link selection unit 131 and the direction determination unit 132.

  The CPT generation unit 140 generates a conditional probability table (CPT) based on the prior probabilities stored in the prior probability storage unit 121 and the conditional probabilities stored in the conditional probability storage unit 122. It is generated and stored in the CPT storage unit 220. Here, the CPT generation unit 140 includes an NB execution unit 141 that generates a CPT corresponding to graph data by a simple Bayes method (Naive Bayes method) using the correlation of each node pair calculated by the graph generation unit 130. .

  For example, when the conditional probability calculation method has the node “I want to watch a movie” as a child node and the node “home” and the node “work day” as a parent node, the node “home” and the node “work day” Suppose that the node “I want to watch a movie” is conditionally independent. Here, the NB execution unit 141 is p (want to watch a movie = 1, home, work day) = p (want to watch a movie = 1) × p (home | want to watch a movie = 1) × p (working day | I want to watch = 1) = 0.054, p (I want to watch a video = 0, home, work day) p (I want to watch a video = 0) x p (Home | I want to watch a video = 0) x p (work day | I want to watch a video = 0) = 0.02 (here, p (home | watching movie == 0) and p (working date | watching movie = 0) are also obtained by interpolation).

  On the other hand, according to Bayes' theorem, the ratio of p (watching videos = 0, home, working days) to p (home = 0, watching videos, working days) is p (watching videos = 1 | home, working) Day) and p (want to watch video = 0 | home, working day) is equal to the ratio of p (want to watch video = 1 | want to work, working day) ≈0.73 and p (home = 0 | I want to watch a video, work day) ≒ 0.27. In this way, when the conditional probability is similarly obtained for p (want to watch a movie = 1 | home = 0, work day = 0) and the like, the NB execution unit 141, as shown in FIG. And the conditional probability of the node “I want to watch a movie” corresponding to the node “working day”. The NB execution unit 141 generates a CPT including the conditional probability calculated in this way for each node, and stores the CPT in the CPT storage unit 220.

  The Bayesian network storage unit 200 stores the graph data generated by the probability network generation unit 100 and the CPT. A Bayesian network is represented by such graph data and CPT. The Bayesian network storage unit 200 includes a graph storage unit 210 and a CPT storage unit 220.

  The graph storage unit 210 stores graph data indicating a parent-child relationship between a plurality of nodes in the Bayesian network. The graph data stored in the graph storage unit 210 is generated by the graph generation unit 130. FIG. 9 is a diagram illustrating an example of graph data generated by the graph generation unit 130 and stored in the graph storage unit 210.

  The CPT storage unit 220 stores a conditional probability table (CPT) indicating a conditional probability that an event of a child node occurs corresponding to an event of the parent node for each parent-child relationship between nodes in the Bayesian network. The CPT stored in the CPT storage unit 220 is generated by the CPT generation unit 140. FIG. 11 is a diagram illustrating an example of CPT data stored in the CPT storage unit 220. Here, the sequence number is assigned to the parent node in order from 0 for each child node. The pattern number represents the combination pattern of the parent node in binary number, and the value of the parent node is set to the digit of the sequence number. Assigned. Here, the CPT stored in the CPT storage unit 220 and the graph data stored in the graph storage unit 210 may be accommodated in the same database so that the table is efficiently normalized.

The estimation unit 300 includes an application determination unit 310, an observation value storage unit 320, and an observation value reception unit 330.
The application determination unit 310 is based on the observation value stored in the observation value storage unit 320, the CPT stored in the CPT storage unit 220, and the graph data stored in the graph storage unit 210, and the observation value reception unit 330. The other event corresponding to the event indicated by the observation value received by is extracted. In addition, the application determination unit 310 determines an application server 500 that provides an information service corresponding to the extracted event, and transmits an information service recommendation request to the determined application server 500. For example, in the storage area of the application determination unit 310, identification information for identifying the application server 500 that provides an information service corresponding to an event is stored in association with each node stored in the Bayesian network.

The observation value storage unit 320 stores an observation value indicating whether or not an event corresponding to the Bayesian network has occurred. The observation value storage unit 320 stores observation values received by the observation value reception unit 330 from the sensor terminal 600. FIG. 12 is a diagram illustrating an example of observation value data stored in the observation value storage unit 320. Here, an observation value “1” corresponding to the node “home” and an observation value “0” corresponding to the node “working day” are stored.
The observation value receiving unit 330 receives the observation value transmitted from the sensor terminal 600 and stores it in the observation value storage unit 320.

  The application server 500 is a computer device that provides a defined information service. For example, when the application determination unit 310 of the information processing apparatus 10 determines that there is a high probability that “watch a video” has occurred in the current event of the user, the video information is distributed among the plurality of application servers 500. An information service recommendation request is transmitted to the application server 500. When the information service recommendation request is transmitted from the application determination unit 310 of the information processing apparatus 10, the application server 500 transmits a provision menu screen that distributes moving image information to the corresponding user terminal 400, for example. On the user terminal 400, a moving image distribution menu screen transmitted from the application server 500 is displayed. Thereby, the user of the user terminal 400 can receive the recommendation of the information service according to the current own state.

  The sensor terminal 600 is a computer terminal that acquires data indicating the state of the user, mood, and the like and transmits the data to the observation value receiving unit 330. The sensor terminal 600 includes, for example, a vital sensor that acquires data such as a user's body temperature and blood pressure, a mobile phone terminal that has a GPS function that acquires user's position data, an acceleration sensor that acquires data indicating the moving speed of the user, An illumination sensor, an input terminal to which user schedule data is input, and the like can be applied.

Next, an operation example of the information processing system 1 will be described. FIG. 13 is a sequence diagram illustrating an operation example of the information processing system 1.
Here, it is assumed that nodes as shown in FIG. 2 are defined in advance and stored in the prior probability storage unit 121. The information processing apparatus 10 discloses the functions of the prior information acquisition unit 110 and the observation value reception unit 330, the prior information acquisition unit 110 starts accepting data from the user terminal 400, and the observation value reception unit 330 Data reception from the sensor terminal 600 is started. The probability network creation unit 100 receives prior information transmitted from the user terminal 400 (step S100).

  The prior information acquisition unit 110 stores the prior probability included in the prior information received from the user terminal 400 in the prior probability storage unit 121, and stores the conditional probability in the conditional probability storage unit 122 (step S110). When the prior information acquisition unit 110 notifies the prior information complement unit 123 that the prior information has been received from the user terminal 400 (step S120), the prior probability storage unit 121 may include the prior probability storage unit 121 or the conditional probability storage unit 122. Among the information stored in the prior probability table or the conditional probability table stored in step S130, the missing information is complemented (step S130).

  When the advance information acquisition unit 110 is notified that the completion process by the advance information complement unit 123 has been completed (step S140), the advance information acquisition unit 110 makes a link creation request to the graph generation unit 130 (step S150). The graph generation unit 130 reads the prior probability and the conditional probability with reference to the prior probability storage unit 121 and the conditional probability storage unit 122 (step S160). The graph generation unit 130 performs link selection processing by the MWST method and link direction determination processing (step S200).

  FIG. 14 is a flowchart showing in detail an operation example of link selection processing by the MWST method. The link selection unit 131 of the graph generation unit 130 has not selected all links, and starts link selection processing using all links as selection candidates (step S201). The link selection unit 131 calculates the mutual information amount of each node pair (step S202), and sorts the calculated mutual information amount in descending order (step S203). The link selection unit 131 temporarily selects a candidate whose mutual information amount is not 0 and is the largest among the links included in the candidate list, and deletes the candidate from the candidate list (step S204).

  And the link selection part 131 determines whether a candidate exists in a candidate list, and when it determines with there not existing, a process is complete | finished (step S205: No). On the other hand, if it is determined that it exists (step S205: Yes), it is determined whether an undirected loop exists in the graph of the temporary selection link and the selection link (step S206). If the link selection unit 131 determines that an undirected loop exists (step S206: Yes), the temporary selection link is deselected (step S207), and the process returns to step S204.

  On the other hand, if the link selection unit 131 determines in step S206 that there is no undirected loop (step S206: No), 1 is added to the number of selected links (step S208). Then, the link selection unit 131 determines whether or not the number of selected links matches the value of the number of nodes−1 (step S209). If the link selection unit 131 determines that the number of selected links matches the value of the number of nodes−1 (step S209: Yes), the link selection unit 131 ends the process. If it is determined that the number of selected links does not match the value of the number of nodes −1 (step S209: No), the temporarily selected link is set as the selected link (step S210), and the process returns to step S204.

FIG. 15 is a flowchart illustrating an operation example of the link direction determination process by the direction determination unit 132 of the graph generation unit 130.
First, the direction determination unit 132 acquires from the link selection unit 131 a list of invalid links selected by the link selection unit 131 and a mutual information amount list between nodes (step S220). The direction determining unit 132 generates a list of nodes adjacent to the plurality of nodes (step S221). Further, the direction of all links is undetermined (step S222). Then, the processing from step S224 to step S225 is repeated for each node adjacent to the plurality of nodes. In step S224, the direction determining unit 132 extracts an adjacent node having a correlation of 0 between adjacent nodes (step S224). Then, the direction determining unit 132 determines the link direction as the repeated central node direction from the extraction node (step S225).

  After repeating the processing from step S223 to step S225 for each node adjacent to the plurality of nodes, the direction determining unit 132 determines whether the directions of all the links have been determined (step S226). Here, when the direction determination unit 132 determines that the directions of all the links have been determined (step S226: Yes), the process ends. On the other hand, if it is determined that the directions of all the links have not been determined (step S226: No), the direction undecided links are extracted (step S227), and the number of direction links is determined among the nodes connected to the extracted links. The largest node is extracted (step S228). Then, the direction determination unit 132 determines the extraction node as a link destination for the link connected to the extraction node (step S229), and returns to step S226.

  Returning to FIG. 13, when link selection processing and link direction determination processing are performed, a graph based on the created link structure is stored in the graph storage unit 210. Then, the CPT generation unit 140 is based on the graph structure generated by the graph generation unit 130, the prior probability stored in the prior probability storage unit 121, and the conditional probability stored in the conditional probability storage unit 122. A CPT is generated (step S310) and stored in the CPT storage unit 220 (step S320). In this way, a Bayesian network is created.

  On the other hand, the user of the user terminal 400 makes a sensor service contract with the system administrator of the information processing system 1 (step S400). In addition, when the user of the user terminal 400 makes a contract with the system administrator of the information processing system 1 for the application server (step S410), the application server 500 starts a subscription reservation of the user situation distributed from the estimation unit 300. (Step S420). When the sensor terminal 600 transmits the generated observation value to the estimation unit 300 (step S430), the estimation unit 300 reads the Bayesian network from the graph storage unit 210 and the CPT storage unit 220, and displays the observation value on the read Bayesian network. Collectively, the event corresponding to the event of the user situation is estimated and determined (step S450). The estimation unit 300 notifies the application server 500 of the determined user event. The application server 500 transmits an information service corresponding to the user event to the user terminal 400 (step S470).

  In this case, for example, the sensor terminal 600, which is a mobile phone terminal equipped with a GPS function and a scheduler, uses the GPS function, and the current state of the user is “home”, and the user is not “working day” by the scheduler. When the observed value indicating that the value is 0 is transmitted to the observed value receiving unit 330, the transmitted observed value is stored in the observed value storage unit 320. Based on the observation value stored in the observation value storage unit 320 and the Bayesian network stored in the Bayesian network storage unit 200, the application determination unit 310 estimates the probability of “watching a video” by the Belief Propagation method or the like. As can be seen from the CPT in FIG. 10, the probability is estimated to be 0.47, and no moving picture is recommended. Further, when time passes and the mobile phone notifies that it is a “working day”, the probability of “watching a video” becomes 0.73, which is notified to the application server 500, and the video information service is provided. Recommended. In addition, although the observation value received from the sensor terminal 600 has been described as being a binary value of 0 or 1, an occurrence probability such as 0.3 may actually be received.

In the present embodiment, the example in which the information processing apparatus 10 includes the probability network generation unit 100, the Bayesian network storage unit 200, and the estimation unit 300 has been described. May be provided. FIG. 16 is a diagram illustrating an example in which the user terminal 1400 includes a probability network generation unit 1410, a graph storage unit 1420, and an estimation unit 1430. Each unit provided in the user terminal 1400 has the same function as each unit having the same name provided in the information processing apparatus 10. FIG. 17 is a diagram illustrating an example in which the application server 1500 includes a probability network generation unit 1510, a graph storage unit 1520, and an estimation unit 1530. Each part with which application server 1500 is provided is the same function as each part with the same name with which information processor 10 is provided.
Note that this embodiment is merely an embodiment of the present invention, and does not limit the scope of the claims.

  By using the above invention, the Bayesian network generator only has to input the correlation of each node pair regardless of the structure of the graph, and it is possible to prevent the number of setting parameters from increasing exponentially. In addition, since only the correlation of each node pair should be input, input that requires specialized knowledge considering the combination of multiple conditions becomes unnecessary, making entry easier and general users with no specialized knowledge Etc. can also generate Bayesian networks.

  In addition, there are different user preferences / situations that should be estimated for each user even if the observed values from the sensors are the same. For example, even if the observation value “stopped” is obtained from the acceleration sensor, the situation varies depending on the user, such as “free time” or “busy”, and there are services that you want to use, such as “want to watch videos” or “want to collect information” Different. Therefore, a Bayesian network for estimating these needs to be generated for each user. Even in such a case, according to the present embodiment, since a Bayesian network for each user can be prepared in advance, it is possible to recommend an information service according to an event for each user.

  As described above, according to the present invention, when constructing a Bayesian network, link selection is performed by a method of approximately extracting a node group having a direct correlation using the MWST method or the like. By using the average occurrence probability (prior probability) of some nodes in addition to the correlation, the conditional probability is statistically obtained by simple Bayes or the like to facilitate the construction of the Bayesian network. As a result, a generator (user) who manually constructs a Bayesian network selects a link indicating a direct cause-and-effect relationship for a given node group and a parameter indicating the strength of the cause-and-effect relationship (a link between nodes by link selection). There is no need to set items and numbers depending on the structure that occurs. That is, link selection in a Bayesian network requires extraction of node pairs having a direct causal relationship, and is difficult to execute unless a node group can be looked down on. Therefore, a creator who does not have specialized knowledge about a node group cannot realize link selection. Therefore, the information processing system 1 according to the present embodiment acquires the causal relationship of each node pair from the creator without distinguishing whether it is direct or indirect, and is in a direct correlation from these relationships. A node pair is approximately extracted using the MWST method, and for a node having a direct correlation with a plurality of nodes, the direction of the link with the nodes (the origin of the link) is determined based on the presence or absence of correlation between surrounding nodes. (See G. Rebane, J. Pearl, "The Recovery of Causal Poly Trees From Statistical Data,"). This allows Bayesian network creators to make link selections without knowing whether the relationship is direct / indirect and which of the node pairs with the relationship is the cause / result.

Further, in parameter setting, it is necessary to set conditional probabilities when all the combinations of values that can be taken by the cause node (hereinafter, parent group) group for each node are observed. For example, if the link is selected so that “I want to collect information” is the result of “free time”, “business hours”, or “going out”, the “information collection” during “free time, business hours and going out” The CPT is made up of a set of probabilities that result in all combinations of causes, such as the probability of “I want to collect information” and the probability of “I want to collect information” when “I want to collect information” when “I am not free, work hours, and not out.” Therefore, in a situation where the number of parent groups increases, the number of settings increases dramatically, and it is necessary to assume a combination of parent group values in CPT generation, which requires specialized techniques. . Therefore, the system also assists in parameter setting. In addition to the correlation of each node pair similar to link selection, the average occurrence probability (prior probability) of some nodes is used to statistically calculate the CPT using the simple Bayes method or the like. Ask for. The correlation between node pairs here refers to, for example, conditional probabilities in the case of “free time” and “want to collect information”, which are the causes and results of each other, where a single cause becomes a conditional probability condition. It is different from the conventional one.
As a result, in this embodiment, a link selection and CPT generation is automatically generated on the system side only from the correlation of the node pairs and the prior probabilities of some nodes, thereby generating a Bayesian network even for users who do not have expertise. This makes it possible to easily generate a Bayesian network for each user.

  The program for realizing the function of the processing unit in the present invention is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read into the computer system and executed to construct a Bayesian network. May be performed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer system” includes a WWW system having a homepage providing environment (or display environment). The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) in a computer system that becomes a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those holding programs for a certain period of time are also included.

  The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. 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, and what is called a difference file (difference program) may be sufficient.

DESCRIPTION OF SYMBOLS 1 Information processing system 10 Information processing apparatus 100 Probability network creation part 110 Prior information acquisition part 120 Prior information holding part 121 Prior probability storage part 122 Conditional probability storage part 123 Prior information complement part 130 Graph generation part 131 Link selection part 132 Direction determination Unit 140 CPT generation unit 141 NB execution unit 200 Bayesian network storage unit 210 graph storage unit 220 CPT storage unit 300 estimation unit 310 application determination unit 320 observation value storage unit 330 observation value reception unit 400 user terminal 500 application server 600 sensor terminal

Claims (10)

An information processing apparatus for generating a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing a user state and a causal relationship between the plurality of events,
A prior probability storage unit in which a prior probability indicating a probability that an event corresponding to the node occurs is associated and stored for each node;
Conditional probabilities indicating the probability that an event corresponding to another node occurs on condition that an event corresponding to the node has occurred are stored in association with each combination of the node and the other node. A conditional probability storage unit;
A graph storage unit that stores graph data indicating a parent-child relationship between the plurality of nodes;
A conditional probability table storage unit that stores a conditional probability table indicating a conditional probability that an event of a child node occurs corresponding to an event of the parent node for each parent-child relationship;
The prior information including the prior probability and the conditional probability is received, the prior probability included in the received prior information is stored in the prior probability storage unit, and the received prior information is stored in the prior information. A prior information acquisition unit for storing the conditional probability included in the conditional probability storage unit;
Based on the prior probability stored in the prior probability storage unit and the conditional probability stored in the conditional probability storage unit, the conditional probability table is generated to store the conditional probability table. A conditional probability table generator to be stored in
An information processing apparatus comprising: Based on the prior probability stored in the prior probability storage unit and the conditional probability stored in the conditional probability storage unit, the graph is generated and stored in the graph storage unit The information processing apparatus according to claim 1, further comprising: a generation unit. The graph generation unit
Based on the prior probability and the conditional probability, a mutual information amount indicating a correlation of a node pair with another node is calculated for each of the plurality of nodes, and the calculated mutual information amount is a relatively high node A link selection unit for associating the link that associates the node pair in descending order with each other unless a cyclic route occurs; and
A node in which a plurality of links are associated by the link selection unit as a central node, and among the plurality of nodes adjacent to the central node, the node whose mutual information amount is below a predetermined threshold, and the central node, A direction determination unit that sets the direction of the link that associates with the direction of the central node,
The information processing apparatus according to claim 2, wherein the graph data is generated based on a link between the nodes and a direction of the link. Of the plurality of nodes stored in the prior probability storage unit or the conditional probability storage unit, prior information corresponding to the node for which prior information has not been received by the prior information acquisition unit, An advance information complementing unit that generates based on the prior information received by the information acquisition unit and stores the generated prior information in the prior probability storage unit or the conditional probability storage unit is provided. The information processing apparatus according to any one of claims 1 to 3. The information processing apparatus according to any one of claims 1 to 4, wherein the conditional probability table generation unit generates the conditional probability table using a simple Bayes method. An observation value storage unit in which an observation value indicating whether or not the event has occurred is stored;
An observation value receiving unit that receives the observation value and stores the received observation value in the observation value storage unit;
Based on the observation value stored in the observation value storage unit, the conditional probability table stored in the conditional probability table storage unit, and the graph data stored in the graph storage unit, the observation An estimation unit that extracts another event corresponding to the event indicated by the observation value received by the value receiving unit;
The information processing apparatus according to any one of claims 1 to 5, further comprising: An information processing system that provides an information service corresponding to a state of the user based on a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing the state of the user and a causal relationship between the plurality of events Because
A prior probability storage unit in which a prior probability indicating a probability that an event corresponding to the node occurs is associated and stored for each node;
Conditional probabilities indicating the probability that an event corresponding to another node occurs on condition that an event corresponding to the node has occurred are stored in association with each combination of the node and the other node. A conditional probability storage unit;
A graph storage unit that stores graph data indicating a parent-child relationship between the plurality of nodes;
A conditional probability table storage unit that stores a conditional probability table indicating a conditional probability that an event of a child node occurs corresponding to an event of the parent node for each parent-child relationship;
The prior information including the prior probability and the conditional probability is received, the prior probability included in the received prior information is stored in the prior probability storage unit, and the received prior information is stored in the prior information. A prior information acquisition unit for storing the conditional probability included in the conditional probability storage unit;
Based on the prior probability stored in the prior probability storage unit and the conditional probability stored in the conditional probability storage unit, the conditional probability table is generated to store the conditional probability table. A conditional probability table generator to be stored in
An observation value receiving unit that receives the observation value and stores the received observation value in the observation value storage unit;
Based on the observation value stored in the observation value storage unit, the conditional probability table stored in the conditional probability table storage unit, and the graph data stored in the graph storage unit, the observation An estimation unit that extracts another event corresponding to the event indicated by the observation value received by the value receiving unit;
An information processing apparatus comprising:
A user terminal that transmits the prior information input from the user to the information processing apparatus;
A sensor terminal that generates an observation value corresponding to the event and transmits the observation value to the information processing apparatus;
An application server that transmits an information service corresponding to the other event extracted by the information processing apparatus to the user terminal;
An information processing system comprising: An information processing system that provides an information service corresponding to a state of the user based on a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing the state of the user and a causal relationship between the plurality of events Because
A prior probability storage unit in which a prior probability indicating a probability that an event corresponding to the node occurs is associated and stored for each node;
A conditional probability indicating the probability that an event corresponding to another node will occur on condition that an event corresponding to the node has occurred is stored in association with each combination of the node and the other node. A conditional probability storage unit;
A graph storage unit that stores graph data indicating a parent-child relationship between the plurality of nodes;
A conditional probability table storage unit that stores a conditional probability table indicating a conditional probability that an event of a child node occurs corresponding to an event of the parent node for each parent-child relationship;
Receiving prior information including the prior probability and the conditional probability input from the user, storing the prior probability included in the input prior information in the prior probability storage unit, and receiving A prior information acquisition unit that stores the conditional probability included in the prior information in the conditional probability storage unit;
Based on the prior probability stored in the prior probability storage unit and the conditional probability stored in the conditional probability storage unit, the conditional probability table is generated to store the conditional probability table. A conditional probability table generator to be stored in
An observation value receiving unit that receives the observation value and stores the received observation value in the observation value storage unit;
Based on the observation value stored in the observation value storage unit, the conditional probability table stored in the conditional probability table storage unit, and the graph data stored in the graph storage unit, the observation An estimation unit that extracts another event corresponding to the event indicated by the observation value received by the value reception unit and receives the information service from an application server that provides an information service corresponding to the extracted event;
A user terminal comprising:
A sensor terminal that generates an observation value corresponding to the event and transmits the observation value to the user terminal;
An application server that transmits to the user terminal an information service corresponding to the other event extracted by the user terminal;
An information processing system comprising: An information processing system that provides an information service corresponding to a state of the user based on a Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing the state of the user and a causal relationship between the plurality of events Because
A prior probability storage unit in which a prior probability indicating a probability that an event corresponding to the node occurs is associated and stored for each node;
Conditional probabilities indicating the probability that an event corresponding to another node occurs on condition that an event corresponding to the node has occurred are stored in association with each combination of the node and the other node. A conditional probability storage unit;
A graph storage unit that stores graph data indicating a parent-child relationship between the plurality of nodes;
A conditional probability table storage unit that stores a conditional probability table indicating a conditional probability that an event of a child node occurs corresponding to an event of the parent node for each parent-child relationship;
The prior information including the prior probability and the conditional probability is received, the prior probability included in the received prior information is stored in the prior probability storage unit, and the received prior information is stored in the prior information. A prior information acquisition unit for storing the conditional probability included in the conditional probability storage unit;
Based on the prior probability stored in the prior probability storage unit and the conditional probability stored in the conditional probability storage unit, the conditional probability table is generated to store the conditional probability table. A conditional probability table generator to be stored in
An observation value receiving unit that receives the observation value and stores the received observation value in the observation value storage unit;
Based on the observation value stored in the observation value storage unit, the conditional probability table stored in the conditional probability table storage unit, and the graph data stored in the graph storage unit, the observation An estimation unit that extracts another event corresponding to the event indicated by the observation value received by the value reception unit, and transmits an information service corresponding to the extracted event to the user terminal;
An application server comprising:
A user terminal that transmits the prior information input from the user to the application server;
A sensor terminal that generates an observation value corresponding to the event and transmits the observation value to the application server;
An information processing system comprising: A Bayesian network including information on a plurality of nodes corresponding to a plurality of events representing a user state and a causal relationship between the plurality of events is generated, and an event corresponding to the node occurs for each node. A prior probability storage unit in which a prior probability indicating a probability is associated and stored, and a conditional probability indicating a probability that an event corresponding to another node occurs on the condition that an event corresponding to the node has occurred, A conditional probability storage unit that is stored in association with each combination of the node and the other node, a graph storage unit that stores graph data indicating a parent-child relationship between the plurality of nodes, and the parent-child relationship A conditional probability table storage unit that stores a conditional probability table that indicates a conditional probability that an event of a child node occurs corresponding to the event of the parent node A management method,
The prior information including the prior probability and the conditional probability is received, the prior probability included in the received prior information is stored in the prior probability storage unit, and the received prior information is stored in the prior information. Storing the conditional probability included in the conditional probability storage unit;
Based on the prior probability stored in the prior probability storage unit and the conditional probability stored in the conditional probability storage unit, the conditional probability table is generated to store the conditional probability table. Storing it in a part;
An information processing method comprising:

Images