JP5364732B2 - Content selection candidate extraction device, content selection candidate extraction method, behavior prediction device, and behavior prediction method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To widen a possibility for a user to select any selection candidate of contents by more accurately inferring the action of the user in simple configuration and varying selection candidates of the user action or a destination to be presented to the user. <P>SOLUTION: A car navigation system 10 performs action prediction processing for predicting the action of a user on the basis of an action history, an action habit that is a feature of an action which can be statistical from the action history, and general common sense stored in a general common sense data table 12e in accordance with user preference and user attributes. The car navigation system 10 further performs preference estimation processing for predicting user preference related to user destination selection on the basis of a category selection tendency, and an estimated category estimated based on abstraction characteristics stored in a category estimation table 12i in accordance with the user preference and the user attributes. <P>COPYRIGHT: (C)2012,JPO&amp;INPIT

Description

  The present invention relates to a content selection candidate extraction apparatus, a content selection candidate extraction method, and a behavior that extract content selection candidates that can be selected by a user under a predetermined situation from content that can be classified by at least one classification item, and present the same to the user The present invention relates to a prediction device and a behavior prediction method.

  A car navigation device, which is an in-vehicle device mounted on a vehicle, provides information on a destination as content and information on the destination and destination information on the destination. The destination is essential information for the route search in the car navigation device together with the departure place of the vehicle. Here, it is important from the viewpoint of convenience of content selection to select and present to the user the optimum content (destination) desired by the user from the enormous amount of content.

  As an example of the content selection method, it is possible to predict a user's behavior more accurately, extract content selection candidates that are highly likely to be selected based on the prediction, and present the extracted selection candidates to the user. For example, Patent Document 1 discloses a user who can estimate user information such as a user's environment, situation, state, request, behavior, and situation in detail and more accurately in consideration of the influence given by time and space. An information estimation device is disclosed.

  Further, in Patent Document 2, in the car navigation device, the starting point and the destination by time in the route search performed in the past are stored in the past history together with information on the route search frequency, and the car navigation device is activated. Discloses a car navigation device that searches for a starting point and a destination having the highest route search frequency in the time zone in the past history and uses the starting point and the destination for the current route search.

JP 2002-92029 A JP-A-8-271277

  However, in the related art represented by the above-mentioned Patent Document 1, although it is possible to more accurately predict the user's behavior and the like, a complicated calculation process that requires many storage resources and high computing power in the computer device Therefore, there are problems of manufacturing cost due to a complicated configuration and a problem that processing time is required.

  Further, in the conventional technique represented by the above-mentioned Patent Document 2, since the start point and the destination with the highest setting frequency in the past history are acquired without considering other factors considering only the time zone, it is absolutely There is a possibility that a starting point and a destination that cannot be selected may be acquired, and there is a possibility that the convenience of guessing and presenting the user's destination using the past history may be lost.

  Further, even if the conventional techniques represented by Patent Documents 1 and 2 described above are combined, user actions and destinations are estimated based on user information and past history, so that candidates such as user actions and destinations are narrowed down. The selection candidates such as user actions and destinations presented to the user are only a subset of the user information and past history, so there are few variations, and the possibility of user selection has not been expanded. .

  The present invention has been made to solve the above problems (problems), and more accurately infers the user's behavior with a simple configuration and presents the user behavior, destination, etc. presented to the user. It is an object of the present invention to provide a content selection candidate extraction device, a content selection candidate extraction method, a behavior prediction device, and a behavior prediction method that allow a selection candidate to have a variation and expand the possibility of user selection.

  In order to solve the above-described problems and achieve the object, the present invention provides a content selection candidate that can be selected by a user under a predetermined situation from content that can be classified by at least one classification item based on an input estimation condition. And a selection history recording unit that records a selection history of the content selected by the user, and the selection history recorded by the selection history recording unit. Based on the periodicity extracting means for extracting the periodicity latent in the content selection action by the user, the selection history, and the periodicity extracted by the periodicity extracting means, by the user A selection candidate extracting unit that extracts selection candidates of content that can be selected, and the selection candidate extracting unit. In order to present the extracted content as the selection candidate to the user in order of priority, based on the consistency of the specific data of the classification item of the content with respect to the estimation condition, the periodicity, or the number of selections on the day of the content It has a priority provision means which provides a priority, It is characterized by the above-mentioned.

  Further, the present invention is the above invention, wherein the selection candidate extraction unit extracts the selection candidate based on a common sense condition that defines a situation in which the content cannot be the selection candidate in a common sense under the predetermined situation. It further includes selection candidate exclusion means for excluding content from the selection candidates extracted by the means.

  Further, the present invention provides the above-described invention in which an abstraction characteristic that abstracts and characterizes the specific data of the classification item corresponds to the specific data of the classification item of the content extracted as the selection candidate by the selection candidate extraction unit. Analogical classification item acquisition means for acquiring, as an analogy classification item, specific data of the classification item of other content whose similarity with the abstraction characteristic of the classification item is equal to or greater than a predetermined value based on the abstraction characteristic table to be added and stored The priority assigning means includes the content extracted as the selection candidate by the selection candidate extraction means and the content specified by the analogy classification item acquired by the analogy classification item acquisition means. Further, priority is given based on the selection tendency, purpose, characteristic factor, or attribute factor.

  In addition, the present invention extracts content selection candidates that can be selected by a user under a predetermined situation from content that can be classified by at least one classification item based on an input estimation condition, and selects content to be presented to the user A selection history recording step of recording a selection history of content selected by the user, and a selection action of the content by the user based on the selection history recorded by the selection history recording step Selection for extracting selection candidates of content that can be selected by the user based on the periodicity extraction step for extracting periodicity latent in the user, the selection history, and the periodicity extracted by the periodicity extraction step A candidate extraction step and, under the predetermined circumstances, the content is common sense A selection candidate exclusion step for excluding content from the selection candidate extracted by the selection candidate extraction step based on common sense conditions that define a situation that cannot be a selection candidate, and the selection candidate by the selection candidate extraction step The abstraction of the classification item based on the abstraction characteristic table that stores the abstract data that abstracts and characterizes the specific data of the classification item in association with the specific data of the classification item of the content extracted as An analogy classification item acquisition step of acquiring specific data of the classification item of other content having characteristics and similarity equal to or higher than a predetermined value as an analogy classification item, content extracted as the selection candidate by the selection candidate extraction step, and The container specified by the analogy category item acquired by the analogy category item acquisition step. In order to present to the user in order of priority, the consistency of the specific data of the content classification item with respect to the estimation condition, the periodicity, the number of times of selection of the content on the day, the selection tendency for the content, And a priority assigning step for assigning a priority based on a feature factor or an attribute factor of the content.

  Further, the present invention predicts the user's behavior specified by the classification item of the selection candidates of content that can be selected by the user from the content that can be classified by at least one classification item based on the input estimation condition. A behavior prediction device, a selection history recording unit that records a selection history of content selected by the user, and an action of selecting the content by the user based on the selection history recorded by the selection history recording unit The classification item of a selection candidate of content that can be selected by the user based on the periodicity extracting means for extracting periodicity latent in the user, the selection history, and the periodicity extracted by the periodicity extracting means The category item extracting means for extracting the category item extracted by the category item extracting means, Priority giving means for giving priority based on the consistency of the specific data of the classification item of the content with respect to the estimation condition, the periodicity, or the number of selections on the day of the content to present to the user in order of priority; It is characterized by having.

  Also, in the present invention according to the above invention, the category item extracting unit may not commonly make the content corresponding to the category item extracted by the category item extracting unit as the selection candidate under the predetermined situation. The method further includes a classification item excluding unit that excludes the classification item from the selection candidates based on a common sense condition that defines the situation.

  Further, the present invention predicts the user's behavior specified by the classification item of the selection candidates of content that can be selected by the user from the content that can be classified by at least one classification item based on the input estimation condition. A behavior prediction method, a selection history recording step for recording a selection history of content selected by the user, and an action for selecting the content by the user based on the selection history recorded by the selection history recording step The classification item of the selection candidates of content that can be selected by the user based on the periodicity extracting step for extracting periodicity latent in the user, the selection history, and the periodicity extracted by the periodicity extracting step A category item extracting step for extracting the category item and extracting the category item under the predetermined situation A classification item exclusion step of excluding a classification item from the selection candidate based on common sense conditions that define a situation in which content corresponding to the classification item extracted by the step cannot be the selection candidate in common sense; In order to present to the user in order of priority the classification items extracted by the classification item extraction step and not excluded from the selection candidates by the classification item exclusion step, the specifics of the classification items of the content with respect to the estimation condition Prioritizing step of giving priority based on data consistency, the periodicity, the number of times of selection of the content on the day, or a selection tendency for the content, the purpose of the content, a characteristic factor or an attribute factor of the content It is characterized by including.

  Further, the present invention extracts specific data of content classification items that can be selected by a user under a predetermined situation based on input of specific data of the classification items of content that can be classified by at least one classification item. A classification item content estimation device to be presented to a user, wherein an abstraction characteristic table for storing the inputted specific data of the classification item in association with an abstract characteristic that abstracts and characterizes the specific data of the classification item Based on the analogy classification item acquisition means for acquiring the specific data of the classification item of other content whose similarity and the abstract property of the classification item are not less than a predetermined value as analogy classification items, and by the analogy classification item acquisition means In order to present to the user in order of priority the content specified by the specific data of the acquired classification items of other content, Selection tendency towards Ntsu, the purpose of the content, and having a priority assigning means for assigning priorities on the basis of the characteristic factors or attributes factors of the content.

  The present invention further includes a selection category item recording unit that records specific data of the category item of the content selected by the user in the above invention, and the category item recorded in the selection category item recording unit In this case, specific data of content classification items that can be selected by the user under the predetermined condition is extracted and presented to the user.

  Further, the present invention extracts specific data of content classification items that can be selected by a user under a predetermined situation based on input of specific data of the classification items of content that can be classified by at least one classification item. A classification item content estimation method to be presented to a user, wherein an abstraction characteristic table that stores the inputted specific data of the classification item in association with an abstract characteristic that abstracts and characterizes the specific data of the classification item Based on the analogy classification item acquisition step of acquiring specific data of the classification item of the other content whose similarity with the abstraction characteristic of the classification item is equal to or greater than a predetermined value as an analogy classification item, and the analogy classification item acquisition step In order to present to the user in order of priority the content specified by the specific data of the acquired classification items of other content Selection tendency for the content, the purpose of the content, characterized in that it comprises a priority giving step of giving a priority on the basis of the characteristic factors or attributes factors of the content.

  According to the present invention, it is possible to easily perform a content selection operation by extracting content selection candidates that can be selected by the user and presenting them to the user.

  Further, according to the present invention, by excluding contents that cannot be selected by common sense from the selection candidates in advance, the user can easily perform a selection operation from a list consisting only of contents that are truly selectable, There is an effect that it is possible to prevent unintentional selection of content.

  In addition, according to the present invention, the user can select content by expanding the selectable target to similar content based on other classification items inferred from specific data of content classification items selected in the past. Thus, the content can be selected from a wide range.

  According to the present invention, it is possible for the user to take corresponding content by extracting the content that can be selected by the user based on the past content selection history and the estimation condition and arranging them in order of priority under the predetermined condition. There is an effect that it is possible to estimate a behavior with high sex.

  Further, according to the present invention, it is possible to prevent a user from predicting an action that cannot be taken by previously excluding contents that cannot be selected by common sense from selection candidates.

  According to the present invention, since other classification items are inferred from the specific data of the classification item of the content selected in the past, it is possible to expand to similar content and list the content from the list and present it to the user. The effect of becoming.

  In addition, according to the present invention, it is possible to analogize similar classification items by inputting specific data of content classification items selected in the past by the user.

FIG. 1 is a diagram illustrating an outline of processing of the car navigation device according to the embodiment. FIG. 2 is a diagram illustrating an example of the estimation condition. FIG. 3 is a functional block diagram illustrating the configuration of the car navigation apparatus according to the embodiment. FIG. 4 is a diagram illustrating a table example of the destination content table. FIG. 5 is a diagram illustrating an example of a user profile table. FIG. 6 is a diagram illustrating a table example of the action history table. FIG. 7 is a diagram illustrating a table example of the behavior habit table. FIG. 8 is a diagram illustrating a table example of the general common sense data table. FIG. 9 is a diagram illustrating a table example of the same day selection history table. FIG. 10 is a diagram illustrating a table example of a genre selection tendency table. FIG. 11 is a diagram illustrating an example of a temporary table obtained by extracting the latest N records from the action history table. FIG. 12 is a diagram illustrating a situation in which the “specialization flag” is turned on. FIG. 13 is a diagram illustrating a table example of the genre analogy table. FIG. 14 is a diagram illustrating an example of the attribute value weighting table. FIG. 15 is a diagram illustrating a calculation example of the priority score. FIG. 16 is a diagram illustrating an example of a recommendation information table. FIG. 17 is a flowchart illustrating a behavior prediction processing procedure executed by the car navigation device according to the embodiment. FIG. 18 is a flowchart illustrating a preference estimation processing procedure executed by the car navigation device of the embodiment. FIG. 19 is a diagram illustrating a display example of recommended information.

  Exemplary embodiments according to a content selection candidate extraction device, a content selection candidate extraction method, a behavior prediction device, and a behavior prediction method according to the present invention will be described below in detail with reference to the accompanying drawings. In the following embodiments, the content selection candidate extraction device, the content selection candidate extraction method, the behavior prediction device, and the behavior prediction method according to the present invention are shown as a car navigation device mounted on a vehicle, In order to select the information associated with the destination as the content, a selection list is displayed that ranks the destination that is expected to be selected by predicting the destination setting behavior of the driver of the vehicle. We will show the case.

  However, the present invention is not limited to this. That is, a content selection list display device that displays a selection list by positioning content that is expected to be selected by a user based on a user's selection tendency or the like based on a user's selection tendency or the like. Applicable to general equipment. The content selection list display device may be integrated with a general content providing device.

  First, an outline of processing of the car navigation device according to the embodiment will be described. FIG. 1 is a diagram illustrating an outline of processing of the car navigation device according to the embodiment. In the following embodiments, “behavior” refers to an action of a user reaching a “destination” by a vehicle. As shown in the figure, the car navigation device 10 outputs “recommended information” that is a list of destinations recommended as the car navigation device 10 with “estimation conditions” as input information.

  As shown in FIG. 2, the “estimation condition” includes “target time zone”, “date (including day of the week)”, “weather”, and “selection history of the day”. These pieces of information are automatically acquired from the system of the car navigation device 10 and / or from outside the system. However, the “target time zone”, “date (including day of the week)”, and “weather” may be input by the user using an input device (not shown).

  The “target time zone” is information for setting a “time zone” which is one of the attributes of “recommended information” to be output. The time zone can be acquired based on the system time of the car navigation device 10. In the embodiment, there are five “time zones”, for example, “morning”, “morning”, “afternoon”, “night”, and “midnight”. These five time zones are ordered in ascending order of “morning”, “am”, “afternoon”, “night”, and “midnight”.

  For example, “morning” is a time zone from 4:00 to 9:00, “morning” is a time zone from 9:00 to 12:00, and “afternoon” is from 12:00 to 18:00. It is assumed that “night” is a time zone from 18:00 to 23:00, and “midnight” is a time zone from 23:00 to 4:00 on the next day. “Recommended information” has a time zone suitable for being set as a destination. For example, “a bay area with a beautiful night view” is “night” in the above-mentioned time zone, which is suitable for being selected as a destination.

  “Date (including day of the week)” is information indicating a date and a day of the week. “Date (including day of week)” can be acquired based on the system calendar of the car navigation apparatus 10. “Weather” is information indicating weather conditions at the time of the day, and includes information such as sunny, rainy, and snowy. “Weather” can be acquired from the outside by wireless communication using a weather information service.

  “Selection history of the day” is a selection history of the destination selected on the current day. “Selection history of the day” can be acquired from the selection table of the current day 12 f of the car navigation device 10 described later.

  The car navigation apparatus 10 performs “behavior prediction processing” that outputs these “prediction area” and “prediction genre” with these “estimation conditions” as inputs. The “prediction area” is area information to which a destination to be included in “recommended information” as a result of the behavior prediction process belongs. “Predicted genre” is a genre major classification and a genre minor classification to which destinations included in “recommended information” as a result of the behavior prediction process belong.

  The “behavior prediction process” is performed when the car navigation apparatus 10 is activated or when a user instructs to start the process.

  The “behavior prediction process” is stored in an action history stored in an action history table 12c described later and an action habit table 12d described later according to user preferences and user attributes stored in the user profile 12b described later. , Based on behavior habits that are the characteristics of behavior that can be statistically determined from the behavior history, and general common sense that is an attribute of a destination that is a reason to be excluded from “recommended information” stored in the general common sense data table 12e described later This is a process of predicting user behavior (needs).

  The action history is a destination that has been actually set from a destination content table 12a, which will be described later, that stores all the information of destinations that can be selected and set by the car navigation device 10. It is a history of information on the ground.

  Next, the car navigation device 10 inputs a genre selection tendency (or “predicted genre” as a result of the behavior prediction process) stored in a genre selection tendency table 12g, which will be described later, and outputs “recommended information” “preference”. Perform estimation processing. The genre selection tendency stored in the genre selection tendency table 12g is information extracted from the action history table 12c through a predetermined process.

  The “preference estimation process” is estimated by a genre selection tendency (or “behavior prediction process”) stored in a genre selection tendency table 12g described later according to user preferences and user attributes stored in a user profile 12b described later. The “predicted genre”) and the analogy genre stored in the genre analogy table 12i described later are processes for predicting preferences related to the user's destination selection.

  The opportunity for performing the “preference estimation process” is when the car navigation device 10 is activated or when the user gives an instruction to start the process.

  Next, the configuration of the car navigation apparatus according to the embodiment will be described. FIG. 3 is a functional block diagram illustrating the configuration of the car navigation apparatus according to the embodiment. The car navigation apparatus 10 according to the embodiment is an apparatus that is mounted on a vehicle and extracts and displays destination candidates that are likely to be selected when setting a destination. In addition, as long as the structure which performs content selection candidate extraction, action prediction, and estimation within a classification item can be cooperated with a car navigation apparatus, a separate apparatus may be sufficient.

  As shown in the figure, a car navigation device 10 according to an embodiment includes a control unit 11, a storage unit 12, and a display unit 13 that is a display unit such as a display device having a predetermined display screen.

  First, the storage unit 12 will be described prior to the description of the control unit 11. The storage unit 12 is a volatile or non-volatile storage unit, and includes a destination content table 12a, a user profile 12b, an action history table 12c, an action habit table 12d, a general common sense data table 12e, and a day selection history. It has a table 12f, a genre selection tendency table 12g, an attribute weighting table 12h, and a genre analogy table 12i.

  The destination content table 12a is a table that stores a destination and information associated therewith and is prepared in advance. As shown in the table example of the destination content table 12a in FIG. 4, "Destination", "Area information", "Genre major category", "Genre minor category", "Chain store name", "Content holder", "Event" Columns such as “information”, “price range / fee”, “popular ranking information”, and “atmosphere” are included. “Genre major classification”, “genre minor classification”, “chain store name”, “content holder”, “event information”, “price / charge”, “popular ranking information”, and “atmosphere” are called genre attributes.

  The “destination” is a spot on the map set as a destination by the car navigation device 10. “Destination” includes, for example, a facility name such as “A1 restaurant”. “Area information” is an area name on the map to which the corresponding “destination” belongs. In “area information”, for example, “destination” corresponds to “A1 restaurant” and “a1 area” is entered.

  “Genre classification” is an action performed by the user at the corresponding “destination”. In the “genre major classification”, for example, “destination” corresponds to “A1 restaurant” and “eat” is entered. The “genre subcategory” is a category as a spot to which the corresponding “destination” belongs. The “genre subcategory” is a classification of facilities such as “family restaurant”, “shopping center”, “leisure facility”, “government office”, and “park”. “Genre small classification” includes, for example, “family restaurant” corresponding to “destination” corresponding to “A1 restaurant”.

  The “chain store name” is the name of the chain store to which the corresponding “destination” belongs when the “destination” is a store. “Chain Store Name” includes, for example, “A Restaurant Churn” corresponding to “A1 Restaurant” in “Destination”. If “Destination” is not a store, “Null” is entered.

  The “content holder” is the name of an information provider who provides detailed guide information regarding the corresponding “destination”. In the “content holder”, for example, “Destination” corresponds to “A1 restaurant” and “X1 company” is entered.

  “Event information” is information indicating a period during which the corresponding “destination” is available or optimal for use (including a period during which a special event or the like is held). “Event information” includes, for example, “6/1 to 9/30” corresponding to “C1 amusement park” in “Destination”. Here, “event information” is “6 / 1-9 / 30” means that “C1 amusement park” is available during the period, is optimal for use during the period, or an event is held. It expresses that it has the characteristic of being. If “event information” does not exist, “Null” is entered.

  The “price range / charge” is information indicating an expense necessary for using the corresponding “destination” or an approximate value of the expense. “Price / Fee” includes, for example, “1000 yen” corresponding to “A1 restaurant” in “Destination”. Note that “Null” is entered when there is no “price range / charge”, that is, when usage is free.

  “Popularity ranking information” is a ranking indicating the degree of popularity given by the “content holder” to the corresponding “destination”. “Popular ranking information” includes, for example, “10th place” corresponding to “A1 restaurant” in “Destination”.

  “Atmosphere” is information indicating the atmosphere of the facility of the corresponding “destination”. “Atmosphere” includes, for example, information indicating mood such as “calm”, “quiet”, “lively”, “fun”, “young”, “couple”, “family (children)”, “ There is information indicating the target generation such as “for seniors”. For “atmosphere”, for example, “destination” corresponds to “A1 restaurant” and “lively” is entered.

  The user profile 12b is a table storing information related to the user and information related to the user's attributes, and is a table prepared in advance by setting necessary information as profile information. As shown in the table example of the user profile 12b in FIG. 5, columns such as “favorite genre”, “holiday setting”, “car type”, “user gender”, “user age”, “favorite content holder”, and the like are included. The “favorite content holder” may be estimated from “user gender”, “user age”, and the like.

  Under “favorite genres”, columns of “genre major classification” and “genre minor classification” are further provided. The “genre major classification” and the “genre minor classification” specify what kind of action the user likes to perform and what kind of facility he / she prefers to perform the action. For example, referring to FIG. 5, under “Favorite Genre”, there are entries of “Everyone in Genre” is “Eat” and “Family Restaurant” is in “Family Restaurant”. The fact that they prefer to “eat” at “family restaurants” will be identified.

  In FIG. 5, the profile information preset in the user profile 12b includes “genre major classification”, “genre minor classification”, “holiday setting”, “vehicle type”, “user gender”, “user age”, “ Although only “favorite content holders” are described, the present invention is not limited to these. That is, based on the columns stored in the destination content table 12a, "favorite chain store name", "favorite source category", "favorite event information", "desired price range / charge", "desired popularity ranking information" , “Desired atmosphere” may be included. These profile information are profile information respectively corresponding to “chain store name”, “event information”, “price range / charge”, “popular ranking information”, and “atmosphere” in the destination content table 12a.

  “Holiday setting” stores the day of the week and date that the user particularly sets as “holiday”. With this setting, the car navigation apparatus 10 does not treat Saturday and Sunday as holidays, but treats the set day of the week or date as holidays and performs the above-described “behavior prediction process”.

  “Vehicle type” is a type of a user's vehicle on which the car navigation device 10 is mounted. The “user gender” is the gender of the user of the vehicle. “User age” is information indicating the actual age of the user or the generation of the user. The “favorite content holder” refers to a content holder that is frequently used when a user obtains information about a destination.

  The action history table 12c is a table in which necessary columns are extracted and stored from the destination content table 12a that are actually set as destinations by the user. As shown in the table example in FIG. 6, the action history table 12 c includes “date”, “day of the week”, “time”, “time zone”, “destination”, “area information”, “genre classification”, “ It has a column “genre subcategory”, and records are sorted in descending order of “date”, “day of the week”, “time”, and “time zone”.

  Here, “date”, “day of the week”, “time”, “time zone” are the date, day of the week, time, and time zone when the destination is set as the destination, and “destination”, “area information” , “Genre major category” and “genre minor category” are columns extracted from the destination content table 12a in accordance with the destination setting. In this way, when a certain destination is set in the car navigation device 10 at a certain date, day of the week, time, and time zone, the corresponding “destination”, “area information”, “genre classification”, The information in the “genre subcategory” column is extracted from the destination content table 12a and stored in the action history table 12c. For example, according to FIG. 6, “A soba shop” is set as a destination at 2007/7/30, Monday, 12:00 noon, and this “A soba shop” corresponds to “A soba shop”. “Area information”, “genre major classification”, and “genre minor classification” are “a area”, “eat”, and “soba”.

  Based on the information stored in the behavior history table 12c, the behavior habit table 12d stores, for example, the periodicity and selection characteristics that are latent at a frequency of 70% or more of the whole in the behavior history based on the destination setting of the user. It is a table. As shown in a table example of the behavior habit table 12d in FIG. 7, “genre major classification”, “genre minor classification”, “day periodicity”, “day period periodicity”, “time zone”, “daily” Column of “number of times of selection”.

  Referring to FIG. 7, for example, based on the information stored in the action history table 12c, the action history based on the destination setting of the user is “Food category” is “Eat” and “Genre category” is “Family”. “Restaurant”, “Day periodicity” is “7 days”, “Day period periodicity” is “Sunday”, “Time zone” is “Night”, and “Number of selections per day” is “1 time”. Although it exists, this is a habit of having the user “eat” at the “Family Restaurant” once a day in the “Sunday” and “Night” time periods in the “7-day” cycle. It shows that there is. In the behavior habit table 12d, “Null” is stored for a column having no corresponding periodicity or feature.

  The general common sense data table 12e is a table that defines, as general common sense, a situation in which the facility cannot be used. As shown in a table example of the general common sense data table 12 e in FIG. 8, columns of “genre large category”, “genre small category”, “general common sense regarding time zone”, and “general common sense related to weather” are provided. “General common sense regarding time zone” means the time zone when the facility as the destination specified by the “genre major category” and “genre minor category” is not operated or the window business is not performed. Information. This time zone has seven categories of “morning”, “morning”, “daytime”, “afternoon”, “evening”, “night”, and “midnight”. The “general common sense regarding the weather” is information indicating the weather in which the facility as the destination specified by the “genre subcategory” cannot be used. There are four categories of weather: “clear”, “cloudy”, “rain”, and “snow”.

  Referring to FIG. 8, for example, when the “genre major category” is “eat” and the “genre minor category” is “ramen”, the “morning” column contains “x”. “It is not open in the morning” is “common sense of time”. Even if the facilities specified by the “genre major classification” and “genre minor classification” are obtained by the “forecast area” and the “predictive genre” as the prediction results by the behavior prediction process, the general prediction is used. Since it is excluded from the result, it is possible to prevent a prediction result that is impossible in common sense from being obtained. In the general common sense data table 12e, “Null” is stored for a column for which no corresponding common common sense exists.

  The current day selection history table 12f is a table that stores the destination set in the car navigation device 10 on the current day and its related information as the current day selection history. As shown in a table example of the current day selection history table 12 f in FIG. 9, there are columns of “Destination”, “Genre major category”, and “Genre minor category”. These are columns of entries extracted from the destination content table 12a by setting the destination in the car navigation device on the current day.

  Referring to FIG. 9, for example, an entry where “Destination” is “A1 restaurant”, “Large genre classification” is “Eat”, and “Genre small classification” is “Family restaurant” is an “Family” action “A1 restaurant”, which is a “restaurant”, indicates that the destination has been set today.

  As shown in FIG. 10, the genre selection tendency table 12g includes, for example, “aggregation unit”, “genre subcategory”, and the latest N (N is a natural number) records in the action history table 12c. , A table storing entries having columns of “weight total”, “weight total ratio”, “slope”, “slope trend”, and “specialization flag” corresponding to the “total unit” and “genre subcategory” is there. The method of extracting the latest N records from the action history table 12c and generating the genre selection tendency table 12g is as follows.

  First, the latest N records are extracted from the action history table 12c. This is because individual selection trends depend on trends and change with time, so it is meaningless to extract old records, and we received recent trends by extracting newer records. This is because it is possible to grasp individual behavioral trends.

  For example, a temporary table obtained by extracting the latest eight records from the example of the action history table 12c shown in FIG. 6 is as shown in FIG. The temporary table obtained by extracting the latest eight records from the action history table 12c is expanded in a predetermined temporary storage area. This temporary table has columns of “selected date / time zone”, “genre subcategory”, “weight base value”, and “weight value”.

  At the time of this expansion, for example, “Monday” to “Friday” are converted to “Weekdays”, and “Sat” and “Sunday” are converted to “Holidays”. The converted “weekdays” and “holiday” classifications are merged together with the “date” and “time zone” information and included in the item “selected date / time zone” in the temporary table.

  As the “weight base value”, a value of the maximum value P (positive number) is assigned to the latest record, and then a value obtained by subtracting the increment value Q (positive number) from P is assigned to the new record. In this way, a value obtained by sequentially subtracting the increment value Q from P is assigned to all the extracted records. Note that P and Q are numerical values that can be tuned and can be changed depending on N so that the values assigned to all records are positive numbers.

  According to the example of FIG. 11, a weight base value “2” is assigned to a record whose “selected date / time zone” is “2007/7/30”, and step values are assigned in the most recent order to other records. As “0.1”, weight base values “1.9”, “1.8”,. A “weight value” is obtained by squaring this “weight base value”. By calculating the “weight value” in this way, the latest selection tendency is made more prominent.

  Then, the records calculated up to “weight value” are aggregated in a predetermined aggregation unit to calculate the sum of “weight values” (“total weight”), and the result is stored in the genre selection tendency table 12g of FIG. Store. With regard to the selection tendency, for example, in the “category subcategory” of “eat”, it can be expected that the tendency changes in the scenes of “morning”, “daytime”, and “night”. In addition, in the “genre sub-category” of “play”, it can be expected that the trend changes with “weekdays / holidays”. As described above, the selection tendency is totaled for each total unit in which the trend can be predicted according to the content of the “genre subcategory”.

  Next, the ratio of each “weight sum” with respect to the maximum value of “weight sum” (“weight sum ratio”) is calculated, and the result is stored in the genre selection tendency table 12g of FIG. Referring to FIG. 10, since the maximum value of “total weight” is “10.5”, each “total weight” “2.25”, “9.45”, and “10.5” is set to “10. The values “21.43%”, “90.00”, and “100.00%” divided by 5 and multiplied by 100 are stored in the corresponding “weight total ratio” column.

  With the above, it was possible to grasp the selection tendency for each aggregation unit. However, in general, when grasping a selection tendency, it is necessary to pay attention to the tendency to select specially. For example, if there are multiple recommended candidates, and each candidate has various attributes, when there are candidates that have attributes that are specifically selected, other attributes follow the selection trend. Even if not, it may be a recommended candidate by focusing on specialized attributes.

  In order to extract such specialized attributes, the following processing is performed. That is, first, the records in the genre selection tendency table 12g of FIG. 10 are sorted in ascending order of “total weight”. Then, the difference from the “total weight” of the previous record is taken from the maximum value of the “total weight” to the minimum value. This is stored as “inclination” in the genre selection tendency table 12g. For example, the “slope” of a record whose “aggregation unit” is “weekday daytime” and “genre subcategory” is “soba” is one record before the “total weight” “10.5” of the record. Is “1.05” obtained by subtracting “9.45”.

  Then, in the genre selection tendency table 12g, the ratio of each “slope” (“slope trend”) to the maximum value of “total weight” is calculated based on the following equation, and the calculation result is the column of “slope tendency” of each record. To store.

  Then, in the genre selection tendency table 12g, it is sequentially determined whether the “slope tendency” is “specific tendency” T% or more from the minimum value of the “total weight” to the maximum value. Note that T is a tuning parameter. If there is a record whose “slope tendency” is equal to or greater than the special tendency T, all the “specialization flags” of the subsequent records are turned on.

  The situation in which the “specialization flag” is turned on is as shown in FIG. That is, when the genre selection tendency table 12g is as shown in FIG. 10, T is set to 40%. First, the “inclination tendency” “0%” of the “initial value” is changed to the “inclination tendency” “21.4%” of the record “weekday noon” and “udon”. Since “21.4%” is less than T, “weekday daytime” and “udon” are said to be “no specialization”.

  Next, the “inclination tendency” “21.4%” of “weekday noon” and “udon” changes to “inclination tendency” “68.6%” in the record of “weekday noon” and “ramen”. Since “68.6%” is equal to or greater than T, “noon on weekdays” and “ramen” are set to “have a special tendency”, and the “specialization flag” is turned on. Then, all the records after this record are assumed to have “specialization tendency”, and the “specialization flag” is turned on. That is, “ramen” and “soba” are specialized attributes.

  In addition, it becomes possible to extract a specialized attribute reliably by narrowing down and extracting to the nearest N records of the action history table 12c.

  Next, by the genre analogy process based on the “genre subcategory” stored in the genre selection tendency table 12g, another “genre subcategory” can be inferred and the range of “genre subcategory” can be expanded.

  As illustrated in FIG. 13, the genre analogy table 12 i is a table in which at least one “abstraction characteristic” is associated with “genre subcategory”. When it is determined by the genre analogy process that the degree of similarity of the “abstraction property” is equal to or higher than a predetermined level (for example, a degree of similarity of 70% or higher), the “genre small classification” can be acquired by reverse lookup. It is a table. As shown in the table example of the genre analogy table 12 i in FIG. 13, columns of “genre subcategory” and “abstraction characteristics” are provided. The “abstraction property” is an abstract property of information entered in the “genre subcategory”, and includes, for example, “food”, “taste”, “volume”, and the like.

  Specifically, referring to FIG. 13, when “genre subcategory” is “barbecue”, “abstracting characteristics” are “food” as “meat”, “taste” as “rich taste”, “volume feeling” "Is a sense of volume" ... In the following, the “genre sub-category” is described as “analogue genre” where “yaki meat” is “analogue genre”, and “genre subcategory” is estimated as “analogue genre”.

  First, when all entries in the column of “food” are searched, for example, since “food” of “genre subcategory” is “steak” and “beef” is “beef”, it is determined that there is a similarity of 70% or more, for example. A score corresponding to the degree of similarity is given.

  Next, when all the entries in the “taste” column are searched, for example, since “taste” of “steak” of “genre subcategory” is “rich taste”, it is determined that there is similarity, and the similarity is A corresponding score is given (in this case, the “taste” of both is the same, so the similarity is 100%).

  Next, when all entries in the column of “Volume Feel” are searched, for example, “Volume Feel” of “Steak” is “Steak” and “Volume Feel” is determined, so it is determined that there is similarity. A score corresponding to the degree of similarity is given (in this case, since the “volume feeling” of both is the same, the degree of similarity is 100%).

  In this way, “genre” is a result in which all entries of all columns are searched for all “abstract characteristics” and a score corresponding to the similarity is totaled is a predetermined value or more (for example, a similarity of 70% or more). The entry “minor classification” is “analog genre” estimated by “analog genre”. Finally, “analog genre” analogized in this way is added to the genre selection tendency table 12g. According to the genre analogy table 12i of FIG. 13, when the “genre subcategory” is “ramen”, “udon”, or “soba”, “spaghetti” is analogized and added. The records in the genre selection tendency table 12g to which the “analog genre” is added are ranked according to a predetermined method.

  FIG. 14 is an example of a weighting table of attribute values in a certain “genre major classification”. As shown in FIG. 14, the attribute value weighting table 12 h includes columns of “attribute”, “base score”, “specialized tendency weight”, and “specialized tendency T”. The attribute value weighting table 12h has, for example, each attribute (“genre small category”, “chain store name”, “content holder”, “event information”, “atmosphere”) when “genre major category” is “eat”. Is an example of a table storing information for weighting and ranking according to the presence / absence of the aforementioned “specialization tendency”.

  The “base score” is a score that serves as a base when ranking the records in the genre selection tendency table 12g based on each attribute value. The “base score” of “genre subcategory” and “chain store name” is “200”. The “base score” of “content holder” and “atmosphere” is “100”. The “base score” of “event information” is “150”.

  The “specialized tendency weight” is a multiplier by which the base score is multiplied when the records in the genre selection tendency table 12g are ranked based on each attribute value and the attribute has a “special tendency”. The “specialization tendency weight” of “genre subcategory”, “chain store name”, and “atmosphere” is “1.4”. The “specialized tendency weight” of “content holder” and “event information” is “1.2”.

  “Specialization tendency T” is a threshold value for determining whether or not there is “specialization tendency” in each attribute of each record in the genre selection tendency table 12g. “Specialization tendency T” of “genre subcategory”, “chain store name”, “content holder”, “event information”, and “atmosphere” are all “40”%.

  Here, when the information in the attribute value weighting table 12h is applied to the record in the genre selection tendency table 12g to which “analogous genre” is added, weighting and ranking are performed according to the presence / absence of “special tendency”. As shown in FIG. Prior to this process, it is determined whether or not the attribute value of the candidate “genre subcategory” matches the attribute value of the selection tendency. At this time, a record that is a “counting unit” that matches the above-described “estimation condition” is set as a check target.

  For example, as shown in FIG. 15, when “genre subcategory” is “ramen”, (a) the base score is 200 points, and (b) the total weight ratio is 90.00%. (C) 180.00 is calculated by calculating (a) × (b).

  Further, when the “genre sub-category” is “ramen”, referring to the genre selection tendency table 12g in FIG. 10, the “specialization flag” is “1” (on), so this (c) 180 By multiplying .00 by (d) specialized tendency weight 1.4, (e) priority score = 252.00 is finally calculated. Similarly, “Udon”, “Spaghetti”, and “Soba” for “genre subcategory” also have priority scores based on their “base score”, “total weight ratio”, and “presence / absence of specialization”. Calculated.

  When “genre subcategory” is “Udon”, (a) the base score is 200 points, and (b) the total weight ratio is 21.43%, so (a) × (b ) Is calculated, (c) 42.86 is calculated. Further, when the “genre subcategory” is “Udon”, referring to the genre selection tendency table 12g of FIG. 10, the “specialization flag” is “0” (off), so this (c) 42 Multiply .86 by (d) specialization tendency weight 1.0 to finally calculate (e) priority score = 42.86.

  When the “genre subcategory” is “spaghetti”, (a) the base score is 200 points, and (b) the total weight ratio is 0.000%, so (a) × (b ) Is calculated, (c) 0.00 is calculated. Further, when the “genre subcategory” is “spaghetti”, referring to the genre selection tendency table 12g in FIG. 10, there is no corresponding record, so that the “specialization flag” is “0” (off). Assuming that (c) 0.00 is multiplied by (d) special tendency weight 1.0, (e) priority score = 0.00 is finally calculated.

  The priority score is, in principle, the total score given for each genre attribute. However, when a plurality of values are implemented in one attribute, the priority score having the largest value among the priority scores for the plurality of values is set as the priority score.

  In addition, the “base score” and “specialized tendency weight” in the attribute value weighting table 12h greatly contribute to the calculation of the priority score. These are considered to change depending on the “total unit” of the selection tendency and the “genre attribute” as a target. The “aggregation unit” is taken into account by aggregating the selection tendency by the “aggregation unit”. However, it is complicated to calculate the priority score for every “genre attribute”. Therefore, the “genre attribute” is set to “group A (preference / purpose)” (“genre subcategory”, “chain store name”), “group B (characteristic factor)” (“content holder”, “event presence / absence”). ) And “Group C (attribute factor)” (“atmosphere”) are grouped into three groups, and a priority score is calculated for each of these groups, thereby enabling efficient processing. . The display order of the recommended information is determined by the priority score.

  In other words, the above shows a case where the genre selection tendency is grasped based on the “genre subcategory”, the genres are inferred, prioritized based on the priority scores of these genres, and presented as recommended information. However, the present invention is not limited to this, and “chain store name”, “content holder (source)”, “event information”, and “atmosphere” can be similarly processed. In other words, if the selection tendency can be grasped only for the genre attributes “genre subcategory”, “chain store name”, “content holder”, “event information”, and “atmosphere”, it can be presented to the user as specific recommended information. is there.

  The recommended information ordered in this way is as shown in FIG. FIG. 16 is a diagram illustrating an example of a recommendation information table generated based on “genre small classification”. The recommendation information table is expanded in a predetermined temporary storage area. The recommended information table in FIG. 16 is a table in the case where “genre major classification” is “eat”.

  As shown in FIG. 16, the recommended information table includes “group A (preference / purpose)” (“genre subcategory”, “chain store name”) and “group B (characteristic factor)” (“content holder”, “Event information”), “Group C (attribute factor)” (“Atmosphere”), and “Priority score” columns. “Ramen”, “Soba”, “Udon”, “Spaghetti” of “Genre sub-category” are sorted in descending order of each “priority score”. In addition, attribute values of genre attributes such as “chain store name”, “content holder”, “event presence / absence”, and “atmosphere” are stored.

  Here, returning to FIG. 3 again, the control unit 11 will be described. The control unit 11 is a device that controls the overall control of the car navigation device 10, and in particular, as the components related to the embodiment, the behavior history registration unit 11 a, the behavior habit extraction processing unit 11 b, the behavior prediction processing unit 11 c, A genre selection tendency extraction processing unit 11d, a genre analogy processing unit 11e, a preference estimation processing unit 11f, and a selection list display control unit 11g.

  The action history registration unit 11a extracts information corresponding to the “destination” set as the destination in the car navigation device 10 by the user from the destination content table 12a, and the action history table 12c together with the destination as an action history. Process to register in.

  The behavior habit extraction processing unit 11b first (1) based on the behavior history registered and stored in the behavior history table 12c, for example, “periodicity of the day” latent in the behavior history with a frequency of 70% or more of the whole. ”,“ Periodicity of day of the week ”, and“ time zone ”are extracted.

  Next, the behavior prediction processing unit 11c extracts (2) an entry corresponding to the day of the week selected with a frequency of 70% from the behavior history table 12c, and stores the extraction result in the general common sense data table 12e. The lowest priority “10” is assigned by screening based on general common sense. Furthermore, the behavior prediction processing unit 11c (3) priorities “1”, “3”, “4”, “7” are assigned to the above extraction results according to the periodicity and characteristics stored in the behavior habit table 12d. ".

  Then, the behavior prediction processing unit 11c (4) based on the same day selection history stored in the current day selection history table 12f, the same “genre major category” and “genre minor category” entries are included in the behavior history table 12c. When the value is the same as the “number of times of selection per day”, that is, when the number of times of selection per day indicated by the action history has been reached, the priority “ 8 "is given.

  And the action prediction process part 11c gives priority "5" with respect to what was not provided with the priority by the process of said (1)-(4) among the above-mentioned extraction results. Finally, the behavior prediction processing unit 11c sorts the above-described extraction results in ascending order of priority assigned to each, ascending order of “time zone”, and descending order of frequency.

  The genre selection tendency extraction processing unit 11d extracts, from the action history table 12c, for example, “genre subcategory” entries of the latest N records together with “date”, “time”, “day of the week”, and “time zone”. , “Date”, “Time”, “Day of the Week”, and “Time Zone” are “Selected Date / Time Zone”, and “Weight Base Value” and “Weight Value” are assigned to each record.

  Then, “Weekdays / Holidays” and “Time zone” of this “Selected date / time zone” are set as “Aggregation unit”, and “Weight value” is aggregated based on this “Aggregation unit” and “Genre subcategory” And registered in the genre selection tendency table 12g. Further, “total weight”, “total weight ratio”, “inclination”, “inclination tendency”, and “specialization flag”, which are the aggregated “weight values”, are assigned to the respective aggregated records.

  The genre analogy processing unit 11e has at least one “genre subclass” based on the “genre subclass” of the entry registered and stored in the genre selection tendency table 12g or the “genre subclass” input by the user. When it is determined with reference to the genre analogy table 12i associated with the “abstract characteristic” that the similarity of the “abstract characteristic” is equal to or higher than a predetermined level (for example, a similarity level of 70% or higher). By acquiring “genre small classification” by reverse lookup, the analogy genre small classification estimated from the analogy based genre small classification is acquired.

  Note that the “abstraction property” of the genre analogy table 12i learns the “genre subcategory” that is frequently selected by the user (for example, at a rate of 70% or more). It is good also as learning the "abstraction characteristic" which a user selects frequently (for example, in the ratio of 70% or more). In this way, it is possible to expand the range of “genre sub-category” to be processed in the latter-stage preference estimation process.

  The preference estimation processing unit 11f acquires the “base score”, “specialized tendency weight”, and “specialization” acquired by referring to the attribute value weighting table 12h for the “genre subcategory” expanded by the genre analogy processing unit 11e. The ordering is performed based on the priority score calculated based on “tendency T”. At this time, the “destination” acquired from the destination content table 12a is scored based on “preference / purpose”, “characteristic factor”, and “attribute factor”. Then, the preference estimation processing unit 11f sorts the “genre small classifications” in descending order of scoring.

  The selection list display control unit 11g displays “genre subcategories” sorted in descending order of scoring by the preference estimation processing unit 11f on the display unit 13 together with various information corresponding to the “destination” in the destination content table 12a. Display control to display.

  Next, behavior prediction processing executed by the car navigation device of the embodiment will be described. FIG. 17 is a flowchart illustrating a behavior prediction processing procedure executed by the car navigation device 10 according to the embodiment. As shown in the figure, first, the behavior prediction processing unit 11c of the car navigation device 10 extracts candidates having a high frequency on the day of the week from the behavior history based on the day of the “guessing condition” (step S101). The candidates extracted in step S101 are hereinafter referred to as “extraction candidates”.

  Note that, in step S101, extraction is not performed based on whether or not it is a holiday. This is because the tendency of whether or not it is a holiday can be grasped by the tendency of the day of the week. Note that it is possible to set whether or not the holiday is an extraction condition in the user profile 12b, but if there is no setting in the user profile 12b, the day of the week set as a holiday in the user profile 12b is set as a holiday.

  Also, if it is impossible to extract candidates with high frequency for the day of the week from the action history in step S101, the day of the week is ignored and only a predetermined number of candidates with high frequency for weekday / holiday discrimination are extracted. Also good.

Subsequently, the behavior prediction processing unit 11c includes a behavior history stored in the behavior history table 12c,
Based on the common sense stored in the common sense data table 12e, among the extraction candidates extracted in step S101, a priority “10” is set to a candidate that cannot be selected in the time zone (step S102). .

  That is, the priority of the extraction candidates that are designated as “not selected” in the time zone of the “guessing condition” in general common sense and that are not actually selected in the time zone in the action history are lowered. Note that extraction candidates that are not designated as “not selected” in that time zone in general sense may be selected in that time zone, so they are not selected in that time zone in the action history. Since there is a problem in lowering the priority, the condition is “and”.

  Subsequently, the behavior prediction processing unit 11c sets the priority “7” to the extraction candidates that are less likely to be selected according to the weather of the current day based on “weather” of the “estimation condition” (step S103). For example, if the “estimation condition” is “rain” and the general common sense is rain, “not selected” is selected, and the priority of extraction candidates not actually selected by “rain” in the action history is lowered. It should be noted that extraction candidates that are not designated as “not selected” in that time zone in general sense may be selected in the relevant weather. Since there is a problem in lowering the value, the condition is “and”.

  Subsequently, based on the “day of the week” and “time zone” of the “estimation condition”, the behavior prediction processing unit 11c selects a predetermined frequency or higher in “day of the week” and “time zone” that match the “estimation condition” ( For example, priority “1” is set for extraction candidates of 70% or more (step S104). This is because it can be said that there is a very high possibility that the selection frequency of “day of the week” and “time zone” of the “estimation condition” is 70% or more, for example.

  Subsequently, based on the “day of the week” and “time zone” of the “estimation condition”, the behavior prediction processing unit 11c selects candidates having a selection frequency equal to or higher than a predetermined value in “day of the week” and “time zone” that match the estimation condition. Among them, priority “2” is set for candidates that are not selected even if the time zone has passed 1 unit, and priority “10” is set for extraction candidates that are not selected even if the time zone has passed 2 units or more. Is set (step S105). Here, one unit of “time zone” is each time zone that passes in the order of “morning”, “morning”, “afternoon”, “night”, “midnight”, “morning”,. Say.

  Here, it can be estimated that the extraction candidates specializing in “day of the week” and “time zone” are considerably less likely to be selected once the time zone has passed. In the selection history of the day, it can be dropped on the condition that it has already been selected, but when considering the implementation, it may not be possible to grasp the fact that it was actually selected. Extraction conditions that have passed have been lowered in priority.

  Subsequently, the behavior prediction processing unit 11c may select a candidate based on the periodicity, and set the priority “3” as an extraction candidate having a selection frequency equal to or higher than a predetermined value (eg, 70% or higher) in the “time zone”. "Is set (step S106). It should be noted that the extraction conditions that have “periodicity of day” and that are specialized in “time zone” are considered to have a higher priority, but slightly higher priority than those that are specialized in “day of the week”. Because it falls.

  Subsequently, the behavior prediction processing unit 11c may be selected based on periodicity, but the priority “4” is selected as an extraction candidate whose selection frequency is less than a predetermined value (for example, less than 70%) in the “time zone”. Is set (step S107). In other words, “day periodicity” has a high possibility of being selected on the day, but “time zone” varies, so priority is set to “4”. This is because an extraction candidate having “day periodicity” can be assumed to have a higher priority than that extracted for “day of the week”. Examples are “I go to the supermarket every weekend” and “I go to the weekend once every three weeks, but today is once every three weeks”.

  Subsequently, the behavior prediction processing unit 11c may select a predetermined number of times within one day, but sets a priority “8” to an extraction candidate that has already been selected a predetermined number of times (step S108). Specifically, based on the same day selection history table 12f, the frequency of the maximum number of selections per day is tabulated for each extraction candidate. The number of times that the frequency of the maximum number of selections per day is 70% or more based on the total result is defined as “specialized number”. Then, from the selection history of the current day, those that have been selected more than the specialized number of times are excluded from the extraction candidates.

  Subsequently, the behavior prediction processing unit 11c sets the priority “5” to the extraction candidates for which the priority is not set (step S109). Subsequently, the behavior prediction processing unit 11c sorts the extraction candidates in ascending order of priority, ascending order of time zone, and ascending order of frequency (step S110).

  By performing these processes, based on the extraction candidates, the “prediction area” is estimated based on the “area information” included in the extraction candidates, and the “prediction genre” is estimated based on the “genre major classification” and the “genre minor classification”. Is possible. The “prediction area” and “prediction genre” estimated in this way may be displayed as a list on the display screen of the display unit 13.

  Next, the preference estimation process performed with the car navigation apparatus of an Example is demonstrated. FIG. 18 is a flowchart illustrating a preference estimation processing procedure executed by the car navigation device 10 according to the embodiment. As shown in the figure, first, the preference estimation processing unit 11f of the car navigation device 10 extracts the latest N records from the behavior history table 12c (step S201).

  Subsequently, the preference estimation processing unit 11f gives a weight value to the extracted action history (step S202). Subsequently, the preference estimation processing unit 11f aggregates the total weight, which is the sum of the weight values, for each day of the week, time zone, and genre subcategory to obtain an aggregated selection tendency (step S203). Subsequently, the preference estimation processing unit 11f calculates the ratio of the total weight for each aggregation unit (step S204).

  Subsequently, the preference estimation processing unit 11f sorts the aggregation units in ascending order of the total weight, and calculates the difference between the two consecutive total weights as a slope (step S205). Subsequently, the preference estimation processing unit 11f calculates the ratio of each inclination with respect to the maximum weight value for each aggregation unit as an inclination tendency (step S206).

  Subsequently, the preference estimation processing unit 11f turns on the specialization flag of the record whose inclination tendency is equal to or greater than the specialization tendency T (step S207). Subsequently, the preference estimation processing unit 11f analogizes the genre small classification based on the abstraction characteristic (step S208). Subsequently, the preference estimation processing unit 11f adds the analogized small category of genre to the aggregated selection tendency (step S209).

  Subsequently, the preference estimation processing unit 11f calculates the product (priority score) of the base score, the total weight ratio, and the specialized tendency weight for each selection tendency attribute (step S210). Subsequently, the selection history is sorted in descending order of the product (priority score), and the recommended attribute value is added and displayed (step S211).

  FIG. 19 is a display example of “recommended information” that is a processing result obtained through the behavior prediction process and the preference estimation process. As shown in the figure, the “recommended information” may be selected on the display screen 13a of the display unit 13 based on the “estimation condition” and the “genre selection tendency” so that the user can easily select the “recommended information”. By controlling the display of “destination” in descending order, a list of selection candidates can be obtained. In this way, the user can select the “destination” he / she desires in one operation without manually setting from “1”.

  As mentioned above, although the Example of this invention was described, this invention is not limited to this, In the range of the technical idea described in the claim, even if it implements in a various different Example, it is. It ’s good. Moreover, the effect described in the Example is not limited to this.

  In addition, among the processes described in the above embodiment, all or part of the processes described as being automatically performed can be manually performed, or the processes described as being manually performed can be performed. All or a part can be automatically performed by a known method. In addition, the processing procedure, control procedure, specific name, information including various data and parameters shown in the above embodiment can be arbitrarily changed unless otherwise specified.

  Each component of each illustrated device is functionally conceptual and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured.

  Furthermore, each or all of the processing functions performed in each device are entirely or partially a CPU (Central Processing Unit) (or a microcomputer such as an MPU (Micro Processing Unit) or MCU (Micro Controller Unit)) and It may be realized by a program that is analyzed and executed by the CPU (or a microcomputer such as MPU or MCU), or may be realized as hardware by wired logic.

  The present invention predicts the user's behavior more accurately with a simple configuration and presents the user behavior and purpose presented to the user when the content selection device is presented to the user in a content providing device such as a car navigation device. This is useful when it is desired to increase the possibility of user selection by giving variations to selection candidates such as the ground.

DESCRIPTION OF SYMBOLS 10 Car navigation apparatus 11 Control part 11a Action history registration part 11b Behavior habit extraction process part 11c Action prediction process part 11d Genre selection tendency extraction process part 11e Genre analogy process part 11f Preference estimation process part 11g Selection list display control part 12 Storage part 12a Destination content table 12b User profile 12c Behavior history table 12d Behavior habit table 12e General common sense data table 12f Same day selection history table 12g Genre selection tendency table 12h Attribute value weighting table 12i Genre analogy table 13 Display unit 13a Display screen

Claims (4)

A content selection candidate extraction device that extracts selection candidates of contents that can be selected by a user under a predetermined situation from contents that can be classified by at least one classification item based on an input estimation condition and presents the selection candidates to the user ,
A selection history recording means for recording a selection history including the genre and selection date and time of the content selected by the user;
Based on the genre and selection date / time of the content recorded as the selection history by the selection history recording unit, the periodicity that is latent in the content selection action by the user is extracted for each genre of the content recorded as the selection history. Periodicity extraction means;
Selection candidate extraction means for extracting content selection candidates that can be selected by the user based on the selection history and the periodicity extracted by the periodicity extraction means;
Selection candidate exclusion means for excluding content from the selection candidates extracted by the selection candidate extraction means, based on common sense conditions that stipulate a situation in which the content cannot be the selection candidates in common sense under the predetermined situation; ,
In order to present to the user, in order of priority, the content of the content corresponding to the estimation condition, the content is extracted as the selection candidate by the selection candidate extraction unit and not excluded from the selection candidate by the selection candidate exclusion unit. A content selection candidate extraction device comprising priority assignment means for assigning a priority based on the consistency of specific data of classification items, the periodicity, or the number of selections of the content that day. A content selection candidate extraction method for extracting content selection candidates that can be selected by a user under a predetermined condition from content that can be classified by at least one classification item based on an input estimation condition, and presenting the selection candidates to the user ,
A selection history recording step for recording a selection history including the genre of the content selected by the user and the selection date and time ;
Based on the genre of the content recorded as the selection history and the selection date and time by the selection history recording step, the periodicity that is latent in the content selection act by the user is extracted for each genre of the content recorded as the selection history. A periodicity extraction step;
A selection candidate extraction step for extracting selection candidates of content that can be selected by the user based on the selection history and the periodicity extracted by the periodicity extraction step;
Selection candidate exclusion step of excluding content from the selection candidates extracted by the selection candidate extraction step based on common sense conditions that stipulate a situation in which the content cannot be the selection candidates in common sense under the predetermined situation When,
Based on the abstract characteristic table that stores the specific data of the classification item of the content extracted as the selection candidate by the selection candidate extraction step in association with the abstract characteristic that abstracts and characterizes the specific data of the classification item An analogy classification item acquisition step of acquiring specific data of the classification item of other content whose similarity and the abstract property of the classification item are not less than a predetermined value, as an analogy classification item;
In order to present in order of priority to the user the content extracted as the selection candidate by the selection candidate extraction step and the content specified by the analogy classification item acquired by the analogy classification item acquisition step, Priority based on consistency of specific data of classification item of content with estimation condition, periodicity, number of selections of the content on the day, selection tendency for the content, purpose of the content, characteristic factor or attribute factor of the content A content selection candidate extraction method comprising: a priority providing step of assigning a degree. A behavior prediction device that predicts a user's behavior specified by the classification item of a content selection candidate that can be selected by a user from content that can be classified by at least one classification item based on an input estimation condition. ,
A selection history recording means for recording a selection history including the genre and selection date and time of the content selected by the user;
Based on the genre and selection date / time of the content recorded as the selection history by the selection history recording unit, the periodicity that is latent in the content selection action by the user is extracted for each genre of the content recorded as the selection history. Periodicity extraction means;
Based on the selection history and the periodicity extracted by the periodicity extracting means, a classification item extracting means for extracting the classification items of content selection candidates that can be selected by the user;
Under the predetermined situation, the classification item is extracted from the selection candidate based on a common sense condition that defines a situation in which the content corresponding to the classification item extracted by the classification item extraction means cannot be the selection candidate. A classification item exclusion means to be excluded;
In order to present the user in order of priority to the classification items extracted by the classification item extraction unit and not excluded from the selection candidates by the classification item exclusion unit, A behavior predicting apparatus comprising: priority giving means for assigning priority based on the consistency of specific data, the periodicity, or the number of selections of the content on the day. A behavior prediction method for predicting a user's behavior specified by the classification item of a content selection candidate that can be selected by a user from content that can be classified by at least one classification item based on an input estimation condition, ,
A selection history recording step for recording a selection history including the genre of the content selected by the user and the selection date and time ;
Based on the genre of the content recorded as the selection history and the selection date and time by the selection history recording step, the periodicity that is latent in the content selection act by the user is extracted for each genre of the content recorded as the selection history. A periodicity extraction step;
A category item extracting step of extracting the category items of content selection candidates that can be selected by the user based on the selection history and the periodicity extracted by the periodicity extracting step;
Under the predetermined situation, the classification item is selected from the selection candidate based on the common sense condition that defines the situation where the content corresponding to the classification item extracted by the classification item extraction step cannot be the selection candidate. A category exclusion step to exclude;
In order to present to the user in order of priority the classification items extracted by the classification item extraction step and not excluded from the selection candidates by the classification item exclusion step, the classification items of the content with respect to the estimation condition Priority assignment that gives priority based on the consistency of specific data, the periodicity, the number of selections of the content on the day, or the selection tendency for the content, the purpose of the content, the characteristic factors or attribute factors of the content A behavior prediction method comprising steps.