Information processing system, information processing method, and information processing program

Abstract

This invention provides an information processing system, method, and program that easily reflects personal preferences in a schedule. One embodiment of the recommendation system includes: a generation unit that generates multiple schedules, each including at least one location as a stop, based on a relationship where the schedule involves movement between multiple locations as an explanatory variable and the utility of the schedule for the user as a target variable; and an output unit that outputs one or more of the multiple schedules.

Description

Technical Field

[0001] This invention relates to an information processing system, an information processing method, and an information processing program. Background Technology

[0002] Technology is being researched that suggests routes for people traveling, such as when going out or going on trips. For example, Patent Document 1 discloses a computing device that calculates the probability of following a route by performing a binary operation on two probability values: a probability derived from a Markov model based on the ease of movement between locations and a probability derived from a topic model based on individual preferences. The ease of movement between locations is derived from the behavioral records of multiple individuals. The computing device determines a recommended route for each person based on the calculated probability values. Furthermore, the computing device recommends routes that satisfy constraints related to travel time.

[0003] Patent Document 1: Japanese Patent Application Publication No. 2011-170453 Summary of the Invention The computing device described in Patent Document 1 uses the ease of transfer between locations derived from the behavioral records of multiple individuals for route determination, and therefore the determined route may not adequately reflect individual preferences.

[0004] This invention was made to solve this problem by providing an information processing system, method, and program that easily reflects personal preferences in daily routines.

[0005] An exemplary embodiment of the present invention relates to an information processing system comprising: a generation unit that generates multiple schedules, including at least one location as a place of stay, based on a relationship in which the schedule is moved between multiple locations as an explanatory variable and the utility of the schedule to the user as a target variable; and an output unit that outputs one or more of the multiple schedules.

[0006] An exemplary embodiment of the present invention relates to an information processing method in which a computer performs the following processing: generating multiple schedules, including at least one location as a place of stay, based on a relationship where the schedule involves moving between multiple locations as an explanatory variable and the utility of the schedule for the user as a target variable; and outputting one or more of the multiple schedules.

[0007] An exemplary embodiment of the present invention relates to an information processing program that causes a computer to perform the following processing: generating multiple schedules, including at least one location as a place of stay, based on a relationship where the schedule involves moving between multiple locations as an explanatory variable and the utility of the schedule for the user as a target variable; and outputting one or more of the multiple schedules.

[0008] Invention Effects According to the present invention, an information processing system, information processing method and program can be provided that can easily reflect personal preferences in daily routines. Attached Figure Description

[0009] Figure 1 This is a block diagram representing an example of the components of a recommender system.

[0010] Figure 2 This is a graph representing an example of a utility function.

[0011] Figure 3 This is a graph representing another example of a utility function.

[0012] Figure 4 An example indicating the status of a schedule.

[0013] Figure 5 Example of a schedule.

[0014] Figure 6 This is a block diagram illustrating an example of the hardware configuration of the information processing apparatus according to the present invention. Detailed Implementation

[0015] Implementation Method 1 Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. Furthermore, for clarity, the following description and drawings have been appropriately omitted and simplified. The accompanying drawings are merely illustrative of one or more embodiments. For the purpose of illustrating exemplary embodiments, not all features or processes shown in any of the figures are necessarily required, and some features or processes may be omitted.

[0016] [Explanation of Composition] Figure 1 This is a block diagram illustrating an example of the structure of a recommender system. The recommender system 100 includes a storage unit 102, a generation unit 104, a selection unit 106, a suggestion unit 108, a receiving unit 110, and an updating unit 112. The storage unit 102 stores a utility function representing the user's utility regarding a schedule involving movement between multiple locations. The utility function is defined by the state of the schedule as an explanatory variable and the utility of the schedule for the user as a target variable. The utility of the schedule represents the degree of preference for the schedule for the user. The degree of preference may represent, for example, psychological liking, convenience, cost, or other advantages. In this example, the explanatory variable is assigned as a discrete state. Furthermore, the utility function is an example of representing the relationship between a schedule involving movement between multiple locations and the user's utility; however, other forms besides the utility function can also be defined as a relationship.

[0017] The status of the schedule includes the following information.

[0018] (1) More than one location shown in the itinerary (e.g., more than one point of origin and more than one point of destination). (2) The time of arrival at each location and the time of movement from each location (3) Modes of movement between locations (1) For example, it can be defined by location information representing a place. Location information can be represented by latitude and longitude, but is not limited to these. Furthermore, the name used to identify a place can be included in the location information of the schedule. In other words, (2) it can be said that the length of stay at each place is defined. In (3), the mode of travel includes, for example, walking, cycling, car, rail, airplane, etc., but is not limited to these. The information on the mode of travel may also include at least one of the following: speed of travel, direction of travel, and timetable used when traveling from the origin to the destination.

[0019] Figure 2 This is a graph representing an example of a utility function. Figure 2 The x-axis of the utility function represents the state of the schedule as an explanatory variable. The y-axis of the utility function represents the probability density p(i) of the user's utility corresponding to the state i of the schedule as the target variable. Figure 2 In the given equation, the probability densities p(i) for states i1, i2, and i3 are p1, p2, and p3, respectively. The probability density p(i) reaches its maximum value (the only maximum) p1 in state i1, and its value decreases as the state changes to i2 and i3. That is, Figure 2 The utility function is unimodal.

[0020] Figure 3 This is another graph representing a utility function. The explanation of the x-axis and y-axis of the utility function is as follows: Figure 2 Same, therefore omitted. Figure 3 In the given diagram, the probability densities p(i) for states i1, i2, and i3 are p1, p4, and p3, respectively. Figure 3 In the given condition, the probability density p(i) becomes the maximum value p1 in state i1, and while it is not the maximum value in state i2, it becomes a maximum value p4. That is, Figure 3 The utility function is multimodal. p4 is the ratio Figure 2 The larger value of p2.

[0021] Furthermore, the shape of the utility function is not limited to Figure 2 and 3 The shape shown. For example, as a utility function, it can be set to have more than three maxima. Furthermore, the utility function can also be... Figure 2 and 3 Other forms of representation besides the probability distribution shown. For example, a histogram can be cited as another form.

[0022] Figure 4 An example indicating the status of a schedule. Figure 4 The graph shown uses the x-axis to represent location and the y-axis (t-axis) to represent time. Each state i1 to i3 corresponds to... Figure 2 The states i1 to i3 shown in 3 represent the following schedule.

[0023] (Status i1) Home: Departs at 10:00 AM => Football Stadium: Arrives at 10:20 AM / Departs at 12:45 PM => Restaurant: Arrives at 1:30 PM / Departs at 2:30 PM => Home: Arrives at 3:30 PM (Status i2) Home: Departs at 10:00 AM => Restaurant: Arrives at 11:00 AM / Departs at 11:50 AM => Coffee Shop: Arrives at 12:10 PM / Departs at 12:50 PM => Football Stadium: Arrives at 1:05 PM / Departs at 3:10 PM => Home: Arrives at 3:30 PM (Status i3) Home: Departs at 10:00 AM => Restaurant: Arrives at 11:00 AM / Departs at 11:40 AM => Home: Arrives at 12:40 PM / Departs at 1:20 PM => Restaurant: Arrives at 2:20 PM / Departs at 3:00 PM => Home: Arrives at 4:00 PM Furthermore, the mode of transportation between locations in schedule states i1 to i3 is by car. Schedule states i1 to i3 include the information shown in (1) to (3).

[0024] The utility function is, for example, a function generated based on the user's behavior records. As an example, the user's behavior records are obtained by determining location information using a Global Navigation Satellite System (GNSS) method on the user's terminal. The utility function derivation device derives the utility function from the user's behavior records using any method such as econometrics or deep learning. The probability distribution of the utility function for scheduling can be a white-box function (e.g., an econometric model) or a black-box function (e.g., a deep learning model). Furthermore, as described later, the utility function can also be updated by the update unit 112. The derivation device can be located within the recommendation system 100 or outside of the recommendation system 100.

[0025] The locations shown in the schedules of each state in the utility function include locations the user has visited in the past, as shown in the user's behavior records. In this case, the schedules of each state in the utility function can be schedules actually experienced by the user in past actions. The locations shown in the schedules of each state can include locations the user has not visited in the past. For example, the exporting device sets multiple locations as candidate travel destinations and extracts arbitrary combinations of multiple locations through sampling. The exporting device generates a schedule of movement between the multiple locations shown in the extracted combinations as a state. The generation unit 104 generates multiple schedules (i.e., multiple states) by extracting multiple different combinations as combinations of multiple locations. The generated multiple states become the states in the utility function.

[0026] The exporting device can categorize locations and assume that locations belonging to the same category have the same utility value. Categories are generated, for example, as hierarchical categories. As an example, suppose a user has visited store A of a chain store but not store B of the same chain. Store A is located near the user's home, while store B is located far away. In this case, the exporting device determines that stores A and B can be classified into the same chain store category, and can assume that store B's utility is the same as store A's.

[0027] When sufficient user behavior data is available, the exporting device can use this data to cover category information (candidate destinations) of the locations the user might visit. Therefore, the exporting device can generate a utility function reflecting the preferences of all users using only the user's behavior data. Conversely, when insufficient user behavior data is available, the exporting device supplements the category information of the locations the user might visit with information representing utility related to other users' locations or schedules.

[0028] Upon receiving a user's request for a schedule notification from the receiving unit 110, the generation unit 104 generates schedules as notification candidates. Specifically, the generation unit 104 generates and extracts schedules by applying the Markov chain Monte Carlo method to a utility function (i.e., sampling schedules). Here, the Metropolis-Hastins method within the Markov chain Monte Carlo method will be described. However, the method used by the generation unit 104 is not limited to this.

[0029] The generation unit 104 first sets a schedule as the initial schedule. For example, the generation unit 104 can set the schedule of a certain state in the utility function as the initial schedule. The generation unit 104 can set the schedule in the utility function where the utility reaches its maximum value as the initial schedule.

[0030] Furthermore, as another example, suppose that the user's initial location and last location are the same during the period of the schedule specified in the prompt request (hereinafter also referred to as the specified period). In this case, the generation unit 104 can set the schedule in which the user's location remains unchanged throughout the specified period as the initial schedule. For example, the generation unit 104 can set the user's location to home and set the schedule in which the user stays at home throughout the specified period as the initial schedule. Alternatively, it is also possible to consider the case in which the user's initial location and last location are different during the specified period. In this case, the generation unit 104 can set the user's initial location and last location and set the schedule in which the user moves once between the initial location and the last location as the initial schedule.

[0031] Additionally, the generation unit 104 sets an achievable schedule as an initial schedule, which includes at least one of the following constraints: environmental conditions, movement mode constraints, or schedule constraints. Environmental conditions refer to conditions specifying a location within a distance that can be traveled from the user's current location. Movement mode constraints refer to conditions specifying how much movement is possible in time and space using the specified movement mode, and to which location the specified movement mode can be reached. For example, if the movement mode is a car, the constraints would specify the car's speed and the location of roads the car can travel on. Schedule constraints are at least one of the following conditions: time or location of the schedule indicated by the user in the prompt request. Time constraints include information such as the schedule's start time (e.g., departure time from home), schedule's end time (e.g., return time from home), and at least one of the arrival or departure times of more than one destination. Location constraints include information about the location the user wants to move to.

[0032] The generation unit 104 performs any one of the four operations shown below once or more on the set initial schedule, thereby generating a new schedule that can be implemented.

[0033] (A) Add Operation: The operation of adding one mobile destination to the schedule. (B) Deletion operation: The operation of deleting a location for a mobile destination in the schedule. (C) Mode Operation: Operation to change the mode of movement for one move in the schedule. (D) Extension / Shortening Operations: Operations that increase or decrease the stay time at one location in the itinerary. Figure 5Examples of various operations for the schedule are shown below. As the initial state, assume a schedule where the user moves to location x1 by car after a stay of period T1 at location x0, and then moves to location x2 by car after a stay of period T2 at location x1. If operation (A) is applied to the initial schedule by the generation unit 104, for example, a schedule is generated where the user moves to location x3, which is between location x1 and location x2. In this schedule, the user moves to location x1 by car after a stay of period T1 at location x0, and then moves to location x3 by car after a stay of period T3 at location x1. The user then moves to location x2 by car after a stay of period T4 at location x3. Here, period T3 + T4 is equal to period T2. The locations added as destinations and the periods spent at the added locations are arbitrary.

[0034] When the generation unit 104 applies operation (B) to the original schedule, for example, a schedule is generated that deletes location x1 as the destination. That is, location x1, which is located between the initial location x0 and the destination x2, is deleted. In this schedule, the user moves to location x2 by car after a stay of period T5 at location x0. Period T5 is equal to period T1+T2.

[0035] When the generation unit 104 applies operation (C) to the original schedule, for example, the mode of movement from position x1 to position x2 is changed from a car to a bicycle. The time of arrival at position x2 is the same as in the original schedule, but the bicycle's speed is slower than the car's. Therefore, the dwell time at position x1 in the generated schedule is shorter than the period T2. The method of changing the mode of movement is arbitrary.

[0036] When the generation unit 104 applies operation (D) to the original schedule, for example, the dwell time at position x1 is shortened and the dwell time at position x2 is lengthened. It is arbitrary which position's dwell time is extended or shortened.

[0037] Furthermore, the generation unit 104 performs any one of the four operations on the initial schedule one or more times, thereby generating a schedule for any state in the utility function. The converse is also true. That is, performing an operation on any state of the schedule will lead to schedules for all other states. Moreover, the output probabilities in the Markov model follow the utility function. Therefore, the operations on the schedule satisfy the ergodicity of the Metropolis-Hastington method. The generation unit 104 can generate any schedule corresponding to the probability distribution of the utility function.

[0038] The generation unit 104 selects a location as the travel destination from a plurality of candidate locations used in the derivation of the utility function. The generation unit 104 generates multiple schedules from the candidate locations, each including at least one location as a stop. Furthermore, the generation unit 104 can perform the above operations to generate new schedules, making the generated schedule an achievable schedule with at least one of the following conditions: environmental conditions, travel mode constraints, or schedule constraints.

[0039] The selection unit 106 selects one or more schedules from the multiple schedules generated by the generation unit 104. The multiple schedules generated by the generation unit 104 are usually multiple, but the selection unit 106 selects one or a limited number of schedules that meet the conditions specified in the prompt request received from the receiving unit 110.

[0040] The specified conditions may include, for example, at least one of the following conditions.

[0041] (i) The condition of the distance in the utility function between the state corresponding to the selected schedule and the state corresponding to the schedule whose utility is maximized. (ii) The condition for using the state corresponding to the selected schedule as the explanatory variable of the utility function, i.e., the target variable, i.e., the utility value. In (i), the distance in the utility function refers to Figure 2 Or the distance between states on the x-axis in 3. The distance in the utility function corresponds, for example, to the number of times operations (A) to (D) above have been performed. In the example of (i), the state corresponding to the selected schedule can be more than a specified distance away from the state corresponding to the schedule whose utility is maximized in the utility function. This means selecting a schedule with low similarity to the schedule with the highest utility (i.e., the highest preference) for the user. As another example, (i) can be a state corresponding to the selected schedule that is less than a specified distance away from the state corresponding to the schedule whose utility is maximized in the utility function.

[0042] Examples of (ii) are at least one where the utility as the target variable is at its maximum or above a specified value. A maximum utility means that the utility function shows a higher utility for the state corresponding to the selected schedule compared to the state of a similar schedule.

[0043] For example, suppose (i) a condition is specified that the state corresponding to the selected schedule is more than a predetermined value away from the state corresponding to the schedule whose utility in the utility function is maximized. Also suppose condition (ii) is specified. In this case, the selection unit 106 can select a schedule that is different from the user's favorite schedule but has greater utility for the user.

[0044] When the utility function is multimodal, the selection unit 106 can more easily select a schedule that satisfies both conditions (i) and (ii) above than when the utility function is unimodal. For example, suppose that as condition (i), it is specified that "the state corresponding to the selected schedule is at least 3 units away from the state corresponding to the schedule whose utility function is at its maximum value".

[0045] As an example, suppose the utility function is as follows: Figure 2 As shown, states i1 to i3 are as follows: Figure 4 The situation is as shown. At this time, the selection unit 106 can select the schedule of state i2 generated by the generation unit 104. However, the utility value p2 in state i2 is not a large value. At this time, the schedule of state i2 may not satisfy the condition (ii). Furthermore, even if the schedule of state i2 is selected, although the schedule is "unexpected" for the user, the degree of preference is relatively small.

[0046] On the other hand, assuming the utility function is as follows Figure 3 As shown, states i1 to i3 are as follows: Figure 4 The situation is as shown. At this time, the selection unit 106 can select the schedule of state i2 generated by the generation unit 104. Here, the utility value p4 in state i2 becomes a value greater than p2. At this time, the schedule of state i2 is likely to satisfy condition (ii). Then, if the schedule of state i2 is selected, it is not the schedule that the user most often selects, so it is "unexpected", and the degree of preference is relatively high.

[0047] The prompting unit 108 recommends one or more schedules to the user by prompting (i.e., outputting) the selection unit 106 to select one or more schedules. The prompting unit 108 can prompt the user to schedules, for example, by displaying the schedules or by providing voice notifications.

[0048] The receiving unit 110 includes input interfaces such as a touch panel and buttons. The receiving unit 110 receives a schedule prompt request by receiving operations from the user. The prompt request may include at least one of the following information used by the generation unit 104 when generating the schedule: the period of the requested schedule, environmental conditions, constraints on the mode of travel, or constraints on the schedule itself. As described above, the prompt request may include specified conditions for the selection unit 106 to select a schedule. Furthermore, by operating the receiving unit 110, the user can select one or more schedules they prefer from two or more schedules prompted by the prompt unit 108.

[0049] The update unit 112 can update the utility functions stored in the storage unit 102 by using one or more schedules specified by the receiving unit 110. For example, the update unit 112 uses one or more specified schedules to update the user's behavior records. The update unit 112 uses the updated user behavior records to generate new utility functions. The method for generating utility functions is as described above. Alternatively, the update unit 112 can also directly update the utility functions stored in the storage unit 102 using one or more specified schedules.

[0050] [Explanation of Effects] In the recommendation system 100 of the present invention, the generation unit 104 generates multiple schedules, each including at least one of the multiple locations as a stopover, based on a utility function representing the user's utility regarding a schedule involving movement between multiple locations. The suggestion unit 108 suggests one or more of the multiple schedules. Since the suggested schedules are determined based on the utility function representing the user's utility, the user's preferences are easily reflected in the schedules. Therefore, the recommendation system 100 can propose personalized schedules based on the user's preferences for various purposes such as outings and travel. Furthermore, the generation unit 104 can also generate schedules that take into account spatiotemporal constraints.

[0051] Furthermore, the selection unit 106 can select one or more schedules based on the distance in the utility function between the state corresponding to the output schedule and the state corresponding to the schedule whose utility function is maximized. For example, the recommendation system 100 can recommend schedules with low similarity to the schedule with the highest utility for the user. In particular, when the utility function is multimodal, the recommendation system 100 tends to recommend schedules with relatively high utility. The recommendation system 100 can also recommend the schedule with the highest utility for the user or schedules similar to that schedule.

[0052] Furthermore, the updating unit 112 can update the utility function using one or more of the two or more schedules suggested by the prompting unit 108. By providing feedback to the utility function for the specified schedule, the utility of the location indicated by the specified schedule becomes greater in the utility function. Since the utility function becomes a more accurate reflection of the user's preference level, the recommender system 100 is able to make future recommended schedules more accurately reflect the user's preferences.

[0053] The processing of the recommendation system 100 can also be achieved by having the computer's processor execute computer programs.

[0054] Figure 6 This is a block diagram illustrating an example of the hardware configuration of an information processing device (in other words, a computer) that performs the processing of the recommendation system 100. (See reference) Figure 6 The information processing device 90 includes a signal processing circuit 91, a processor 92, and a memory 93.

[0055] The signal processing circuit 91 is a circuit that processes signals according to the control of the processor 92. Additionally, the signal processing circuit 91 may include communication circuitry for transmitting or receiving signals from other devices.

[0056] Processor 92 is connected to memory 93 and performs the processing of the system described in the above embodiments by reading from memory 93 and executing computer programs. Processor 92 has one or more processors. Examples of processor 92 include central processing unit (CPU), microprocessor (MPU), field-programmable gate array (FPGA), etc., but it is not limited to these.

[0057] The memory 93 is composed of volatile memory, non-volatile memory, or a combination of volatile and non-volatile memory. The memory 93 is not limited to one unit; multiple units may be provided. Furthermore, the volatile memory may be, for example, Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), or other Random Access Memory (RAM). The non-volatile memory may be, for example, Programmable Random Only Memory (PROM), Erasable Programmable Read Only Memory (EPROM), other Read Only Memory (ROM), flash memory, or Solid State Drive (SSD).

[0058] The memory 93 is used to store one or more commands. A command is a command used to instruct the computer to perform the algorithm described in this invention. Here, one or more commands are stored as a program in the memory 93. The processor 92 reads the stored program from the memory 93 and executes it. By executing the program, the information processing described in the embodiments can be realized.

[0059] The program can be stored on a non-transitory computer-readable medium or a tangible storage medium. By way of example, and not limitation, computer-readable media or tangible storage media include memory technologies such as RAM, ROM, flash memory, and SSDs. Furthermore, computer-readable media or tangible storage media can include optical disc storage devices such as Compact Disc Read-Only Memory (CD-ROM), Digital Versatile Disk (DVD), and Blu-ray discs. Moreover, computer-readable media or tangible storage media can include magnetic tape, magnetic tape, disk storage, and magnetic storage devices. The program can be transmitted on a temporary computer-readable medium or a communication medium. By way of example, and not limitation, temporary computer-readable media or communication media include propagation signals such as electricity, light, and sound. Temporary computer-readable media can provide the program to the computer via wired or wireless communication paths such as wires and optical fibers.

[0060] The present invention has been described above with reference to the embodiments described above, but the present invention is not limited to the above embodiments. Various modifications that can be understood by those skilled in the art can be made to the structure or details of the present invention within the scope of the present invention.

[0061] Symbol Explanation 100 - Recommendation system, 102 - Storage department, 104 - Generation department, 106 - Selection department, 108 - Prompt department, 110 - Receiving department, 112 - Update department.

Claims

1. An information processing system, characterized in that, have: The generation department generates multiple schedules, each including at least one stop, based on a relationship where the schedule will move between multiple locations as an explanatory variable and the utility of the schedule for the user as the target variable. and The output unit outputs one or more of the multiple schedules.

2. The information processing system according to claim 1, characterized in that, The output unit selects and outputs the one or more schedules based on the distance in the relationship between the explanatory variables corresponding to the one or more schedules output and the explanatory variables corresponding to the schedules whose utility is maximized in the relationship.

3. The information processing system according to claim 1 or 2, characterized in that, The output unit outputs two or more of the multiple schedules. The information processing system further includes an update unit, which updates the relationship using one or more of the two or more output schedules.

4. An information processing method, characterized in that, The computer performs the following processing: Based on the relationship between schedules that will move between multiple locations as explanatory variables and the utility of the schedule for the user as the target variable, generate multiple schedules that include at least one location as a stop. and Output one or more of the multiple schedules.

5. An information processing program, characterized in that, The computer will perform the following processing: Based on the relationship between schedules that will move between multiple locations as explanatory variables and the utility of the schedule for the user as the target variable, generate multiple schedules that include at least one location as a stop. and Output one or more of the multiple schedules.

Images