WO2014103560A1 - 分析装置、分析プログラム、分析方法、推定装置、推定プログラム、及び、推定方法。 - Google Patents
分析装置、分析プログラム、分析方法、推定装置、推定プログラム、及び、推定方法。 Download PDFInfo
- Publication number
- WO2014103560A1 WO2014103560A1 PCT/JP2013/081118 JP2013081118W WO2014103560A1 WO 2014103560 A1 WO2014103560 A1 WO 2014103560A1 JP 2013081118 W JP2013081118 W JP 2013081118W WO 2014103560 A1 WO2014103560 A1 WO 2014103560A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- analysis
- unit
- event
- periods
- state
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/02—Input arrangements using manually operated switches, e.g. using keyboards or dials
- G06F3/023—Arrangements for converting discrete items of information into a coded form, e.g. arrangements for interpreting keyboard generated codes as alphanumeric codes, operand codes or instruction codes
- G06F3/0233—Character input methods
- G06F3/0237—Character input methods using prediction or retrieval techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
Definitions
- the present invention relates to an analysis device, an analysis program, an analysis method, an estimation device, an estimation program, and an estimation method.
- a replacement unit that replaces each event in an impulse-like event sequence with a plurality of rectangular pulses, and an event sequence generated by the replacement of at least one change point of the plurality of rectangular pulses.
- a division unit that divides by time, a state vector generation unit that generates a state vector corresponding to a plurality of rectangular pulse states, and a plurality of states corresponding to a plurality of periods
- An analysis device including an analysis unit that performs a Poisson regression analysis using a vector, an analysis method using the analysis device, and an analysis program that causes a computer to function as the analysis device are provided.
- an estimation device for estimating a response to an event sequence based on an analysis result by the analysis device, wherein each event in the input event sequence is replaced with a plurality of rectangular pulses.
- a division unit that divides the event sequence generated by replacement at the time of at least one change point of the plurality of rectangular pulses, and a plurality of rectangular pulses corresponding to each of the plurality of divided periods.
- An estimation device including an estimation unit for estimating the number of times, an estimation method using the estimation device, and a computer as the estimation device It provides an estimate program to function.
- the structure of the analyzer 10 and the estimation apparatus 20 of this embodiment is shown.
- the flow of the analysis process of the analyzer 10 of this embodiment is shown.
- An example of an event history input by the replacement unit 102 is shown.
- the individual response approximated by a plurality of rectangular pulses is shown.
- generates is shown.
- segments is shown.
- An example of a state vector generated by the state vector generation unit 106 is shown.
- a specific example of a state vector generated by the state vector generation unit 106 is shown.
- An example of the weight vectors w 1 to m stored in the storage unit 116 is shown.
- the flow of the estimation process of the estimation apparatus 20 of this embodiment is shown. 2 shows an example of a hardware configuration of a computer 1900.
- FIG. 1 shows the configuration of the analysis device 10 and the estimation device 20 of this embodiment.
- the analysis apparatus 10 analyzes an event history in which an impulse-like event for an individual and an individual's reaction to the event are recorded. For example, the analysis device 10 records product campaign information advertisements (impulse-like events for individuals) to hundreds of thousands to tens of millions of customers, and records of customer purchases (individual responses) for each customer. Analyze the history of events it contains.
- the analysis apparatus 10 includes a replacement unit 102, a division unit 104, a state vector generation unit 106, a reaction frequency calculation unit 108, an analysis unit 110, and a storage unit 116.
- the replacement unit 102 replaces each event in the impulse-like event series with a plurality of rectangular pulses. Thereby, the replacement unit 102 can approximate the response function corresponding to the event and the individual for each impulse-like event with a plurality of rectangular pulses.
- the replacement unit 102 extracts an event series including the event time and the reaction time of the individual from the event history stored in an external database or the like.
- the replacement unit 102 generates, from the extracted event series, a rectangular pulse having a predetermined time width starting from the time when the event occurs.
- the replacement unit 102 generates a plurality of rectangular pulses having a plurality of different predetermined time widths.
- the replacement unit 102 includes a plurality of event series (for example, a series of direct mail transmission events including advertisements and a series of web advertisement browsing events) in which events corresponding to a plurality of channel types are arranged in time series. ) May be replaced with a plurality of rectangular pulses (for example, direct mail transmission or web advertisement browsing).
- the replacement unit 102 supplies the individual reaction time and information on a plurality of rectangular pulses to the dividing unit 104.
- the dividing unit 104 divides the periods included in the plurality of event series at the time of at least one change point of the plurality of rectangular pulses in at least one of the plurality of event series. For example, the dividing unit 104 divides a period included in the event sequence into a plurality of periods at the rising and falling times in at least one of the plurality of rectangular pulses replaced by the replacing unit 102 from the plurality of event sequences. The dividing unit 104 supplies a plurality of rectangular pulses and information on a plurality of divided periods to the state vector generating unit 106. In addition, the dividing unit 104 supplies the reaction time calculation unit 108 with the reaction time of the individual and information on the plurality of divided periods.
- the state vector generation unit 106 generates a state vector corresponding to the state of the plurality of rectangular pulses for each of the plurality of event sequences, corresponding to each of the plurality of divided periods. For example, the state vector generation unit 106 generates a state vector having a plurality of elements corresponding to the state of each rectangular pulse for each of a plurality of divided periods.
- the state vector generation unit 106 sets the value of the corresponding element of the state vector to 1 and the rectangular pulse has a value of 0.
- the value of the corresponding element of the state vector may be zero.
- the state vector generation unit 106 supplies the generated state vector to the analysis unit 110.
- the reaction frequency calculation unit 108 receives the event sequence and calculates the reaction frequency that is the number of reactions that occurred during the period corresponding to each of the plurality of periods. For example, the reaction number calculation unit 108 counts the number of reactions during the period based on whether the individual reaction time is included in each of the plurality of divided periods. The reaction frequency calculation unit 108 supplies the calculated reaction frequency for each of a plurality of periods to the analysis unit 110.
- the analysis unit 110 performs Poisson regression analysis using a plurality of state vectors corresponding to a plurality of periods and a plurality of reaction times corresponding to the plurality of periods.
- the analysis unit 110 includes a conversion unit 112 and an optimization unit 114.
- the conversion unit 112 converts a plurality of state vectors corresponding to a plurality of periods into a plurality of feature vectors using a mapping function.
- the conversion unit 112 supplies the feature vector to the optimization unit 114.
- the optimization unit 114 optimizes the weight vector so as to maximize the probability that the number of reactions can be obtained from the Poisson distribution based on the scalar score obtained by the inner product of the feature vector and the weight vector in a plurality of periods.
- the scalar score represents the magnitude of the stimulus when the stimulus from the event sequence from multiple channels is replaced with a single impulse-like stimulus at a certain point in time.
- the weight vector is a vector quantity that represents the characteristics of an individual's response to a stimulus caused by an event, and may be different for each individual.
- the optimization unit 114 uses the inner product of the feature vector and the weight vector as a scalar score, and the number of reaction events recorded in the history from the Poisson process in which the average number of reactions per unit time is an exponential function of the scalar score is For the logarithm of the generated probabilities, the weight vector is optimized so that the sum over the periods is maximized.
- the optimization unit 114 causes the storage unit 116 to store the optimized weight vector.
- the storage unit 116 stores the weight vector obtained by the optimization unit 114 for each individual.
- the storage unit 116 may supply the weight vector to the estimation device 20 as required.
- the analysis apparatus 10 divides the event sequence into a plurality of periods to generate state vectors for each period, and the number of reactions expected by the Poisson process based on the state vector and the weight vector in each period is the history response.
- the weight vector is optimized so as to maximize the probability of matching the number of times. For this reason, according to the analyzer 10, it is possible to construct a Poisson process model in consideration of a dynamic event.
- the analysis device 10 divides the event sequence into a plurality of periods at the time of the change point of the rectangular pulse based on the event, thereby reducing the amount of calculation compared to the method of discretizing the event sequence into a period of a certain length. can do.
- the estimation device 20 estimates an individual's response to the event sequence based on the individual weight vector obtained based on the analysis result of the analysis device 10.
- the estimation apparatus 20 includes a replacement unit 202, a state vector generation unit 206, a storage unit 208, and an estimation unit 210.
- the replacement unit 202 replaces each event in the input event series with a plurality of rectangular pulses. For example, the replacement unit 202 inputs an event sequence assumed in the simulation. Similar to the replacement unit 102, the replacement unit 202 generates a rectangular pulse having a predetermined time width starting from the time point of the event included in the input event series. The replacement unit 202 supplies information on a plurality of rectangular pulses for a plurality of event sequences to the state vector generation unit 206.
- the state vector generation unit 206 generates a state vector corresponding to the state of a plurality of rectangular pulses for each of a plurality of event sequences, corresponding to each of a plurality of periods. For example, the state vector generation unit 206 divides periods included in a plurality of event sequences at predetermined time intervals.
- the state vector generation unit 206 generates a state vector having a plurality of elements corresponding to the state of each rectangular pulse in the same manner as the state vector generation unit 106 for each divided period.
- the state vector generation unit 206 supplies the generated state vector for each of a plurality of periods to the estimation unit 210.
- the storage unit 208 receives weight vector information from the storage unit 116 of the analyzer 10 and stores it.
- the storage unit 208 supplies the weight vector to the estimation unit 210.
- the estimation unit 210 estimates the number of reactions indicating the number of reactions corresponding to each of the plurality of periods, from the plurality of state vectors corresponding to each of the plurality of periods, using the analysis result by the analysis apparatus 10. For example, first, the estimation unit 210 converts a plurality of state vectors into a plurality of feature vectors using a mapping function, like the conversion unit 112.
- the estimation unit 210 calculates a scalar score obtained from the inner product of the feature vector and the weight vector for each of a plurality of periods.
- the estimation unit 210 simulates the number of reactions for each of a plurality of periods by a Monte Carlo method from a Poisson distribution having an expected value as the product of the exponent function of the scalar score and the length of the period.
- the estimation unit 210 calculates the total number of reaction times y i in the entire simulation period by calculating the total of the simulated number of reactions.
- the estimation unit 210 can also calculate a distribution relating to the total number of reaction events by changing the seed of the random number and repeating the simulation.
- the estimation device 20 of the present embodiment can estimate a response to an event for each individual using the weight vector obtained by the analysis device 10. Furthermore, the estimation device 20 can quantitatively predict the magnitude of the response of a large number of individuals to a future event by executing a simulation on a plurality of individuals.
- FIG. 2 shows a flow of analysis processing of the analyzer 10 of the present embodiment.
- the analysis apparatus 10 analyzes the response of the individual to the event series in the history by executing the processing from S102 to S114.
- the replacement unit 102 replaces each event in the impulse-like event sequence for a specific individual with a plurality of rectangular pulses. For example, first, the replacement unit 102 inputs an event history stored in a database or the like.
- FIG. 3 shows an example of an event history input by the replacement unit 102.
- the history in FIG. 3 includes individual numbers (ID (i)), and includes names of a plurality of customers (Name) as individuals. Also, the history includes the first to oth event series (where o is a natural number) and includes the reaction time of the individual.
- the history includes the time when the direct mail is transmitted to the customer as the first event sequence (channel 1), and the customer browses the web advertisement as the second event sequence (channel 2). Includes time.
- the history may include the time when the customer purchased the product as an individual response to an event stimulus from a plurality of channels.
- the advertisement of the product is transmitted to “Taro Yamada” by direct mail as 2012/1/24 and 2012/3/20 as the first event series. Shows that we have seen WeB advertisements on 2012/2/4 and 2012/3/4 as the event series, and that “Taro Yamada” purchased the product related to the advertisement on 2012/2/10 as a response It is.
- the replacement unit 102 extracts the time of the first event and the time of the second event related to “Taro Yamada” from the history shown in FIG.
- the replacement unit 102 generates a rectangular pulse having a predetermined time width starting from the time of the extracted event.
- the replacement unit 102 generates rectangular pulses having a time width of one day, one week, two weeks, and four weeks starting from the time of the event.
- FIG. 4 shows the relationship of individual responses approximated by a plurality of rectangular pulses.
- a curved line represented by a dotted line represents an ideal response to an individual's stimulus.
- the individual shows the highest response immediately after the stimulation by the event, and the response to the stimulation gradually decreases with time.
- the absolute strength of the response to the stimulus by the same event and the mode of response decay vary from individual to individual.
- the dotted line curve in FIG. 4A shows an ideal response of an individual sensitive to a stimulus
- the dotted line curve in FIG. 4B shows an ideal response of an individual insensitive to a stimulus.
- an individual who is sensitive to a stimulus has a high response immediately after the stimulation by an event, but the response may rapidly decrease after the stimulation.
- an individual who is insensitive to a stimulus may not have a high response immediately after the stimulus due to the event, but may continue for a relatively long time after the stimulus. .
- the solid line in FIG. 4 shows the response of the individual approximated by a plurality of rectangular pulses.
- an ideal response curve can be approximated by a step function in which a plurality of rectangular pulses having different time widths are superimposed.
- the time width (1 day, 1 week, 2 weeks, and 4 weeks) of each rectangular pulse constituting the step function is the same for all individuals.
- the heights h1 to h4 of each rectangular pulse are different for each individual.
- the ideal response curve of an individual can be approximated by a combination of heights of a plurality of rectangular pulses having a predetermined time width.
- the combination of the heights of a plurality of rectangular pulses that each individual has corresponds to a weight vector w i for each individual described later.
- i 1 of the individual weight vector w 1 1-4 th element W 1_1 ⁇ W of (Taro Yamada) 1_4 are ideal duration 1 day that approximates the response to stimuli of the individual, 1 week, Corresponding to the respective heights h1 to h4 of the rectangular pulses of 2 weeks and 4 weeks.
- channels 1 stimulus and channel 2 stimulus events included in two event series are represented in an impulse form.
- Channel 1 windows 1 to 4 and “channel 2 windows 1 to 4” indicate a plurality of rectangular pulses replaced from events of the event series of the respective channels.
- the replacement unit 102 raises the pulse at the start of the rectangular pulse to a high level state having a value of 1, and lowers the pulse at the end of the pulse after the elapse of a predetermined time width to have a value of 0. It may be in a low level state. Instead, the replacement unit 102 may assign other integers or real numbers to the values of the high level state and the low level state of the rectangular pulse. Further, when a plurality of rectangular pulses partially overlap in the same window in the same window, the rectangular pulse values of the corresponding period are integrated according to some standard to calculate one value, which is used as a state value. . As this criterion, a total value is usually used, but an OR operation using a bit representation may be used. The replacement unit 102 divides the reaction time and the information on the time and value at the start and end of a plurality of rectangular pulses for a specific individual (for example, “Taro Yamada” in the history of FIG. 3). To supply.
- the dividing unit 104 divides the periods included in the plurality of event series at the time of at least one change point of the plurality of rectangular pulses in at least one of the plurality of event series.
- FIG. 6 shows an example of a period included in the event series divided by the dividing unit 104.
- the dividing unit 104 is included in the event series based on the times t1, t2, t3,... Of all the change points such as the rising and falling edges of the plurality of rectangular pulses generated from the plurality of event series.
- the entire period is composed of a plurality of periods (for example, a period T 1 from time t1 to immediately before t2, a period T 2 from time t2 to immediately before t3, a period T from time t3 to immediately before t4). 3 ).
- the dividing unit 104 supplies a plurality of rectangular pulses and information on a plurality of divided periods to the state vector generating unit 106. In addition, the dividing unit 104 supplies the reaction time and information on a plurality of divided periods to the reaction number calculating unit 108.
- the state vector generation unit 106 generates a state vector corresponding to the state of the plurality of rectangular pulses for each of the plurality of event series, corresponding to each of the plurality of divided periods. For example, for the individual i, the state vector generation unit 106 has a plurality of elements corresponding to the state of the rectangular pulse corresponding to the period T j (j is an integer) from time tj to immediately before time t (j + 1). ij is generated.
- FIG. 7 shows an example of a state vector generated by the state vector generation unit 106.
- FIG. 8 shows a specific example of a state vector generated by the state vector generation unit 106.
- FIG. 8A shows the event sequence shown in FIG. 6, and
- FIG. 8B shows the state vector generated by the state vector generation unit 106 in the event sequence period t 7 in FIG. 8A.
- Rectangular pulse value (0), channel 1 window 2 rectangular pulse value (0), channel 1 window 3 rectangular pulse value (0), channel 1 window 4 rectangular pulse value (1), channel 2 window 1 rectangular pulse value (0), channel 2 window 2 rectangular pulse value (0), channel 2 window 3 rectangular pulse value (1), and channel 2 window 4 rectangular pulse value (1) Are generated as elements.
- the state vector generation unit 106 supplies the generated state vector X ij to the analysis unit 110.
- the reaction frequency calculation unit 108 calculates the number of reactions y ij receives the event sequence is the number of reactions that occur during the period T j. For example, the number-of-reactions calculation unit 108 counts the number of individuals i that include the reaction time in each of a plurality of divided periods T j shown in FIG.
- the conversion unit 112 converts the plurality of state vectors Xij into a plurality of feature vectors.
- the conversion unit 112 converts the state vector X ij into the corresponding feature vector ⁇ (X ij ) by using an arbitrary mapping function ⁇ : R d1 ⁇ R d2 designed in advance.
- d1 is the dimension of the state vector
- d2 is the dimension of the feature vector.
- d1 and d2 may be the same.
- the conversion unit 112 by adding a second-order correlation terms in the state vector X ij, may convert the state vector X ij to the feature vector [Phi (X ij).
- the optimization unit 114 optimizes the weight vector so that the probability function is maximized. Specifically, first, the optimization unit 114 calculates a scalar score based on the inner product of the feature vector X ij and the weight vector w i for each period j of the individual i. Next, the optimization unit 114 calculates the logarithm of the probability that the number of reactions y ij is generated from the Poisson distribution in which the product of the calculated exponent function of the scalar score and the length of the period j is an expected value. Finally, the optimization unit 114 optimizes the weight vector w i so that the sum of log probabilities in all periods is maximized.
- the optimization unit 114 may use maximum likelihood estimation, MAP estimation, haze estimation, or the like as a technique for optimizing Equation 1.
- the optimization unit 114 uses the probability function 1 of Equation 1 as a probability function l.
- a function obtained by removing the normalization term for the number of reactions from the logarithmic probability mass function based on the above is used.
- the optimization unit 114 can perform optimization considering the irreversibility of time by removing the normalization term.
- the optimization unit 114 uses a function represented by the following expression as the probability function l.
- y is the actual number of reactions
- z is a scalar score, which is a logarithm of the expected value of the number of reactions
- ⁇ represents the length of the target period.
- the vector w 0 in Equation 1 is a response parameter that does not depend on an individual.
- pen (w 0 , w 1 , w 2 ,... w m ) is a normalization term and normally uses a convex function, but may use a non-convex function.
- the optimization unit 114 prevents excessive adaptation to the history in which the weight vectors w 1 to m are input, and prevents the fitness from being lowered for an event sequence different from the history.
- the optimization unit 114 can result in Equation 1 as a convex optimization problem, and thus can calculate a global optimal solution by the gradient method. Thereby, the optimization part 114 can optimize Formula 1 stably.
- the optimization unit 114 may be used L 2 normalization term as normalization term for overfitting prevention.
- the optimization unit 114 uses the L 2 normalization term as Can be used.
- c 0 and c is a hyper-parameter determined by cross-test, or the like.
- the optimization unit 114 may use the L 1 normalization term instead of the L 2 normalization term.
- the optimization unit 114 may use a mathematical expression that does not include a normalization term for preventing excessive adaptation.
- the optimization unit 114 may assign the state vector X ix period T x for the two periods T x + period T (x + 1). Moreover, the optimization unit 114 for the duration T x + period T (x + 1), and the state vector X ix period T x, period adjacent to the period T x T (x + 1) of the state vector X i (x + 1) May be assigned a state vector obtained from the average of.
- the optimization unit 114 stores the optimized m individual weight vectors w 1 to m in the storage unit 116.
- FIG. 9 shows an example of the weight vectors w 1 to m stored in the storage unit 116.
- the storage unit 116 stores the values of the elements W ie constituting the weight vectors w 1 to m for each individual ID (i).
- the number of elements constituting the weight vector may be the same as the number of rectangular pulses that the replacement unit 102 generates from one event in S102.
- the replacement unit 102 replaces events in the history event series with rectangular pulses by the processing from S102 to S114, so that the state vector generation unit 106 generates a state vector. To do. Since the state vector is stationary in each period, the analysis apparatus 10 causes the analysis unit 110 to perform a Poisson regression analysis on the weight vectors w 1 to m corresponding to the m individuals, and models the response of each individual to the event. be able to.
- the analyzer 10 can approximate the tendency of the response to the stimulation of m individuals by a step function composed of a plurality of rectangular pulses.
- the analysis apparatus 10 can estimate what kind of event a certain individual i is likely to react to and a tendency such as a time lag of a response to the event by calculating the weight vector w i .
- FIG. 10 shows a flow of estimation processing of the estimation apparatus 20 of the present embodiment.
- the estimation device 20 estimates the responses of a plurality of individuals to a predetermined event sequence by executing the processing from S202 to S210.
- the replacement unit 202 replaces each event in the impulse-like event sequence for a specific individual with a plurality of rectangular pulses.
- the replacement unit 102 inputs an event sequence assumed in the simulation.
- the replacement unit 102 may create an event series for the individual i from a previously created advertisement campaign plan for a product using a predetermined algorithm, and input this.
- the replacement unit 202 generates a rectangular pulse having a predetermined time width starting from the time point of the event included in the event series of the individual i.
- the replacement unit 202 supplies the state vector generation unit 206 with information on a plurality of event sequences and start points and time widths of the plurality of rectangular pulses.
- the state vector generation unit 206 generates a state vector corresponding to the state of the plurality of rectangular pulses for each of the plurality of event sequences, corresponding to each of the plurality of periods. For example, first, the state vector generation unit 206 divides a period included in a plurality of event series at a constant time interval ⁇ t, and generates a plurality of divided periods T 1 to T n .
- State vector generator 206 generates the state vector X ij of individual i for each divided period T j.
- the state vector generation unit 206 generates a state vector X ij from a plurality of elements corresponding to the states of a plurality of rectangular pulses at the start point or end point of the period T j .
- the state vector generation unit 206 may generate the state vector X ij from the average value of the state of the rectangular pulse at the start point and end point of the period T j .
- the state vector generation unit 206 supplies the generated state vector X ij to the estimation unit 210.
- the storage unit 208 reads the weight vector w i of the individual i from the storage unit 116 of the analysis apparatus 10 and stores it.
- the storage unit 208 supplies the weight vector w i to the estimation unit 210.
- estimation unit 210 a plurality of the state vector X i1 ⁇ X in which corresponding to each of the plurality of periods T 1 ⁇ T n, the individual corresponding to a plurality of periods T 1 ⁇ T n i
- the number of reactions y ij indicating the number of reactions is estimated.
- the estimation unit 210 converts a plurality of state vectors X ij into a plurality of feature vectors ⁇ (X ij ) in the same manner as the conversion unit 112.
- the estimation unit 210 calculates a scalar score obtained from the inner product of the feature vector ⁇ (X ij ) of the individual i and the weight vector w i for each of the plurality of periods T 1 to T n .
- Estimation unit 210 from the Poisson distribution of the average number of reactions the product of the exponential function and the time intervals ⁇ t of the scalar score in the period T j, and calculates the samples y ij of the reaction times in a period T j by Monte Carlo method.
- the estimation unit 210 When ⁇ t is sufficiently small, the estimation unit 210 generates 0 or 1 as the number of reactions in the period tj.
- the estimation unit 210 estimates the number of reactions Y i of the individual i in the entire period of the event sequence by calculating the sum of the number of reactions y ij in the plurality of periods T 1 to T n .
- the estimation apparatus 20 repeats the processing from S202 to S210 until the number of reactions Y i is estimated for all individuals.
- the estimation apparatus 20 generates a state vector from an assumed event sequence, and uses an inner product of the state vector and a weight vector w i representing a tendency of a response to a stimulus for each individual i.
- the response to the event can be simulated for each i.
- the estimation apparatus 20 can predict the magnitude
- the estimation unit 210 of the estimation device 20 may obtain the probability of reaction to an individual event or the time required for the reaction instead of estimating the number of reactions to the individual event. Good.
- the analysis apparatus 10 determines the response characteristic of the individual i by the weight vector w i . For this reason, the estimation apparatus 20 can optimize a marketing measure by using the value of the weight vector w i and using the reinforcement learning technique and / or the Markov decision process technique.
- the state vector generation unit 206 divides the event sequence at regular time intervals. For this reason, according to the estimation apparatus 20, compared with the case where an event series is divided
- the analysis apparatus 10 and the estimation apparatus 20 of the present embodiment use an advertisement campaign for a product by DM and web browsing as an event series and customer purchase as a reaction to the event
- the event series and reaction of the analysis apparatus 10 and the estimation apparatus 20 May be other.
- the analysis device 10 and the estimation device 20 may use a stimulus to an individual by television, e-mail, telephone, or the like as an event.
- the analysis device 10 and the estimation device 20 may use access to a web page of a customer's product as a reaction.
- the analysis apparatus 10 and the estimation apparatus 20 may include a reaction to the event itself in the event series. For example, when the customer (individual) purchases (reacts) a certain product in response to a certain DM advertisement (event), the analysis device 10 and the estimation device 20 handle the purchase itself as an event for the next purchase. Thereby, the analysis apparatus 10 and the estimation apparatus 20 can model the case where the reaction itself by the individual causes the next reaction, such as the habit of purchasing.
- the analysis device 10 and the estimation device 20 may analyze and predict the propagation of information in the social network.
- the analysis device 10 and the estimation device 20 may divide the text posted on the short text posting website into predetermined topic categories, and post topics of different categories as events by different channels.
- the analysis device 10 and the estimation device 20 may use another post following the post as a reaction to the event. Thereby, the analysis apparatus 10 and the estimation apparatus 20 can analyze and predict the state of propagation of information in the social network.
- FIG. 11 shows an example of a hardware configuration of a computer 1900 that functions as the analysis device 10, the estimation device 20, and the like.
- a computer 1900 is connected to a CPU peripheral unit having a CPU 2000, a RAM 2020, a graphic controller 2075, and a display device 2080 that are connected to each other by a host controller 2082, and to the host controller 2082 by an input / output controller 2084.
- Input / output unit having communication interface 2030, hard disk drive 2040, and CD-ROM drive 2060, and legacy input / output unit having ROM 2010, flexible disk drive 2050, and input / output chip 2070 connected to input / output controller 2084 With.
- the host controller 2082 connects the RAM 2020 to the CPU 2000 and the graphic controller 2075 that access the RAM 2020 at a high transfer rate.
- the CPU 2000 operates based on programs stored in the ROM 2010 and the RAM 2020 and controls each unit.
- the graphic controller 2075 acquires image data generated by the CPU 2000 or the like on a frame buffer provided in the RAM 2020 and displays it on the display device 2080.
- the graphic controller 2075 may include a frame buffer for storing image data generated by the CPU 2000 or the like.
- the input / output controller 2084 connects the host controller 2082 to the communication interface 2030, the hard disk drive 2040, and the CD-ROM drive 2060, which are relatively high-speed input / output devices.
- the communication interface 2030 communicates with other devices via a network by wire or wireless.
- the communication interface functions as hardware that performs communication.
- the hard disk drive 2040 stores programs and data used by the CPU 2000 in the computer 1900.
- the CD-ROM drive 2060 reads a program or data from the CD-ROM 2095 and provides it to the hard disk drive 2040 via the RAM 2020.
- the ROM 2010, the flexible disk drive 2050, and the relatively low-speed input / output device of the input / output chip 2070 are connected to the input / output controller 2084.
- the ROM 2010 stores a boot program that the computer 1900 executes at startup and / or a program that depends on the hardware of the computer 1900.
- the flexible disk drive 2050 reads a program or data from the flexible disk 2090 and provides it to the hard disk drive 2040 via the RAM 2020.
- the input / output chip 2070 connects the flexible disk drive 2050 to the input / output controller 2084 and inputs / outputs various input / output devices via, for example, a parallel port, a serial port, a keyboard port, a mouse port, and the like. Connect to controller 2084.
- the program provided to the hard disk drive 2040 via the RAM 2020 is stored in a recording medium such as the flexible disk 2090, the CD-ROM 2095, or an IC card and provided by the user.
- the program is read from the recording medium, installed in the hard disk drive 2040 in the computer 1900 via the RAM 2020, and executed by the CPU 2000.
- a program installed in the computer 1900 and causing the computer 1900 to function as the analysis apparatus 10 includes a replacement module, a division module, a state vector generation module, a reaction frequency calculation module, an analysis module, a conversion module, and an optimization module.
- the program that includes the storage module and causes the computer 1900 to function as the estimation device 20 includes a replacement module, a state vector generation module, a storage module, and an estimation module.
- These programs or modules work on the CPU 2000 or the like to change the computer 1900 into a replacement unit 102, a division unit 104, a state vector generation unit 106, a reaction frequency calculation unit 108, an analysis unit 110, a conversion unit 112, an optimization unit 114,
- the storage unit 116, the replacement unit 202, the state vector generation unit 206, the storage unit 208, and the estimation unit 210 may function.
- the information processing described in these programs is read into the computer 1900, so that the replacement unit 102, the division unit 104, and the state vector generation, which are specific means in which the software and the various hardware resources described above cooperate.
- the specific analysis apparatus 10 and the estimation apparatus 20 according to a use purpose are constructed
- the CPU 2000 executes a communication program loaded on the RAM 2020 and executes a communication interface based on the processing content described in the communication program.
- a communication process is instructed to 2030.
- the communication interface 2030 reads transmission data stored in a transmission buffer area or the like provided on a storage device such as the RAM 2020, the hard disk drive 2040, the flexible disk 2090, or the CD-ROM 2095, and sends it to the network.
- the reception data transmitted or received from the network is written into a reception buffer area or the like provided on the storage device.
- the communication interface 2030 may transfer transmission / reception data to / from the storage device by a DMA (direct memory access) method. Instead, the CPU 2000 transfers the storage device or the communication interface 2030 as a transfer source.
- the transmission / reception data may be transferred by reading the data from the data and writing the data to the communication interface 2030 or the storage device of the transfer destination.
- the CPU 2000 is all or necessary from among files or databases stored in an external storage device such as a hard disk drive 2040, a CD-ROM drive 2060 (CD-ROM 2095), and a flexible disk drive 2050 (flexible disk 2090).
- This portion is read into the RAM 2020 by DMA transfer or the like, and various processes are performed on the data on the RAM 2020. Then, CPU 2000 writes the processed data back to the external storage device by DMA transfer or the like.
- the RAM 2020 and the external storage device are collectively referred to as a memory, a storage unit, a storage device, or the like. These may correspond to the storage unit 116 and the storage unit 208 of the present embodiment.
- the CPU 2000 can also store a part of the RAM 2020 in the cache memory and perform reading and writing on the cache memory. Even in such a form, the cache memory bears a part of the function of the RAM 2020. Therefore, in the present embodiment, the cache memory is also included in the RAM 2020, the memory, and / or the storage device unless otherwise indicated. To do.
- the CPU 2000 performs various operations, such as various operations, information processing, condition determination, information search / replacement, etc., described in the present embodiment, specified for the data read from the RAM 2020 by the instruction sequence of the program. Is written back to the RAM 2020. For example, when performing the condition determination, the CPU 2000 determines whether or not the various variables shown in the present embodiment satisfy the conditions such as large, small, above, below, equal, etc., compared to other variables or constants. If the condition is satisfied (or not satisfied), the program branches to a different instruction sequence or calls a subroutine.
- the CPU 2000 can search for information stored in a file or database in the storage device. For example, in the case where a plurality of entries in which the attribute value of the second attribute is associated with the attribute value of the first attribute are stored in the storage device, the CPU 2000 displays the plurality of entries stored in the storage device. The entry that matches the condition in which the attribute value of the first attribute is specified is retrieved, and the attribute value of the second attribute that is stored in the entry is read, thereby associating with the first attribute that satisfies the predetermined condition The attribute value of the specified second attribute can be obtained.
- the program or module shown above may be stored in an external recording medium.
- an optical recording medium such as DVD or CD
- a magneto-optical recording medium such as MO
- a tape medium such as an IC card, and the like
- a storage device such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet may be used as a recording medium, and the program may be provided to the computer 1900 via the network.
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Data Mining & Analysis (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Educational Administration (AREA)
- Operations Research (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Complex Calculations (AREA)
Abstract
Description
pen(w0,w1,w2,…wm)は、正規化項であって通常は凸な関数を用いるが非凸な関数を用いてもよい。最適化部114は、正規化項を用いることにより、重みベクトルw1~mが入力された履歴に過剰適合し、履歴と異なるイベント系列に対して適合度が低下することを防止する。
を用いることができる。なお、c0及びcは、交叉検定等によって定めるハイパーパラメータである。最適化部114は、L2正規化項に代えて、L1正規化項を用いてもよい。また、最適化部114は、過剰適合防止用の正規化項を含まない数式を用いてもよい。
Claims (20)
- インパルス状のイベント系列におけるそれぞれのイベントを複数の矩形パルスに置換する置換部と、
前記イベント系列を前記複数の矩形パルスの少なくとも1つの変化点の時刻で分割する分割部と、
分割された複数の期間のそれぞれに対応して、前記複数の矩形パルスの状態に応じた状態ベクトルを生成する状態ベクトル生成部と、
前記複数の期間に対応する複数の前記状態ベクトルを用いてポアソン回帰分析を行う分析部と、
を備える分析装置。 - 前記置換部は、複数の前記イベント系列におけるそれぞれのイベントを前記複数の矩形パルスに置換し、
前記分割部は、前記複数のイベント系列の少なくとも1つにおける前記複数の矩形パルスの少なくとも1つの変換点の時刻で分割し、
前記状態ベクトル生成部は、分割された複数の期間のそれぞれに対応して、前記複数のイベント系列のそれぞれについての前記複数の矩形パルスの状態に応じた前記状態ベクトルを生成する
請求項1に記載の分析装置。 - 前記複数の期間のそれぞれに対応して、前記イベント系列を受けて当該期間中に発生した反応の回数である反応回数を算出する反応回数算出部を更に備え、
前記分析部は、前記複数の状態ベクトル、および、前記複数の期間に対応する複数の前記反応回数を用いてポアソン回帰分析を行う
請求項2に記載の分析装置。 - 前記分析部は、
前記複数の状態ベクトルを複数の特徴ベクトルに変換する変換部と、
前記複数の期間において、前記特徴ベクトルと重みベクトルとの内積により得られるスカラースコアに基づくポアソン分布から前記反応回数が得られる確率を最大化するように重みベクトルを最適化する最適化部と、
を有する請求項3に記載の分析装置。 - 前記最適化部は、前記スカラースコアに基づいて前記反応回数の発生確率を算出する確率関数の、前記複数の期間における合計が最大化するように前記重みベクトルを最適化する請求項4に記載の分析装置。
- 前記最適化部は、前記確率関数として、単位時間あたりのイベント発生回数の平均が前記スカラースコアの指数関数となるポアソン過程において、対象の期間の間に前記反応回数のイベントが発生する確率を算出するための対数確率質量関数から前記反応回数に対する正規化項を除去した関数を用いる請求項5に記載の分析装置。
- 前記分析部は、前記複数のイベント系列の入力対象となる複数の個体のそれぞれについて前記ポアソン回帰分析を行う請求項1から7のいずれか一項に記載の分析装置。
- 前記分析部は、前記複数の期間のうち一部を間引く請求項1から8のいずれか一項に記載の分析装置。
- 前記分析部は、前記複数の期間のうち基準未満の長さの期間を間引く請求項9に記載の分析装置。
- 請求項1から10のいずれか一項に記載の分析装置による分析結果に基づいてイベント系列に対する反応を推定する推定装置であって、
入力された前記イベント系列におけるそれぞれのイベントを複数の矩形パルスに置換する置換部と、
複数の期間のそれぞれに対応して、前記複数の矩形パルスの状態に応じた状態ベクトルを生成する状態ベクトル生成部と、
前記分析装置による分析結果を用いて、前記複数の期間のそれぞれに対応する複数の前記状態ベクトルから、前記複数の期間のそれぞれに対応する反応の回数を示す反応回数を推定する推定部と、
を備える推定装置。 - コンピュータを、インパルス状のイベント系列に対するポアソン回帰分析を行う分析装置として機能させる分析プログラムであって、
前記コンピュータを、
入力された前記イベント系列におけるそれぞれのイベントを複数の矩形パルスに置換する置換部と、
前記イベント系列を前記複数の矩形パルスの少なくとも1つの変化点の時刻で分割する分割部と、
分割された複数の期間のそれぞれに対応して、前記複数の矩形パルスの状態に応じた状態ベクトルを生成する状態ベクトル生成部と、
前記複数の期間に対応する複数の前記状態ベクトルを用いてポアソン回帰分析を行う分析部と、
して機能させる分析プログラム。 - コンピュータを、請求項12に記載の分析プログラムによる分析結果に基づいてイベント系列に対する反応を推定する推定装置として機能させる推定プログラムであって、
前記コンピュータを、
入力された前記イベント系列におけるそれぞれのイベントを複数の矩形パルスに置換する置換部と、
複数の期間のそれぞれに対応して、前記複数の矩形パルスの状態に応じた状態ベクトルを生成する状態ベクトル生成部と、
前記分析装置による分析結果を用いて、前記複数の期間のそれぞれに対応する複数の前記状態ベクトルから、前記複数の期間のそれぞれに対応する反応の回数を示す反応回数を推定する推定部と、
して機能させる推定プログラム。 - コンピュータが、インパルス状のイベント系列におけるそれぞれのイベントを複数の矩形パルスに置換する段階と、
前記コンピュータが、前記イベント系列を前記複数の矩形パルスの少なくとも1つの変化点の時刻で分割する段階と、
前記コンピュータが、分割された複数の期間のそれぞれに対応して、前記複数の矩形パルスの状態に応じた状態ベクトルを生成する段階と、
前記コンピュータが、前記複数の期間に対応する複数の前記状態ベクトルを用いてポアソン回帰分析を行う段階とを備える、
分析方法。 - 前記置換する段階において、複数の前記イベント系列におけるそれぞれのイベントを前記複数の矩形パルスに置換し、
前記分割する段階において、前記複数のイベント系列の少なくとも1つにおける前記複数の矩形パルスの少なくとも1つの変換点の時刻で分割し、
前記生成する段階において、分割された複数の期間のそれぞれに対応して、前記複数のイベント系列のそれぞれについての前記複数の矩形パルスの状態に応じた前記状態ベクトルを生成する、
請求項14に記載の分析方法。 - 前記複数の期間のそれぞれに対応して、前記イベント系列を受けて当該期間中に発生した反応の回数である反応回数を算出する反応回数算出段階を更に備え、
前記ポアソン回帰分析を行う段階において、前記複数の状態ベクトル、および、前記複数の期間に対応する複数の前記反応回数を用いてポアソン回帰分析を行う、
請求項15に記載の分析方法。 - 前記ポアソン回帰分析を行う段階は、
前記複数の状態ベクトルを複数の特徴ベクトルに変換する変換段階と、
前記複数の期間において、前記特徴ベクトルと重みベクトルとの内積により得られるスカラースコアに基づくポアソン分布から前記反応回数が得られる確率を最大化するように重みベクトルを最適化する最適化段階とを含む、
請求項16に記載の分析方法。 - 前記ポアソン回帰分析を行う段階において、複数の前記イベント系列の入力対象となる複数の個体のそれぞれについて前記ポアソン回帰分析を行う、
請求項14から17のいずれか一項に記載の分析方法。 - 前記ポアソン回帰分析を行う段階において、前記複数の期間のうち一部を間引く請求項14から18のいずれか一項に記載の分析方法。
- 請求項14に記載の分析方法による分析結果に基づいてイベント系列に対する反応を推定する推定方法であって、
コンピュータが、入力された前記イベント系列におけるそれぞれのイベントを複数の矩形パルスに置換し、
前記コンピュータが、複数の期間のそれぞれに対応して、前記複数の矩形パルスの状態に応じた状態ベクトルを生成し、
前記コンピュータが、前記分析方法による分析結果を用いて、前記複数の期間のそれぞれに対応する複数の前記状態ベクトルから、前記複数の期間のそれぞれに対応する反応の回数を示す反応回数を推定する
推定方法。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2014554237A JP6072078B2 (ja) | 2012-12-25 | 2013-11-19 | 分析装置、分析プログラム、分析方法、推定装置、推定プログラム、及び、推定方法。 |
US14/655,635 US10121156B2 (en) | 2012-12-25 | 2013-11-19 | Analysis device, analysis program, analysis method, estimation device, estimation program, and estimation method |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2012-281789 | 2012-12-25 | ||
JP2012281789 | 2012-12-25 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2014103560A1 true WO2014103560A1 (ja) | 2014-07-03 |
Family
ID=51020654
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2013/081118 WO2014103560A1 (ja) | 2012-12-25 | 2013-11-19 | 分析装置、分析プログラム、分析方法、推定装置、推定プログラム、及び、推定方法。 |
Country Status (3)
Country | Link |
---|---|
US (1) | US10121156B2 (ja) |
JP (1) | JP6072078B2 (ja) |
WO (1) | WO2014103560A1 (ja) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020026326A1 (ja) * | 2018-07-31 | 2020-02-06 | 日本電気株式会社 | 情報処理装置、制御方法、及びプログラム |
WO2020026327A1 (ja) * | 2018-07-31 | 2020-02-06 | 日本電気株式会社 | 情報処理装置、制御方法、及びプログラム |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
NZ591354A (en) | 2010-02-26 | 2012-09-28 | Kraft Foods Global Brands Llc | A low-tack, UV-cured pressure sensitive acrylic ester based adhesive for reclosable packaging |
CN104699717B (zh) * | 2013-12-10 | 2019-01-18 | 中国银联股份有限公司 | 数据挖掘方法 |
US11403531B2 (en) * | 2017-07-19 | 2022-08-02 | Disney Enterprises, Inc. | Factorized variational autoencoders |
US11423325B2 (en) | 2017-10-25 | 2022-08-23 | International Business Machines Corporation | Regression for metric dataset |
CN112579920B (zh) * | 2020-12-09 | 2023-06-20 | 成都中科大旗软件股份有限公司 | 一种基于事件实现跨时空传播分析方法 |
WO2023054783A1 (ko) * | 2021-09-29 | 2023-04-06 | 주식회사 프렌트립 | 파라미터 간의 연관도를 기반으로 레저 활동 레슨에 대한 파라미터를 예측하는 전자 장치의 제어 방법 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2005236862A (ja) * | 2004-02-23 | 2005-09-02 | Kddi Corp | ログ分析装置、ログ分析プログラムおよび記録媒体 |
JP2009229899A (ja) * | 2008-03-24 | 2009-10-08 | Toshiba Corp | 音声認識装置および音声認識方法 |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1630866A (zh) | 1999-01-27 | 2005-06-22 | 理查德桑德斯国际公司 | 用于模拟人类对刺激的反应的方法 |
US6430539B1 (en) | 1999-05-06 | 2002-08-06 | Hnc Software | Predictive modeling of consumer financial behavior |
US6904408B1 (en) * | 2000-10-19 | 2005-06-07 | Mccarthy John | Bionet method, system and personalized web content manager responsive to browser viewers' psychological preferences, behavioral responses and physiological stress indicators |
JP2002288295A (ja) | 2001-03-27 | 2002-10-04 | Digicube Co Ltd | 流通在庫シミュレーション装置、記録媒体、及びプログラム |
US20080290987A1 (en) * | 2007-04-22 | 2008-11-27 | Lehmann Li | Methods and apparatus related to content sharing between devices |
US7519553B2 (en) * | 2007-05-02 | 2009-04-14 | International Business Machines Corporation | Method and system for debt collection optimization |
JP5170537B2 (ja) | 2008-03-13 | 2013-03-27 | 富士通株式会社 | 利用者状態推定システム、利用者状態推定方法および利用者状態推定プログラム |
US20100049538A1 (en) * | 2008-08-22 | 2010-02-25 | Durban Frazer | Method and apparatus for selecting next action |
US8543445B2 (en) * | 2009-12-21 | 2013-09-24 | Hartford Fire Insurance Company | System and method for direct mailing insurance solicitations utilizing hierarchical bayesian inference for prospect selection |
JP5423512B2 (ja) | 2010-03-18 | 2014-02-19 | 日本電気株式会社 | 放送型データストリーム保護システム、方法及びユーザ端末特定プログラム |
US9047560B2 (en) * | 2011-06-29 | 2015-06-02 | Microsoft Technology Licensing, Llc | Using event stream data to create a decision graph representing a race participant where leaf nodes comprise rates defining a speed of the race participant in a race simulation |
-
2013
- 2013-11-19 US US14/655,635 patent/US10121156B2/en active Active
- 2013-11-19 WO PCT/JP2013/081118 patent/WO2014103560A1/ja active Application Filing
- 2013-11-19 JP JP2014554237A patent/JP6072078B2/ja active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2005236862A (ja) * | 2004-02-23 | 2005-09-02 | Kddi Corp | ログ分析装置、ログ分析プログラムおよび記録媒体 |
JP2009229899A (ja) * | 2008-03-24 | 2009-10-08 | Toshiba Corp | 音声認識装置および音声認識方法 |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020026326A1 (ja) * | 2018-07-31 | 2020-02-06 | 日本電気株式会社 | 情報処理装置、制御方法、及びプログラム |
WO2020026327A1 (ja) * | 2018-07-31 | 2020-02-06 | 日本電気株式会社 | 情報処理装置、制御方法、及びプログラム |
JPWO2020026326A1 (ja) * | 2018-07-31 | 2021-08-02 | 日本電気株式会社 | 情報処理装置、制御方法、及びプログラム |
JPWO2020026327A1 (ja) * | 2018-07-31 | 2021-08-02 | 日本電気株式会社 | 情報処理装置、制御方法、及びプログラム |
US20210311009A1 (en) * | 2018-07-31 | 2021-10-07 | Nec Corporation | Information processing apparatus, control method, and non-transitory storage medium |
JP7074194B2 (ja) | 2018-07-31 | 2022-05-24 | 日本電気株式会社 | 情報処理装置、制御方法、及びプログラム |
JP7140191B2 (ja) | 2018-07-31 | 2022-09-21 | 日本電気株式会社 | 情報処理装置、制御方法、及びプログラム |
Also Published As
Publication number | Publication date |
---|---|
US20150339680A1 (en) | 2015-11-26 |
JP6072078B2 (ja) | 2017-02-01 |
JPWO2014103560A1 (ja) | 2017-01-12 |
US10121156B2 (en) | 2018-11-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6072078B2 (ja) | 分析装置、分析プログラム、分析方法、推定装置、推定プログラム、及び、推定方法。 | |
CN105631698B (zh) | 用于策略部署的风险量化 | |
US8190537B1 (en) | Feature selection for large scale models | |
Keane et al. | A computationally practical simulation estimation algorithm for dynamic panel data models with unobserved endogenous state variables | |
US20090024407A1 (en) | Indicating which of forecasting models at different aggregation levels has a better forecast quality | |
US10515378B2 (en) | Extracting relevant features from electronic marketing data for training analytical models | |
CN106030589A (zh) | 使用开源数据的疾病预测系统 | |
JP2008152321A (ja) | 顧客セグメント推定装置 | |
WO2008042784A2 (en) | Comparing taxonomies | |
JP5963320B2 (ja) | 情報処理装置、情報処理方法、及び、プログラム | |
JP5984147B2 (ja) | 情報処理装置、情報処理方法、及び、プログラム | |
US20190080352A1 (en) | Segment Extension Based on Lookalike Selection | |
Zellner et al. | Bayesian and non-Bayesian analysis of the seemingly unrelated regression model with Student-t errors, and its application for forecasting | |
Djeundje et al. | Identifying hidden patterns in credit risk survival data using generalised additive models | |
CN110543474A (zh) | 一种基于全埋点与潜在因子模型的用户行为分析方法和装置 | |
EP3918545A1 (en) | Method and system for optimizing an objective having discrete constraints | |
JP2015114988A (ja) | 処理装置、処理方法、およびプログラム | |
US20170046726A1 (en) | Information processing device, information processing method, and program | |
JP2015187773A (ja) | データ解析装置、データ解析プログラム及びデータ解析方法 | |
JP5860828B2 (ja) | 行動確率推定装置、方法、及びプログラム | |
JP2020071523A (ja) | 見積方法、課金方法、コンピュータ、及びプログラム | |
JP6784000B2 (ja) | 情報処理装置、情報処理方法、及び、プログラム | |
JP6570501B2 (ja) | 情報提供装置、情報提供方法、および情報提供プログラム | |
WO2023021690A1 (ja) | モデル生成装置、モデル生成方法、プログラム | |
JP7392838B2 (ja) | 推定装置、推定システム、推定方法及びプログラム |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 13868413 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2014554237 Country of ref document: JP Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 14655635 Country of ref document: US |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 13868413 Country of ref document: EP Kind code of ref document: A1 |