KR101680241B1 - Method for predicting human personality based on data regarding human position having specific data type - Google Patents
Method for predicting human personality based on data regarding human position having specific data type Download PDFInfo
- Publication number
- KR101680241B1 KR101680241B1 KR1020150089051A KR20150089051A KR101680241B1 KR 101680241 B1 KR101680241 B1 KR 101680241B1 KR 1020150089051 A KR1020150089051 A KR 1020150089051A KR 20150089051 A KR20150089051 A KR 20150089051A KR 101680241 B1 KR101680241 B1 KR 101680241B1
- Authority
- KR
- South Korea
- Prior art keywords
- information
- data
- neural network
- network model
- value
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G06N99/005—
Abstract
Discloses a toughness evaluation method for evaluating toughness using an artificial neural network model that has been learned using learning data that includes input layer information about time remaining at a plurality of positions and output layer information indicating information about toughness values. The input layer information includes N * M pieces of data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes , The value of the data E p, q is set such that the person was at the position p at the time point indicated by the time index q by the particular person for any p, q (p is a natural number from 1 to N and q is a natural number from 1 to M) And a binary number that is determined depending on whether or not it is a binary number.
Description
BACKGROUND OF THE
An example of a method of extracting personality data using position data is disclosed in the 40th Korea Information Processing Society Abstracts, Vol. 20, No. 2 (2013.11). This method relates to a technique of outputting data concerning the personality of the person in the output layer when a value related to the probability that a person exists in each of the specific places is input to the input layer of the artificial neural network.
The research of this paper is meaningful in that it is the initial study on the method of extracting personality data of human using position data through artificial neural network. However, the representation method of the data provided to the input layer of the artificial neural network is relatively complex in that the expression is provided as a real number having a value of 0 to 1. Therefore, learning of artificial neural network is inefficient and implementation is relatively difficult.
In the present invention, a technique for extracting personality using a Deep Belief Network is provided.
According to the results of the conventional psychological research, it is said that the human movement pattern is influenced by the personality of each person. For example, an extroverted person would prefer going out rather than an introverted person. Conversely, if the positional data statistical value of a specific person is compared with the average statistical value of positional data of a general person, if the outgoing frequency of the specific person is determined to be high, the person can be determined to have an extrovert nature.
Data on human location have been very interesting in many related fields and due to the help of big data technology and data mining techniques due to portable position sensing devices and smart phones, .
Although there are a lot of psychological expressions representing human personality, in one embodiment of the present invention, BFI (Big Five Inventory), which is most commonly used about personality expression method, can be used.
The personality of a particular person and the preferred location of the particular person can be correlated at a common sense level. In the present invention, an algorithm for simulating the effect of human position on toughness can be designed and provided.
In one aspect of the present invention, a configuration focused on toughness is provided. To this end, a methodology for establishing a human movement model can be used. This methodology is described in Hyunuk Kim and Ha Yoon Song, "Formulating Human Mobility Model in a Form of Continuous Time Markov Chain", Procedia CS, Vol. 10, pp. 379-396, 2012.C. In the present invention, a process of predicting toughness using a human position can be developed and provided.
It is clear that the human movement pattern is affected by toughness, and this has already been studied. For example, a person with an extraversion may like outdoor activity more than a person with an introversion.
In one technique according to one aspect of the present invention, among many toughness models, it introduces a Big Five toughness characteristic which is a generally well accepted model for toughness. In the technique according to one aspect of the present invention, an artificial neural network such as DBN can be used as a tool for mutually combining human mobility and toughness.
In psychology, the big-five factor for personality represents the five domains of human personality and is used to describe human personality. In the initial study of human personality, 16 factor theories were published. The results of the factor analysis suggest that 5-factor models with 5 factors are more appropriate. Other studies have confirmed the validity of the above 5-factor model. The Big Five framework on personality traits has been presented as a robust model for understanding the relationship between personality and various behaviors.
The big-five factor is the five basic big-five personality factors (OCEAN), such as openness to experience, conscientiousness, extroversion, agreeableness and neuroticism. It can be divided. Of these, openness to experience can be measured by a conflicting measure of [inventive / curious] and [consistent / cautious]. Art, emotion, adventure, unconventional ideas, curiosity, and a variety of experiences. Openness to experience reflects the degree of intellectual curiosity, creativity, and preferences for new and diverse things. The word "openness to experience" has been replaced by the term intellect. Sincerity can be measured by an opposing measure of [efficient / organized] and [easy-going / careless]. Self-control, obligation behavior, and attainment goals, planned behavior, organizational, and dependency, rather than impulsive behavior. Extroversion can be measured by an opposing measure of [outgoing / energetic] and [solitary / reserved]. Activity, positive emotions, surrogacy, assertiveness, sociality, and the tendency to pursue stimulation within other people, and talkativeness. Affinity can be measured by a conflicting measure of [friendly / compassionate] and [cold / unkind]. It relates to a sympathetic mind and a tendency to be cooperative rather than to be suspicious or hostile to others. Neuroticism can be measured by a conflicting measure of [sensitive / nervous] and [secure / confident]. It is associated with a tendency to experience uncomfortable feelings such as anger, anxiety, depression, and vulnerability. Neuroticism refers to the degree of emotional stability and impulse control, and may be referred to as emotional stability. The Big Five Inventory (BFI) was developed to identify the traces of people by dividing the five factor into 44 questionnaires. Each questionnaire has a 5-point scale. The toughness feature can be specified by using the Big Five Inventory (BFI), the most frequently used self-report inventory, and the five factor score from the BFI is a function of the major output of the algorithm presented in the present invention As an example.
In one embodiment of the present invention, human characteristics can be expressed numerically using these five factors. In one embodiment, this value is closer to 0 as the personality score is lower, and may be closer to 5 as the personality score is lower.
Artificial neural networks (briefly neural networks or neural networks) were first presented in 1943 by W. McCulloch and W. Hartley. If the sum of the stimulus input to the neuron is larger than the threshold value, the neuron is activated but if it is smaller than the threshold value, the neuron is not activated.
The artificial neural network may include an output layer, a hidden layer, and an input layer. Each layer may be composed of any number of nodes.
Here, the 'node' can be defined as an abstract concept object that can change its value through a special procedure with a specific value, and can be linked with another node by a link. The input layer is a set of one or more nodes having a specific value assigned by the user and the output layer is a set of one or more nodes having a result of the procedure according to a specific procedure set by the user, Hidden layer " may refer to a set of one or more nodes storing intermediate results and temporal values that temporarily appear when a user performs a predetermined procedure.
There may be links between the nodes of the 'input layer' and the nodes of the 'hidden layer', and between the nodes of the 'hidden layer' and the nodes of the 'output layer' It may have a specific weight to be given.
The toughness evaluation method provided in accordance with an aspect of the present invention is a toughness evaluation method using an artificial neural network model that is learned using learning data including input layer-information and output layer- Is a toughness evaluation method for evaluating a toughness. The artificial neural network model is learned by providing the input layer-information to the input layer of the artificial neural network model and providing the output layer-information to the output layer of the artificial neural network model. The toughness evaluation method includes inputting input layer information on a time when a specific person stays at a plurality of positions to the input layer of the artificial neural network model to acquire toughness value from the output layer of the artificial neural network model . The input layer information includes N * M pieces of data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes. , The value of the data E p, q is set such that the person is at the position p at the time point indicated by the time index q by the particular person , with respect to any p, q (p is a natural number from 1 to N and q is a natural number from 1 to M) And has a binary number that is determined depending on whether or not it has been.
Here, the toughness evaluation method may further include a step of determining toughness of the specific person by comparing the toughness value of the obtained specific person with the average toughness value of the plurality of people.
A computing apparatus provided according to another aspect of the present invention is capable of executing the toughness evaluation method.
A computer-readable medium provided according to another aspect of the present invention records a program for causing the computing device to execute the toughness evaluation method.
According to still another aspect of the present invention, there is provided a toughness evaluation model generation method comprising: learning an artificial neural network model using learning data including input layer-information about an amount of time spent in a plurality of positions and output layer- To generate a toughness evaluation model. Providing the input layer information to the input layer of the artificial neural network model and providing a set of the output layer information corresponding to the input layer information to the output layer of the artificial neural network model, . The input layer information includes N * M pieces of data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes. , The value of the data E p, q is set such that the person is at the position p at the time point indicated by the time index q by the particular person , with respect to any p, q (p is a natural number from 1 to N and q is a natural number from 1 to M) And has a binary number that is determined depending on whether or not it has been. Here, the binary number may mean '1' or '0'.
A computing apparatus provided according to another aspect of the present invention is capable of executing the toughness evaluation model generation method.
A computer-readable medium provided according to another aspect of the present invention records a program that causes the computing device to execute the toughness evaluation model generation method.
According to the present invention, it is possible to provide a technique of extracting personality using a Deep Belief Network.
FIG. 1 shows an example of the result of prediction of the positional data change according to the BFI value.
FIG. 2 illustrates learning data for learning an artificial neural network according to a first comparative example of the present invention.
FIG. 3 shows personality data values output according to position data input to the artificial neural network, using the artificial neural network learned using the learning data of FIG.
FIG. 4 shows the difference from the average personality number (i.e., the first row of FIG. 2) for the personality data results from FIG.
FIG. 5 shows a graph in which the contents of FIG. 4 are easily changed.
FIG. 6 shows a general topology of an artificial neural network to which learning data shown in FIG. 2 is applied.
FIG. 7 shows an example of a learning model that is learned using the learning data shown in FIG.
Fig. 8 shows a model learning method according to the first comparative example of the present invention.
FIG. 9 is a schematic representation of a dictionary information collection process according to an embodiment of the present invention.
10 is a diagram for explaining an example of using a network model that reflects a characteristic value of a specific person and outputs a probability value that the specific person is located at a specific place.
FIG. 11 shows an example of a network model that is learned and provided using learning data, which can be used to provide a prediction method according to the second mode of the second comparative example.
12 shows an example of a graph in which a plurality of positions defined by predetermined criteria are associated with acquired time, according to an embodiment of the present invention.
13 shows an example of data formed by modifying collected geolocation data and data on toughness values according to an embodiment of the present invention.
14 shows an example of a table comparing characteristics of training data with characteristics of a test result using an artificial neural network model trained according to an embodiment of the present invention.
15A shows an example of an artificial neural network model that can be used to perform the toughness evaluation method according to an embodiment of the present invention.
FIG. 15B is a more detailed representation of the input layer of the artificial neural network model shown in FIG. 15A.
Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be understood, however, that the invention is not limited to the disclosed embodiments, but may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, Is provided to fully inform the user.
≪ Comparative Example 1 of the present invention &
Although BFIs are likely to be transformed into numerical data, it is difficult to actually compare human BFI personality data with moving data. Therefore, in the first comparative example of the present invention, it is assumed that the position data weight value changes according to each BFI element of the human, and the position data change aspect can be applied when each BFI changes.
Each element of personality data BFI satisfies orthogonality with each other. Orthogonality is used to describe the relationship between two or more variables, which can be expressed as 'orthogonal' when there is no relation between the variables. In a similar sense, it can be described as 'independence'. Since the influence of each personality element on the position data also has to satisfy the orthogonality, the personality factor affects the position data also satisfies the orthogonality.
FIG. 1 is a diagram illustrating a change in position data according to each BFI element so as to satisfy orthogonality. In the prior art, since the biased relationship between the personality data and the position data can not be clearly known, the table shown in Fig. 1 assumes the correlation between the personality and the position relationship data. If the score of each personality factor is different from the 'average personality score' of the corresponding factor, the position data changes according to the degree of difference. Here, the 'average personality score' may mean the average of personality scores of a plurality of randomly extracted persons. For example, the mean value of 100 neurogenic values arbitrarily extracted from a stable society may be lower than the 100 neural values arbitrarily extracted in a society undergoing war.
For example, if a person's neuroticism score is higher than the average personality score for neuroticism, it affects the probability of being placed in the indoor activity 'home', reflecting the person who dislikes going out, We can reduce the weights of the activities 'San' and 'Church'.
The row indicated by 'openness' in FIG. 1 indicates that the probability that a particular person having an openness higher than the average personality score on openness is located in 'home', 'school', 'church', or 'mountain' Indicates a higher or lower probability that a person who is not a person is located in a 'home', 'school', 'church', or 'mountain'. '+' Indicates that a certain person with higher openness than the average has a higher probability of being located in the place indicated by the column than the person who does not, and '-' means the opposite. In Figure 1, a person with a higher openness than the average is assumed to have a higher probability of being located in a school or mountain than a person who is not.
To illustrate another example in FIG. 1, a row represented by 'Neutoricism' indicates that a particular person having neuroticism higher than the average personality score of neurosis is a 'house', 'school', 'church' The probability of being located on the 'mountain' is higher or lower than the probability that the person who is not in 'house', 'school', 'church', or 'mountain' is located. In FIG. 1, a certain person having a higher level of neuroticism than the average has a higher probability of being located at home than a person who is not. And it is assumed that a certain person with higher neuroticism than the above average has a lower probability of being located in the church or mountain than the other.
In FIG. 1, the relationship of the position-time data according to the BFI value is inferred from a psychological point of view. However, the relationship of the BFI value according to the individual position-time data pattern can be presented according to the present invention algorithm have.
[Experiment]
Various examples of analyzing personality data result values according to position data can be presented.
First, we can see the position data and the personality data as cause and result, and the relation between them as X, Y according to the function definition. However, there are drawbacks to this method, and it is necessary to understand the complex connection causation in order to understand the relationship between X and Y, ie, position and personality.
In addition to analyzing by function, we can think about solving with pattern analysis algorithm. The pattern analysis algorithm is an algorithm for analyzing a specific pattern according to learned data for a certain phenomenon. For example, there is a neural network (ie, neural network or artificial neural network) that imitates a human brain. The pattern analysis algorithm derives a probabilistic result according to the learned information without having to grasp the relation between X and Y, unlike the method using the function. In the first comparative example of the present invention, by using the features of the pattern analysis algorithm, the relationship between the position data and the personality data is grasped by using the pattern analysis algorithm.
In the first comparative example of the present invention, an artificial neural network algorithm is used and a total of four input nodes and five output nodes are used. The previous input nodes receive the position data values, and the following five values are the values to be input to the output node of the sample data, and the personality data corresponding to the personality change is received as the output value. Then, if four position data are input to the input node for the learned neural network, five personality data corresponding to the input data will be output from the output node.
FIG. 2 illustrates learning data for learning an artificial neural network according to a first comparative example of the present invention. Hereinafter, the process and logic for deriving the learning data of FIG. 2 will be described.
The first row of Figure 2 shows the values actually obtained from a plurality of people. The second to eleventh rows of FIG. 2 are obtained by transforming the data presented in the first row by predetermined logic, based on the position-to-tangency relationship of FIG. 1, Results are shown.
In the table of FIG. 2, each row has a combination of BFI values having a certain characteristic and a probability value to be located at each position (house, school, church, mountain). The BFI displayed in each row has a characteristic of high and low for each personality element. The input personality data value is based on the tendency distribution diagram of the personality provided by BFI. The tendency of personality follows the standard distribution.
The first row represents the personality value of the mean. If personality has the characteristics of 0.64, 0.62, 0.669, 0.74, 0.568 in order from extrinsic to extrinsic, it means that personality has an average character over all elements.
The second through eleventh rows represent the top 10% and bottom 10% of the standard distribution of each BFI value factor.
The second row shows the average sample data of people with high openness of the top 10%. If a person has a high level of openness at the top 10% level, then the openness value for this person will be 0.74 and the other values will remain unchanged. Because of the orthogonality of the five factors, the combination of arbitrary five factors is allowed without sacrificing the generality of the BFF.
The third row represents the average sample data of people with low openness of the bottom 10%. If a person has a low openness of the lower 10% level, then the openness value for this person will be 0.54 and the other values will remain unchanged.
The fourth row represents the average sample data of the people with the highest 10% higher extroversion. If a person has a high extroversion of the top 10% level, then the extroversion value for this person would be 0.718, and the other values would remain unchanged.
The fifth row represents the average sample data of people with low outwardness of the bottom 10%. If a person has a low outwardness of the lower 10% level, the extroversion value for this person will be 0.515, and the other values remain unchanged.
The sixth row represents the average sample data of the people with high integrity of the top 10%. If a person has a high integrity level of the top 10%, then the integrity value for this person will be 0.76 and other values will remain unchanged.
The seventh row represents the average sample data for people with low integrity of the bottom 10%. If a person has low sincerity at the lower 10% level, the sincerity value for this person will be 0.56, and the other values will remain unchanged.
The eighth row represents the average sample data of people with high affinity of the top 10%. If a person has a high affinity at the top 10% level, the affinity value for this person will be 0.84, and the other values will remain unchanged.
The ninth row represents the average sample data of people with low affinity for the bottom 10%. If a person has a low affinity at the lower 10% level, the affinity value for this person will be 0.64, and the other values will remain unchanged.
The 10th row represents the average sample data of people with high neurotic disorder of the top 10%. If a person has high neurotics at the top 10% level, the neurotic value for this person will be 0.674 and the other values will remain unchanged.
The eleventh row represents the average sample data of people with low 10% lower neurosis. If a person has low neurotics at the lower 10% level, the neurotic value for this person is 0.46 and the other values remain unchanged.
For example, the fourth row shows a high extroversion, which has a higher value when compared to the first row, which has an extrinsic value of 0.718 indicating an averaging characteristic. If the personality value of a person has an extroversion value of 0.718, it indicates that it has an upper 10% extroversion.
In the case of the location data to be set in the node of the input layer of the artificial neural network, it is highly likely that the location data are close to 1, and the nearer the value is, the less likely it exists at the location. For example, if the position data value of 'home' in FIG. 2 is 0.6 and the position data value corresponding to 'school' is 0.75, it indicates that the person is likely to be in 'school'. This value indicates a probability value or a weight value as a probability value, but it does not necessarily indicate the probability value itself.
Each position data is based on the data of the first row of FIG. 2 and is calculated according to each personality characteristic shown in FIG. 2 according to the tendency (+, -) to reflect the influence of the position data according to each personality assumed in FIG. The position data value was changed. This process of change is based on the predictions of the correlation between personality and location according to the results of psychological research.
If the specific location data value is positive in the personality, the location data is designed to have a weight value of 1.5 times the location data when the average personality is the reference. Here, the weight value of 1.5 may be changed to another value greater than 1.0.
Conversely, if they have a negative effect, they will have a value of 0.5 times the average personality data. Here, the weight value of 0.5 may be changed to another value smaller than 1.0.
For example, according to FIG. 1, extroversion affects the church negatively and positively affects san. In FIG. 2, the fourth row shows high extroversion. In the data of the first row showing ordinary characteristics, 'church' has a negative effect and gives a value of 0.15 which is 50% And it is 0.45.
The artificial neural network can be learned by using the learning data shown in FIG. Then, an arbitrary position value is inputted to the input node of the artificial neural network which has completed the learning, and the personality data learning result according to the position data can be obtained.
FIG. 3 shows personality data values analyzed according to input position data for an artificial neural network learned using the learning data of FIG. In this case, the input data values are the same as the sample data input data shown in FIG. When the actual position data other than the position data of FIG. 2 is input to the learned artificial neural network, the artificial neural network can be verified better.
Analysis of the data in FIG. 3 reveals which elements of the characteristics are largely reflected in the output value compared with the average characteristics in FIG. For example, in the data of the fourth row in FIG. 3, the personality scores are (0.651, 0.704, 0.69, 0.719, 0.584) and the average personality scores in FIG. 2 are (0.64, 0.62, 0.669, 0.74, 0.568). Comparing these two personality scores shows that the score for the second element (extroversion) has increased significantly. The second personality factor can be deduced to be a person with extroversion if it has an extrinsic location weight of (0.6, 0.5, 0.45, 0.15) in order of home, school, church, and mountain.
FIG. 4 shows the difference between the personality data result value shown in FIG. 3 and the average personality value. The smaller the absolute numerical value of a particular personality element, the more the personality element is similar to the average personality. On the contrary, the larger the value of the personality, the more the value of the personality is different from the average personality. If both sleep, the personality factor indicates a tendency to be higher than the average personality, and if it is negative, the personality tendency is low.
Since the value of the position data input in FIG. 3 considers only one influence for each personality element, the values in FIG. 4 should have a high absolute value one by one, except for the first row. In the case of the first row, it corresponds to the characteristic of the average of FIG. 2, and the difference value is 0.002, which can be estimated to be within the error value.
The personality data value has a standard deviation value of about 0.07, so that if there is an error close to 0.07, the personality tendency is different from the average. For example, in the sixth row of FIG. 4, the extroversion value is as high as 0.084. Therefore, it can be seen that the sixth row shows high extroversion characteristics. In addition, the affinity is a small value of -0.021, which is within the error value.
FIG. 5 shows a graph in which the contents of FIG. 4 are easily changed.
The components on the X axis indicate the personality BFI, and the values of the BFI when the Y axis component is the average nature and the difference value are shown. For example, in the case of 'high openness', the black stick is high, indicating openness, which is higher than average personality.
FIG. 6 shows a general topology of an artificial neural network to which learning data shown in FIG. 2 is applied.
When the learning input layer data (house, school, church, mountain data in Fig. 2) prepared in advance in the input layer of the artificial neural network is applied, data having a value different from the learning output layer data (BFI data in Fig. 2) do. This is because the weight of the artificial neural network may not have a perfectly learned value. The difference value between the output layer data for learning that is prepared in advance and the output data output from the output layer of the artificial neural network is regarded as the error value of the output layer. Specific methods for learning artificial neural networks are well known in the prior art literature.
[Learning model according to the first comparative example of the present invention]
FIG. 7 shows an example of a learning model that is learned using the learning data shown in FIG.
This learning model is composed of four
[Model learning method according to the first comparative example of the present invention]
Fig. 8 shows a model learning method according to the first comparative example of the present invention.
First, a learning model may be provided that includes an input layer, a hidden layer, and an output layer, each of which includes one or more nodes and are weighted by each other.
Then, in step S1, input layer learning data composed of values related to the probability of being located at a plurality of predetermined positions may be input to the input layer to obtain an output value from the output layer.
Then, in step S2, the output value obtained above can be compared with output layer learning data paired with the input layer learning data to calculate an error value. At this time, the output layer learning data may be a score value for the five toughness factors shown in FIG. 2, for example.
Then, in step S3, the weight included in the learning model can be updated so as to minimize the error value. This minimization scheme can be implemented using a known mathematical algorithm. By repeating these steps, the accuracy of the learning model can be improved.
[Position prediction computing device for implementing the method according to the first comparative example]
The location prediction computing apparatus for implementing the method according to the first comparative example may include a data acquisition unit and a processing unit.
The processing unit may be configured to predict the toughness of the particular person using a learning model that includes an input layer, a hidden layer, and an output layer that are each made up of nodes associated with each other by weighted links. The processing unit may be configured such that (1) a value related to the toughness of the specific person is outputted from the output layer when a value related to the probability that a specific person exists in each of a plurality of positions is input to the input layer.
The data obtaining unit may obtain information on a value related to a probability that the specific person exists in each of a plurality of locations and provide the information to the processing unit.
≪ Comparative Example 2 of the present invention &
The second comparative example of the present invention is an example modified from the first comparative example. The second comparative example further includes a method of commercially using the result of acquiring toughness data from the position-time data.
Companies that sell or service consumer goods want to obtain customer information about their prospects and / or actual customers (hereafter, customers). The business of providing such customer information to the above companies is emerging as a new industry. In particular, if companies can know personality information about the personality of a customer, it can be utilized for individual marketing that responds to each customer in a customized way. However, if customers do not provide information about their personality on their own, companies will not know this information. Therefore, each company can perform the above-mentioned individual marketing only if the customers can predict the information about the personality of each customer, even if the customer does not provide information about his / her personality.
For example, some fast food outlets offer a special breakfast menu for the worker in the morning, and not all workers will visit this store to get a breakfast menu. These stores may want to obtain information about the nature of the customers who visit them, and it would be very useful if there is another way to obtain such personality information because it is difficult to increase the participation rate in the questionnaire process for obtaining such personality information .
In addition, if a customer having a specific characteristic can know the possibility of coming near a customer's store operated by a specific company at a specific time, it may be possible to carry out a marketing method of operating the customer's store at different times by using such information .
In the second comparative example, well-known clustering techniques and network models are used to provide information on the toughness described above. These related arts will be described in detail with reference to the following description of the invention.
In the second comparative example, the positional information of the specific person is used to infer the toughness of the specific person who does not provide the information about his / her toughness by himself / herself. In order to do so, the specific person must provide his / her location information. In recent years, there has been a tendency to agree to provide his / her location in order to utilize the location information providing service provided by the mobile user equipment. .
Clms
When the network model in which the above-described learning is completed is provided, information on the personality of the persons other than the learning data providers who have agreed to provide their own location information can be obtained.
In the second comparative example, a 'prediction method' for predicting human toughness or position can be provided.
It is necessary to collect the advance information necessary for executing the prediction method according to the second comparative example. The process of collecting such prior information can be referred to as a 'pre-information gathering process'.
There are two modes, i.e., a first mode and a second mode, in the prediction method using a set of dictionary information collected through the dictionary information collection process.
Hereinafter, the 'pre-information gathering process', the first mode, and the second mode will be described.
[Prior information gathering process]
FIG. 9 is a schematic representation of a dictionary information collection process according to an embodiment of the present invention.
In the dictionary information collection process, information about a plurality of persons, for example, a total of M persons, is collected. Specifically, the information may include the following various processes.
First, it may include a 'location information gathering process' for collecting a location on a globe (hereinafter simply referred to as 'location') of a specific person who wants to collect information according to time. The positional tuple (= {time, latitude, longitude}), which is composed of (1) time, (2) latitude of the earth position in which the specific person exists, As a basic data type. In order to obtain information about a location, a particular person may carry a location information collection device capable of collecting location information on the earth, and the sets of location tuples collected by the location information collection device are related to the location information collection server Lt; / RTI > Such transmission may be performed in real time via wired / wireless communication, or the location information collection device may deliver the location information collection server offline. If there are other parameters that can replace the longitude and latitude, such a parameter may be substituted for the longitude and latitude to produce the set of location tuples. If the reliability of the location information belonging to the set of location tuples is questionable, the validation of the location information may be examined and a filtering process may be performed to remove the invalid information.
Second, a 'CTMC generation process' for a specific person to generate the CTMC for the specific person using the set of location tuples for the specific person obtained through the position information collection process described above may be included.
In order to generate CTMC (Continuous Time Markov Chain) according to the second comparative example, a plurality of clusters obtained by applying a clustering technique to data constituting a set of location tuples of a specific person can be used.
When the number of data related to the location information of the specific person collected by the location information collection server reaches a significant number and a set related to the location information of the specific person is formed, the location information collection server uses the clustering technique to locate the location of the specific person Clustering can be done. When the location information of a specific person is clustered, the information that the specific person has stayed in the area within a specific range for a considerable period of time appears as clusters as the number of the area where the particular person stays. Location points that do not belong to the cluster are considered to be points that are not considered as primary because they are points that appear during the movement.
As a representative technique used for clustering, there is 'expectation maximization clustering', which enables probability based clustering. The power law distribution (transition period of human movement) is a typical objective function that can be used for expectation-based clustering. Thus, in one embodiment of the present invention, a power distribution similar to the exponential distribution can be used. This distribution can be referred to as a modified exponential distribution in the present invention, where the parameter is the distance of human movement from the center of the residence area. The following equation (1) represents a probability distribution that can be used in an embodiment of the present invention.
Here, lambda is a controllable parameter indicating the maximum distance of the cluster, which can be fixed to a constant value. Also, x represents the distance between the current position of the specific person and the center of the cluster. At this time, since the human movement model is known to follow a so-called heavy-tailed distribution called Levy Walk, it can be judged that the use of this equation is suitable.
Considering this fact, the following steps can be repeated.
1. Initialize the cluster with location points belonging to the cluster.
2. Calculate the number of points in the cluster.
3. Calculate the probability that each location point belongs to a cluster with a probability distribution as shown in [Equation 1].
4. Correct the probability of points belonging to the cluster at the moving speed of the person represented by each point.
5. Repeat the EM clustering algorithm.
The probability of moving from one cluster to another can be calculated by counting the number of location points belonging to the cluster using the multiple clusters and calculating the posterior of the time relationship in which the location points appear among the clusters. The probability of moving from one cluster to another is not fixed but can be represented by a specific function. However, fixed values may be used for convenience of calculation. When a probability function to go from one cluster to another cluster is obtained, a cluster is defined as a state and a continuous time markov chain (CTMC) is constructed in which each probability function is a transition probability function. can do. The CTMC thus constructed can be evolved to be more accurate by changing not only the fixed position but also the position information of a specific person. When CTMC is configured and the current position of a particular person belongs to a specific state of the CTMC (that is, when the location of a specific person is specified), the CTMC transition probability function can be used to stochastically predict the next position of a particular person.
As described above, in the CTMC according to an embodiment of the present invention, when a probability function to go from one cluster to another cluster among the plurality of clusters is obtained, the cluster is set as a state, May refer to a continuous time marker chain with a transition probability function.
Information for assisting understanding of the CTMC generation process is disclosed in Korean Patent Application No. 10-2012-0119403 (Filing Date: 2012-10-26), 10-2012-0060839 (Filing Date: 2012-06-07), 10-2012-0071212 (Filing date: 2012-06-29).
Third, the preliminary information collection process may include a 'personality information collection process' for collecting information on the personality of a specific person who wants to collect information. The personality information gathering process can be performed through a questionnaire on personality in the online or offline manner to the specific person. According to the survey, you can get values about several kinds of personality scales that can identify the personality of a person. For example, a survey using the well-known BFI personality scale can be used to obtain specific numeric values for five different types of personality measures (OCEAN) for each particular person. In one embodiment of the present invention, human personality can be quantified using these five personality scales. For example, the lower the personality score, the closer to 0, and the higher the personality score, the closer the score is to 5.
In the second comparative example, as described above, it is possible to introduce a big five-toughness characteristic, which is a generally well-accepted model for toughness among many toughness models. The five factor score from the BFI is the main data used in the algorithm presented in the second comparative example.
The above-mentioned questionnaire for obtaining the personality information of a specific person can be made in a manner in which the questionnaire person and the specific person face each other. However, since this method is inefficient, a surveillance through IT technology may be carried out with the consent of the specific person to overcome the inefficiency.
In the second comparative example, the 'personality information collection process' described above may be performed by including a step of providing the personality information collection server with data stored in the personality inputting device from a specific person.
The above-described preliminary information collection step can be performed independently and repeatedly for all persons. Thus, the above-mentioned dictionary information can be provided as a different set of information for each person.
When the preliminary information collection step is completed, the prediction method according to the first mode and / or the prediction method according to the second mode described below can be performed.
[Prediction Method According to First Mode]
The prediction method according to the first mode of the second comparative example relates to a technique for calculating a probability value that the specific person will be located at a specific place by using the characteristic value of the specific person.
10 is a diagram for explaining an example of using a network model that reflects a characteristic value of a specific person and outputs a probability value that the specific person is located at a specific place. According to this example, using a network model comprising an
At this time, a 'time value' may be input to some nodes of the
[Table 1]
In addition, a 'characteristic value' of the specific person can be input to another node of the
Then, when the above-mentioned (1) time value and (2) characteristic value are input to the
To implement the above example, the weight associated with each link of the network model shown in Fig. 10 may be determined by learning. 10, the individual data for learning can be made up of values relating to a specific time, a probability that an arbitrary person will be located at a specific place at a certain time, and the toughness of the arbitrary person . A plurality of such individual data may be gathered to provide a complete data set for learning, wherein different individual data included in the entire data set may have at least one of a specific time and an identification value for an arbitrary person.
The entire learning process for learning the network model can be performed by repeating the individual learning process a plurality of times. In each individual learning process, for example, 2 shown in the third row of [Table 1] is input to the
In summary, the learning process of FIG. 10 may be regarded as learning the possibility that the specific person exists at a specific time when the BFF characteristic value of a specific person is known at a specific time.
When the BFF of another person not included in the learning data is known and the BFF of the other person is input to the network model together with the specific time when the learning is completed, The possibility of being located at a specific place can be outputted.
[Position prediction computing device for implementing the prediction method according to the first mode]
The location prediction computing apparatus for implementing the prediction method according to the first mode may include a data acquisition unit and a processing unit.
The processing unit may be configured to predict the position of the specific object using a learning model that includes an input layer, a hidden layer, and an output layer each of which is associated with each other by weighted links. The processing unit may output a probability value that the specific object is located at a specific place in the output layer when (1) a time value is input to the input layer, and (2) a characteristic value of the specific object is inputted.
The data obtaining unit may be adapted to obtain the time value and the characteristic value and provide the obtained time value and the characteristic value to the processing unit. The data acquiring unit acquires the above information from the external data collecting device and collects and collects the individual data temporarily or semi-permanently. Alternatively, the data acquiring unit acquires necessary data from a database existing in an external server and temporarily or semi- Can be stored.
Contents that can be helpful for implementing the prediction method according to the first mode described above are disclosed in Korean Patent Application No. 10-2014-0000361 (filing date: 2014-01-02).
[Prediction Method According to Second Mode]
FIG. 11 shows an example of a network model that is learned and provided using learning data, which can be used to provide a prediction method according to the second mode of the second comparative example.
The network model according to Fig. 11 is composed of five
In the process of learning the network model according to FIG. 11, the weights Vij and Wjk between the nodes can be set to a predetermined initial value. In the individual learning process, a value related to a specific time is input to the
When the above learning is completed, inputting to the input node a probability or a statistic value that exists at a specific time and at a specific place of the other person at the specific time with respect to another person not included in the learning data, The characteristic value of another person, such as the BFF toughness value, may be output.
In FIG. 11, a
[Position prediction computing apparatus for implementing the prediction method according to the second mode]
The location prediction computing apparatus for implementing the prediction method according to the second mode may include a data acquisition unit and a processing unit.
The processing unit may be configured to predict the toughness of the particular person using a learning model that includes an input layer, a hidden layer, and an output layer that are each made up of nodes associated with each other by weighted links. The processing unit may be configured such that (1) a value related to the toughness of the specific person is outputted from the output layer when a value related to the probability that a specific person exists in each of a plurality of positions is input to the input layer.
The data obtaining unit may obtain information on a value related to a probability that the specific person exists in each of a plurality of locations and provide the information to the processing unit.
Contents that can be helpful for implementing the prediction method according to the second mode described above are disclosed in Korean Patent Application No. 10-2014-0045450 (filing date: 2014-04-16).
The location information collecting apparatus and the personality information collecting apparatus described above may be the same apparatus, and may be simply referred to as a " user apparatus " herein. The location information collection server and the personality information collection apparatus described above may be the same server, and may be simply referred to as a 'server' herein. The server described above may not only collect information but also process or process the collected information to implement the embodiments of the present invention described above. According to an embodiment, the user device may include a role of the server. In the present specification, the 'server' and the 'user device' may be collectively referred to as a 'computing device'.
The network model illustrated in FIG. 10 is referred to as a first network model in this specification, and the network model illustrated in FIG. 11 may be referred to as a second network model in this specification.
Although the present invention has been described using the embodiment for acquiring the positional information by using the GPS device for convenience, the positional information obtained by using the positioning apparatus or the system other than the GPS is described in various embodiments You can understand what is applicable to the example.
<First Embodiment of Present Invention>
Providing location data or location-time data to the input layer of an artificial neural network (i.e., a neural network) through the examples described so far can help to understand the technique of outputting values about personality in the output layer.
At this time, in the above-described examples, a probability value existing at a specific position is used as the data type of the position data provided to the input layer. The probability value has a value between 0 and 1.
The first embodiment of the present invention relates to a technique for further simplifying the data type of the position-time data provided in the input layer of the neural network.
Although the first comparative example and the second comparative example described above are compared with the first embodiment which is the best embodiment according to the present invention, the first comparative example and the second comparative example described above can also be treated as embodiments of the present invention have.
To use the collected position-time data as training data, a preprocessing process is required. If the information about the location is captured by the latitude and longitude of GPS, or divided into small location categories, the size of the data becomes too large. Therefore, we can use classification without using latitude and longitude, and it is possible to reclassify small position category into large category. Among the position-time data collected through the position information collecting device, the position data can be divided into a large classification. For example, each category is a house, a school, a restaurant, a pub, and a nightlife. Also, if time is divided in units of one second or one minute in time, the size of data becomes large, and the memory can not cope with the operation. Therefore, for example, the time every 30 minutes can be used. For example, a total of 34 time points can be used for the time data, from 7:00 am to 2:00 pm for 30 minutes. The position-time data created in this criterion can have the form as shown in FIG.
12 shows an example of a graph in which a plurality of positions defined by predetermined criteria are associated with acquired time, according to an embodiment of the present invention.
In FIG. 12, the horizontal axis represents time. In this example, a total of 34 time points are defined on the horizontal axis for 30 minutes from 7:00 am to 2:00 pm.
12, the vertical axis indicates a plurality of positions defined by reclassifying the collected latitude and longitude according to the predetermined criteria. In the example of FIG. 12, indices of 0, 1, 2, 3, and 4 are shown on the vertical axis, and each index may be, for example, a house, a school, a restaurant, a pub, or a nightlife.
Data that can be represented as shown in FIG. 12 can be transformed into a data type as shown in FIG. 13 for use in an artificial neural network model, for example, a DBN (deep belief network) training.
13 shows an example of data formed by modifying collected geolocation data and data on toughness values according to an embodiment of the present invention.
As shown in Fig. 13, the information about the position illustrated in Fig. 12 can be displayed as a one-dimensional array having a shape of a two-dimensional array of 34 * 5 like train_array_x. Each element of the array can have a value of '0' or '1'. In this format, '0' represents a case where the specific time does not exist at a specific position, and '1' represents a case where it exists. That is, a total of 34 * 5 data listed in train_array_x can be regarded as a set of 34 sets of 34 data defined for each of the above five positions. For example, the first 34 of the total 34 * 5 pieces of data may be data indicating whether or not they were in 'home' for, for example, 34 points in the 30 minute interval. For example, the second 34 of the total 34 * 5 pieces of data may be data indicating whether or not they were in 'school' for, for example, thirty-four points in the 30 minute interval.
If you only exist in one location from 7:00 am to 2:00 pm in order to learn about the movement pattern, the position-time data may not be used for DBN training. In other words, it can be used for training only when there is movement.
DBN should be able to learn what kind of personality it has according to human movement patterns. Also, for the calculation of personality data, each person's personality should be numerically represented. In an embodiment of the present invention, the above-mentioned BFI (Big Five Inventory) can be used for characterization of personality. BFI is a theory established by McCrae and Costa. BFI classifies human personality into five elements, BFF (Big Five Factors). Each category has openness, enthusiasm, sympathetic, and nervousness. The average value of each element is 2.5. If there is a personality factor whose value is larger than the average value, the personality factor is higher than others. You can perform personality tests on a number of subjects to obtain data to train your DBN, and you can quantify your personality through BFI. Since each element has five elements, it is possible to make the second training data in the form of train_array_y as a one-dimensional array having a length of 5 as shown in FIG. This personality data represents the nature of the first training data, the position-time data.
Personality data can be expressed in one-to-one correspondence with position-time data. In this case, there are personality data as many as the number of position-time data.
In some cases, the location-time data may be collected multiple times for one experimenter, and the location of the stay may vary with each acquisition. In this case, the personality data may be expressed as one-to-N with position-time data.
In the present invention, since a pattern is assumed to exist according to human personality, a model capable of recognizing and classifying a pattern can be used for implementation of an embodiment of the present invention. For example, DBN can be used as a pattern classification model. DBN is a machine learning model that can classify patterns of uneducated data after training using training data.
In order to know the personality according to the human movement pattern, the model should know what kind of personality pattern the specific movement pattern represents and should be able to predict the personality only by the movement pattern. DBN is a suitable model for such pattern classification and prediction. DBN can be regarded as a layered model of several restricting boltzman machines (RBM), which is another machine learning model. First, the RBM consists of one hidden layer and one visible layer, and there are units in each layer. Units on the same layer are not connected and only have connections between units on different layers. There are also weights in the connections between the units. First, the training can be pre-training in Greedy-Layer wise fashion. Given an input to the RBM, the unit values and connection weights change between the visible and hidden layers. The value of the result of this training becomes the input of the next RBM layer. Fine tuning can be done after training to the highest level RBM by repeating the following training. You can fine tune the trained unit values and weights and the bias values that correct them. Once training is done, certain position-time data can be classified with a certain personality pattern. When the position-time data is given to the trained DBN, the personality type will be displayed, so that personality can be predicted.
An experiment using an embodiment of the present invention can be performed. For example, this experiment trains the DBN using the training data as inputs to the eight position-time data and personality data collected first. And uses the time and location data into the training data as test data to see if the trained DBN predicted the personality. The BFF value of the personality data used as training data can be compared with the BFF value as a result of the model. Thus, the smaller the difference between the two personality data, the more predictable the personality.
14 shows an example of a table comparing characteristics of training data with characteristics of a test result using an artificial neural network model trained according to an embodiment of the present invention.
14, there is an experimenter who has similar characteristics according to the position-time data used in the training and the predicted characteristics according to the test results. For the first experimenter, the nature of the training data and the nature of the test results are quite similar. In addition, it can be seen that personality factors other than the nervousness of
15A shows an example of an artificial neural network model that can be used to perform the toughness evaluation method according to an embodiment of the present invention.
FIG. 15B is a more detailed representation of the input layer of the artificial neural network model shown in FIG. 15A.
Hereinafter, a toughness evaluation method according to an embodiment of the present invention will be described with reference to FIGS. 15A and 15B.
The toughness evaluation method according to an exemplary embodiment of the present invention is a method for evaluating toughness of a
The artificial
The toughness evaluation method includes the steps of inputting input layer information on the time at which a specific person stays at each of the M positions into the input layer of the artificial neural network model to obtain toughness values for the particular person from the output layer of the artificial neural network model .
The
The
The hidden
Each of the total N * M input layer nodes is provided with data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes .
The value of the data E p, q for any p, q (where p is a natural number from 1 to N, q is a natural number from 1 to M) indicates whether the person was at position p at the time point indicated by the time index q And may be a binary number determined according to whether or not it is a binary number. For example , the value of the binary number E p, q may be '1' (or '0') if the person was at the position p at the time point indicated by the time index q. Conversely , the value of the binary number E p, q may be '0' (or '1') if the person is not at the position p at the time point indicated by the time index q.
15B, the number of nodes indicated by
The preprocessing process used in the first embodiment of the present invention is not limited to capturing the position information by GPS latitude and longitude, or by dividing it into small location categories. For example, the preprocessing process may use a large category, such as a house, a school, a restaurant, . Here, houses, schools, restaurants, pubs, and entertainment sites can be selected using the job code provided by the government. However, in the modified embodiment of the present invention, the information about the position may be divided into a plurality of clusters determined by the clustering algorithm described above without being classified by the latitude and longitude of the GPS or the small location category.
While the present invention has been described with reference to exemplary embodiments, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the invention. Accordingly, the true scope of the present invention should be determined by the technical idea of the appended claims.
Claims (7)
Wherein the artificial neural network model is learned by providing the input layer-information to an input layer of the artificial neural network model and providing the output layer-information to an output layer of the artificial neural network model,
Inputting the input layer information on the time at which a particular person stays for each of a plurality of locations into the input layer of the artificial neural network model to obtain a toughness value for the particular person from the output layer of the artificial neural network model, / RTI >
The input layer information includes N * M pieces of data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes It is provided in the form of a one-dimensional array,
The value of the data E p, q is set to the position p at the time point indicated by the time index q for any p, q (p is a natural number from 1 to N and q is a natural number from 1 to M) And a binary number having a value of '0' or '1'
Toughness evaluation method.
The computing device further includes a step of determining toughness of the specific person by comparing the toughness value of the acquired specific person with an average toughness value of a plurality of people,
The one-
A series of binary values determined according to whether the particular person was at one location for a plurality of viewpoints, and
Another series of binary values presented in succession to the series of binary values as another series of binary values determined according to whether the particular person was at another location relative to the plurality of time points
≪ / RTI >
Toughness evaluation method.
Wherein the artificial neural network model is learned by providing the input layer-information to an input layer of the artificial neural network model and providing the output layer-information to an output layer of the artificial neural network model,
The computing device inputs input layer information on the time at which a particular person stays for a plurality of locations into the input layer of the artificial neural network model to obtain toughness values for the particular person from the output layer of the artificial neural network model Step < RTI ID = 0.0 >
The input layer information includes N * M pieces of data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes It is provided in the form of a one-dimensional array,
The value of the data E p, q is calculated as to whether or not the specific person is at the position p at the time point indicated by the time index q, with respect to any p, q (p is a natural number of 1 to N and q is a natural number of 1 to M) Characterized by having a binary number having a value of '0' or '1'
Computing device.
Inputting layer-information on the time at which a particular person stays with respect to each of a plurality of positions into the input layer of the artificial neural network model to obtain a toughness value for the particular person from the output layer of the artificial neural network model A computer-readable recording medium having recorded thereon a program,
Wherein the artificial neural network model is learned by providing the input layer-information to an input layer of the artificial neural network model and providing the output layer-information to an output layer of the artificial neural network model,
The input layer information includes N * M pieces of data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes It is provided in the form of a one-dimensional array,
The value of the data E p, q is calculated as to whether or not the specific person is at the position p at the time point indicated by the time index q, with respect to any p, q (p is a natural number of 1 to N and q is a natural number of 1 to M) Characterized by having a binary number having a value of '0' or '1'
A computer-readable recording medium on which a program is recorded.
The computing device provides the input layer-information to the input layer of the artificial neural network model and provides a set of the output layer-information corresponding respectively to the input layer-information to the output layer of the artificial neural network model, Learning the model,
The input layer information includes N * M pieces of data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes It is provided in the form of a one-dimensional array,
The value of the data E p, q depends on whether or not the person was at the position p at the time point indicated by the time index q, for any p, q (p is a natural number from 1 to N and q is a natural number from 1 to M) Characterized in that it has a binary number with a value of '0' or '1'
A method of generating a toughness evaluation model.
Wherein the computing device provides the input layer information to the input layer of the artificial neural network model and provides a set of the output layer information corresponding to the input layer information to the output layer of the artificial neural network model, And a step of learning a neural network model is performed,
The input layer information includes N * M pieces of data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes It is provided in the form of a one-dimensional array,
The value of the data E p, q depends on whether or not the person was at the position p at the time point indicated by the time index q, for any p, q (p is a natural number from 1 to N and q is a natural number from 1 to M) Characterized in that it has a binary number with a value of '0' or '1'
Computing device.
The input layer information includes N * M pieces of data E i, j (i is a natural number of 1 to N, j is a natural number of 1 to M) composed of information on N positions and M time indexes It is provided in the form of a one-dimensional array,
The value of the data E p, q depends on whether or not the person was at the position p at the time point indicated by the time index q, for any p, q (p is a natural number from 1 to N and q is a natural number from 1 to M) Characterized in that it has a binary number with a value of '0' or '1'
A computer-readable recording medium on which a program is recorded.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020150089051A KR101680241B1 (en) | 2015-06-23 | 2015-06-23 | Method for predicting human personality based on data regarding human position having specific data type |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020150089051A KR101680241B1 (en) | 2015-06-23 | 2015-06-23 | Method for predicting human personality based on data regarding human position having specific data type |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| KR101680241B1 true KR101680241B1 (en) | 2016-12-06 |
Family
ID=57576387
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| KR1020150089051A KR101680241B1 (en) | 2015-06-23 | 2015-06-23 | Method for predicting human personality based on data regarding human position having specific data type |
Country Status (1)
| Country | Link |
|---|---|
| KR (1) | KR101680241B1 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109118763A (en) * | 2018-08-28 | 2019-01-01 | 南京大学 | Vehicle flowrate prediction technique based on corrosion denoising deepness belief network |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140064130A (en) * | 2012-11-19 | 2014-05-28 | 홍익대학교 산학협력단 | Probabilistically predicting location of object |
-
2015
- 2015-06-23 KR KR1020150089051A patent/KR101680241B1/en active IP Right Grant
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140064130A (en) * | 2012-11-19 | 2014-05-28 | 홍익대학교 산학협력단 | Probabilistically predicting location of object |
Non-Patent Citations (1)
| Title |
|---|
| 김승연 외, 위치 데이터를 이용한 성격 데이터의 추론, 제40회 한국정보처리학회 추계학술발표대회 논문집, 제20권 제2호, 2013.11.* * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109118763A (en) * | 2018-08-28 | 2019-01-01 | 南京大学 | Vehicle flowrate prediction technique based on corrosion denoising deepness belief network |
| CN109118763B (en) * | 2018-08-28 | 2021-05-18 | 南京大学 | Vehicle flow prediction method based on corrosion denoising deep belief network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kaunang et al. | Students' academic performance prediction using data mining | |
| CN108121785A (en) | A kind of analysis method based on education big data | |
| CN106663240A (en) | Systems and methods for data-driven identification of talent | |
| CN103902566B (en) | A kind of personality Forecasting Methodology based on microblog users behavior | |
| Yu | Academic Performance Prediction Method of Online Education using Random Forest Algorithm and Artificial Intelligence Methods. | |
| CN108109089A (en) | A kind of education can computational methods | |
| JPWO2018168220A1 (en) | Learning material recommendation method, learning material recommendation device, and learning material recommendation program | |
| Rangnekar et al. | Career prediction model using data mining and linear classification | |
| Intisar et al. | Classification of online judge programmers based on rule extraction from self organizing feature map | |
| Liao et al. | Course drop-out prediction on MOOC platform via clustering and tensor completion | |
| Khodeir | Student modeling using educational data mining techniques | |
| da Silva et al. | Elicitation of criteria weights for multicriteria models: Bibliometrics, typologies, characteristics and applications | |
| Loganathan et al. | Development of machine learning based framework for classification and prediction of students in virtual classroom environment | |
| Trabelsi et al. | The emotional machine: A machine learning approach to online prediction of user's emotion and intensity | |
| Billah et al. | A Data Mining Approach to Identify the Stress Level Based on Different Activities of Human | |
| KR101680241B1 (en) | Method for predicting human personality based on data regarding human position having specific data type | |
| Arif et al. | An Improved Prediction System of Students' Performance Using Classification model and Feature Selection Algorithm | |
| Nguyen et al. | SECC: Simultaneous extraction of context and community from pervasive signals | |
| Mateos-Mora et al. | A guide for the analysis of cultural scenes: a measurement proposal and its validation for the Spanish case | |
| CN108022057A (en) | Learning behavior analyzing method and system | |
| Sghir et al. | Using learning analytics to improve students' enrollments in higher education | |
| CN114936843A (en) | Method and device for evaluating matching degree of personnel and post | |
| Gara et al. | Mining Association Rules on Students Profiles and Personality Types | |
| Stamoulis et al. | Associations of Personality Traits and Emotional Intelligence In Social Network Consumers via Data Mining Techniques | |
| Nguyen | Temporal spike attribution: A local feature-based explanation for temporally coded spiking neural networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| E701 | Decision to grant or registration of patent right | ||
| GRNT | Written decision to grant |