Learning apparatus, mental state sequence prediction apparatus, learning method, mental state sequence prediction method and program
A learning device includes: a psychological state data preprocessing unit that calculates a duration of a psychological state from psychological state sequence data, and generates preprocessed psychological state sequence data including the psychological state and the duration; and a learning unit that learns a psychological state sequence prediction model, using input sequence data including behavior sequence data and the preprocessed psychological state sequence data, and correct sequence data that is preprocessed psychological state sequence data at a time later than the input sequence data.
1. A learning apparatus comprising:
a memory; and
a processor configured to execute:
scanning each of behavior data columns that are extracted from a behavior sequence database,
replacing a missing value or an unexpected value with another value in a case where there is the missing value or the unexpected value, thereby generating past behavior sequence data;
calculating a duration of a mental state from mental state sequence data, and generating preprocessed mental state sequence data including the mental state and the duration; and
learning a mental state sequence prediction model, using input sequence data including the behavior sequence data and the preprocessed mental state sequence data, and correct sequence data that is preprocessed mental state sequence data at a time later than the input sequence data,
wherein the mental state is defined by mood data including valence and arousal, together with information about dates and times when a user makes self-reports, said valence indicating degree of positiveness and negativity of emotion of the user at that time, and said arousal indicating degree of excitement in the emotion of the user, and
wherein the processor is further configured to execute:
calculating the duration of the mental state by calculating a difference of an entry of the mood data and next entry of the mood data.
2. The learning apparatus according to claim 1 , wherein the processor is further configured to execute:
performing soft clustering on the preprocessed mental state sequence data obtained at the generating, and outputting a parameter set related to a mental state and a duration of each component,
wherein the learning includes learning the mental state sequence prediction model from a probability of belonging to each component and the parameter set, the probability of belonging being obtained by the mental state sequence prediction model from the input sequence data.
3. The learning apparatus according to claim 1 , wherein the processor is further configured to execute:
replacing the missing value or the unexpected value with the another value by inserting the mean value of a corresponding column in the behavior sequence database or “0” in the case of numerical data, and inserting a character string indicating that a value is missing in the case of character string data.
4. A mental state sequence prediction apparatus comprising:
a memory; and
a processor configured to execute:
scanning each of behavior data columns that are extracted from a behavior sequence database,
replacing a missing value or an unexpected value with another value in a case where there is the missing value or the unexpected value, thereby generating past behavior sequence data;
calculating a duration of a mental state from mental state sequence data, and generating preprocessed mental state sequence data including the mental state and the duration; and
predicting a future mental state and a duration of the future mental state, by inputting the past behavior sequence data and past preprocessed mental state sequence data obtained at the generating, to a mental state sequence prediction model learned with use of input sequence data including behavior sequence data and the preprocessed mental state sequence data, and correct sequence data that is preprocessed mental state sequence data at a time later than the input sequence data,
wherein the mental state is defined by mood data including valence and arousal, together with information about dates and times when a user makes self-reports, said valence indicating degree of positiveness and negativity of emotion of the user at that time, and said arousal indicating degree of excitement in the emotion of the user, and
wherein the processor is further configured to execute:
calculating the duration of the mental state by calculating a difference of an entry of the mood data and the next entry of the mood data.
5. The mental state sequence prediction apparatus according to claim 4 , wherein
the predicting includes predicting a future mental state and a duration of the future mental state, on a basis of a parameter set and a probability of belonging to a component, the parameter set being related to a mental state and a duration of each component, the parameter set being obtained by soft clustering performed on the preprocessed mental state sequence data during learning, the probability of belonging being obtained by the mental state sequence prediction model.
6. A learning method comprising:
scanning each of behavior data columns that are extracted from a behavior sequence database,
replacing a missing value or an unexpected value with another value in a case where there is the missing value or the unexpected value, thereby generating behavior sequence data;
calculating a duration of a mental state from mental state sequence data, and generating preprocessed mental state sequence data including the mental state and the duration; and
learning a mental state sequence prediction model, using input sequence data including the behavior sequence data and the preprocessed mental state sequence data, and correct sequence data that is preprocessed mental state sequence data at a time later than the input sequence data,
wherein the mental state is defined by mood data including valence and arousal, together with information about dates and times when a user makes self-reports, said valence indicating degree of positiveness and negativity of emotion of the user at that time, and said arousal indicating degree of excitement in the emotion of the user, and
wherein the processor is further configured to execute:
calculating the duration of the mental state by calculating a difference of an entry of the mood data and the next entry of the mood data.
7. A mental state sequence prediction method comprising:
scanning each of behavior data columns that are extracted from a behavior sequence database,
replacing a missing value or an unexpected value with another value in a case where there is the missing value or the unexpected value, thereby generating past behavior sequence data;
calculating a duration of a mental state from mental state sequence data, and generating preprocessed mental state sequence data including the mental state and the duration; and
predicting a future mental state and a duration of the future mental state, by inputting past behavior sequence data and past preprocessed mental state sequence data obtained at the generating, to a mental state sequence prediction model learned with use of input sequence data including behavior sequence data and the preprocessed mental state sequence data, and correct sequence data that is preprocessed mental state sequence data at a time later than the input sequence data,
wherein the mental state is defined by mood data including valence and arousal, together with information about dates and times when a user makes self-reports, said valence indicating degree of positiveness and negativity of emotion of the user at that time, and said arousal indicating degree of excitement in the emotion of the user, and
wherein the processor is further configured to execute:
calculating the duration of the mental state by calculating a difference of an entry of the mood data and the next entry of the mood data.
8. A non-transitory computer-readable recording medium having computer-readable instructions stored thereon, which when executed, cause a computer to function as the learning apparatus according to claim 1 .