Behavior Prediction Model Using Effort and Habituation Features

Solution Overview

Problem

Existing behavior prediction methods require large amounts of data and manual hyperparameter tuning, making accurate predictions difficult, especially when data is limited, and are inefficient in extracting relevant features for prediction.

Innovation Solution

A behavior prediction apparatus that evaluates effort and habituation features from a person's history to train a prediction model using these features and time intervals between behaviors, reducing the complexity of data analysis and hyperparameter tuning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If deep learning techniques are used to automatically extract patterns from past history information, then prediction capability is improved, but large amounts of data are required and significant time is spent on hyperparameter tuning

Engineering Contradiction:

Improveprediction accuracyVSAvoiddata volume

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts and utilizes pre-defined domain knowledge features (effort feature and habituation feature) from the behavior history data, rather than attempting to automatically learn all features from raw data. This extraction approach allows accurate prediction with limited data by focusing on the most relevant pre-identified features.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent performs preliminary feature engineering by explicitly defining effort features and habituation features before model training. This preliminary preparation of meaningful features enables the model to achieve good prediction performance without requiring extensive data or complex automatic feature learning.

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If deep learning techniques are used to automatically extract patterns from past history information, then prediction capability is improved, but significant time is spent on hyperparameter tuning

Engineering Contradiction:

Improveprediction accuracyVSAvoidhyperparameter tuning time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts and utilizes pre-defined domain knowledge features (effort feature and habituation feature) from the behavior history data, rather than attempting to automatically learn all features from raw data. This extraction approach allows accurate prediction with limited data by focusing on the most relevant pre-identified features.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the approach from automatic feature learning with many hyperparameters to using explicitly defined features with fewer parameters. By transforming the problem into one that uses domain knowledge features, the need for extensive hyperparameter tuning is eliminated while maintaining prediction accuracy.

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If automatic feature extraction is used, then the system can handle infinite possibilities, but it requires large amounts of data and complex processing

Engineering Contradiction:

Improvefeature extraction capabilityVSAvoidsystem complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent extracts and utilizes pre-defined domain knowledge features (effort feature and habituation feature) from the behavior history data, rather than attempting to automatically learn all features from raw data. This extraction approach allows accurate prediction with limited data by focusing on the most relevant pre-identified features.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of attempting to extract all possible features from the data (excessive action), the patent focuses on extracting and utilizing only the most critical features - effort and habituation - that are sufficient for accurate prediction. This partial approach reduces complexity while maintaining effectiveness.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20230394363A1Behavior prediction method, behavior prediction apparatus and program

Publication Date: 2023.12.07 NT T INC

AI summary

A computer evaluates, from a first behavior history including, for each of a plurality of behaviors of a person, a time of the behavior and a numerical value indicating the person's state after the behavior, a first feature indicating a first amount of effort the person makes until the numerical value exceeds a threshold at a certain point in time; evaluates, from the first behavior history, a second feature indicating a degree of the person's habituation to a state indicated by the threshold by the certain point in time; and trains a prediction model, in which the first and second features are used as explanatory variables and a time interval from a behavior at the certain point in time to a next behavior in the first behavior history is used as an explained variable, based on the first and second features and the time interval.