Learning Model Generation for User Behavior Recognition

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing learning models generated using big data struggle to identify behavior with high accuracy due to the presence of unnecessary information, leading to reduced precision in behavior recognition.

Innovation Solution

An information processing device acquires attribute information and similar user data, along with appliance condition and sensor information, to generate a learning model that focuses on relevant data, using algorithms like Hidden Markov Models or Convolutional Neural Networks, thereby improving behavior identification accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If big data is used as learning data to generate a learning model, then the quantity of learning data is increased, but the accuracy of behavior identification deteriorates due to unnecessary information

Engineering Contradiction:

Improvequantity of learning dataVSAvoidaccuracy of behavior identification

Core Design Contradiction:

Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent segments the learning data generation process into distinct stages: first acquiring attribute information and identifying similar users, then selectively acquiring appliance condition information and sensor information only for those similar users. This segmentation prevents mixing data from dissimilar users, thereby maintaining data quality while still accumulating sufficient learning data quantity.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by making the data acquisition process user-specific. Instead of uniformly acquiring data from all users, the system acquires appliance condition information and sensor information selectively based on user similarity. This ensures that each user's learning data has high local quality (relevance) while the overall dataset maintains sufficient quantity.

Inventive Principle:

Principle #3Local quality

2Adaptability or versatility

If learning data from multiple users is aggregated to improve model generalization, then the versatility of the learning model is improved, but the precision of behavior recognition deteriorates due to inclusion of dissimilar user data

Engineering Contradiction:

Improvegeneralization capability of learning modelVSAvoidprecision of behavior recognition

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent changes the parameter of user selection from random or uniform sampling to similarity-based selection. By using attribute information as a selection parameter and identifying users with similar attributes, the system transforms the data aggregation process into a targeted collection of relevant data, maintaining precision while achieving versatility through diverse but similar user profiles.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20230030053A1Information processing device, communication system, and generation method

Publication Date: 2023.02.02 MITSUBISHI ELECTRIC CORP

AI summary

An information processing device includes an acquisition unit that acquires attribute information regarding a first user, acquires information indicating a user having attribute information similar to the attribute information regarding the first user, and acquires first acquisition information as at least one item of information out of appliance condition information as information regarding condition of an appliance used by the user and sensor information as information obtained by detecting the user by a sensor and a first generation unit that generates a first learning model, which identifies behavior of the first user, based on the first acquisition information.