At-home Prediction Device Dimension Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting whether a resident is at home in a household face challenges due to the need for device operating times, which can be difficult to obtain, leading to low prediction accuracy when using information like mobile communication terminal states or TV audience ratings, and combining multiple data types results in high-dimensional sparse data that further deteriorates accuracy.
Innovation Solution
An at-home prediction device that acquires household and prediction information, performs dimension compression on this data to reduce bias and create dense data, allowing for more accurate predictions by avoiding mutual influences between data types during compression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple data types are combined for prediction, then prediction coverage is improved, but data sparsity increases and accuracy deteriorates
Solution Approach 1:
The patent combines household information with multiple types of prediction information (mobile terminal state, TV audience rating, etc.) into a unified prediction model. By integrating these diverse data sources through dimension compression, the system achieves comprehensive prediction coverage while maintaining accuracy through proper data fusion techniques.
Solution Approach 2:
The patent applies dimension compression to transform high-dimensional sparse data into lower-dimensional dense representations. This parameter transformation converts the data structure from sparse high-dimensional form to dense low-dimensional form, resolving the accuracy deterioration caused by data sparsity while preserving prediction coverage.
2Measurement precision
If dimension compression is performed on combined data, then data density is improved, but mutual influence between data types may occur
Solution Approach 1:
The patent segments the dimension compression process by performing compression for each type of prediction information separately before combining results with household information. This segmented approach prevents mutual influence between different data types during compression while achieving dense data representation for each segment.
3Measurement precision
If household information is acquired for each household type, then prediction specificity is improved, but data complexity increases
Solution Approach 1:
The patent transforms complex high-dimensional household information into compressed lower-dimensional representations that retain essential characteristics for prediction. This parameter transformation reduces data structure complexity while maintaining prediction specificity by preserving the most relevant features during compression.
Data Source
AI summary
The at-home prediction device is a device that makes a prediction relating to a resident in a household being at home, and the at-home prediction device includes: a household information acquiring unit configured to acquire household information according to the number of households of each type in an area in which the household are located; a prediction information acquiring unit configured to acquire prediction information of a plurality of types other than the household information in the area that is used for the prediction; a dimension compressing unit configured to perform dimension compression on the prediction information together with the household information for each type of the prediction information; a prediction unit configured to make a prediction relating to the resident in the household being at home on the basis of information that is compressed in dimensions; and an output unit configured to output information representing a prediction result.