Sensor-Based Outlier Day Detection With Optimal Feature Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately detect a user's irregular days due to variability in sensor types and reliance on raw data values, making it difficult to transfer models and lacking optimal labeling approaches for outlier detection.
Innovation Solution
A method and system that uses machine learning classification models to label features by applying multiple labeling approaches, selecting an optimal approach for each feature, and determining outlier days based on synthetic input and target features, allowing personalized recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple labeling approaches are applied to detect outlier days, then detection accuracy is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent segments the outlier detection process into multiple distinct labeling approaches (e.g., statistical methods, machine learning classifiers, rule-based systems). Each approach processes the sensor data independently to generate labels, allowing the system to evaluate multiple perspectives on what constitutes an outlier day without requiring one complex monolithic system.
Solution Approach 2:
The patent merges the results from multiple labeling approaches by aggregating their outputs to determine the final outlier classification. This combination allows the system to leverage the strengths of different approaches while compensating for their individual weaknesses, improving overall detection accuracy through ensemble decision-making.
2Reliability
If multiple labeling approaches are used for each feature, then detection reliability is improved, but resource consumption increases
Solution Approach 1:
The patent dynamically adjusts the application of labeling approaches based on data characteristics, device state, and detection needs. The system can activate or deactivate specific labeling approaches depending on the situation, allowing it to maintain high reliability when needed while conserving computational resources during normal operation or when resources are constrained.
Solution Approach 2:
The patent changes parameters such as the number of labeling approaches applied, the complexity of each approach, or the data sampling rate based on resource availability and detection requirements. This allows the system to optimize the balance between reliability and resource consumption by adjusting operational parameters in response to changing conditions.
3Ease of manufacture
If models are trained using raw data values, then model training is straightforward, but model transferability to different sensor types is reduced
Solution Approach 1:
The patent transforms raw sensor data values into standardized parameters or features that are independent of specific sensor types. By changing the representation of the data from raw values to normalized, dimensionless, or universally applicable parameters, the trained models can be transferred across different sensor types and devices without retraining, significantly improving model adaptability.
4Loss of information
If comprehensive data analysis is performed to identify outlier days, then personalized recommendations are improved, but processing time increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing sensor data, pre-computing features, and pre-training models during off-peak times or in advance. This allows the system to have processed data and trained models ready when needed, reducing the processing time required during actual outlier detection and recommendation generation while maintaining comprehensive analysis quality.
Data Source
AI summary
A method for detecting a user's outlier days uses data corresponding to features of the user acquired over multiple days by sensors on the user's electronic device. The data acquired for each day and feature is labeled as regular or irregular by applying N labeling approaches. One of the N labeling approaches compares the data for each feature with how values of previously acquired data for corresponding features are distributed. N labels are generated for the data for each feature and day. The machine learning classification model is trained using one of the N labels for each of the N labeling approaches. An optimal labeling approach is selected from among the N labeling approaches for each feature using the machine learning classification model. For each feature, the method determines whether each of the days is an outlier day for the user using the labels obtained with the optimal labeling approach.