Noise-Suppression Classifiers Using Proximity Scores for Binary Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing supervised machine learning methods for classification tasks, particularly in fields like biology and finance, face challenges due to noisy independent variables contaminated by factors unrelated to the outcome, leading to poor specificity and accuracy in predicting binary outcomes.
Innovation Solution
The method involves calculating proximity scores by normalizing and dampening outliers in independent variables, using a combination of CSC and FCSI variables, and applying compression equations to suppress noise, thereby enhancing predictive power.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional supervised machine learning methods are used for classification, then the system can handle complex data with multiple independent variables, but the predictive accuracy and specificity deteriorate due to noisy variables contaminated by unrelated factors
Solution Approach 1:
The patent extracts and removes noisy independent variables from the dataset that are contaminated by factors unrelated to the outcome. This is achieved through statistical analysis to identify and eliminate variables with high noise levels, thereby improving predictive accuracy by focusing only on relevant features.
Solution Approach 2:
The patent creates a composite modeling approach by combining multiple processing techniques: noise suppression algorithms, variable selection methods, and ensemble modeling. This composite approach integrates different computational strategies to achieve superior predictive performance compared to individual methods.
2Loss of information
If more independent variables are included in the classification model, then the model can capture more information, but the predictive power decreases due to contamination from unrelated factors
Solution Approach 1:
The patent applies local quality by treating different independent variables differently based on their individual noise characteristics. Rather than uniformly processing all variables, the system identifies specific high-quality variables with low noise and gives them greater weight in the prediction model, while suppressing or removing noisy variables.
Solution Approach 2:
The patent changes parameters of the independent variables through normalization, standardization, and transformation processes. This involves adjusting the scale and distribution of variables to reduce noise impact while preserving the underlying signal, thereby maintaining information retention while improving predictive power.
3Ease of manufacture
If conventional classifier methods are used, then the implementation is straightforward, but the sensitivity and specificity are insufficient for accurate binary outcome prediction
Solution Approach 1:
The patent segments the classification process into distinct stages: data preprocessing, noise suppression, variable selection, model training, and validation. This segmentation allows each stage to be optimized independently while maintaining overall implementation simplicity through modular design and automated workflows.
Data Source
AI summary
This disclosure relates to methods for improving the accuracy of supervised machine learning for classification of complex data where a binary outcome prediction is desired from a multitude of independent variables that generally relate to the outcomes but are not highly specific in prediction. These methods are used in biology and in the physical world to determine likely outcomes based upon carefully selected input information termed independent variables. Classification is defined as the process of recognition, understanding, and grouping of objects and ideas into preset categories a.k.a. “sub-populations.” With the help of these pre-categorized training datasets, classification in machine learning programs leverage a wide range of algorithms to classify future datasets into respective and relevant categories. Supervised machine learning involves using a predetermine data set of independent variables with known outcomes to be used as training set for the process of building the predictive model.