Gaussian Copula Mixture Model for Non-Normal Healthcare Data Clustering
Solution Overview
Problem
Gaussian mixture models are limited in their ability to identify clusters in healthcare datasets where data is not normally distributed, as they assume data points within clusters follow a Gaussian distribution.
Innovation Solution
The method involves using a Gaussian copula mixture model (GCMM) that estimates parameters for each cluster using an expectation-maximization algorithm, determining inverse cumulative distributions, and updating parameters based on these distributions to create a model capable of identifying clusters in healthcare datasets, regardless of normal distribution assumptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If Gaussian mixture models are used for clustering, then the model training process is simplified, but the model cannot be applied to scenarios where data is not normally distributed
Solution Approach 1:
The patent transforms the data distribution parameters by applying inverse cumulative distribution functions to convert non-normally distributed data into normally distributed data. This parameter transformation allows the Gaussian mixture model to process non-normally distributed healthcare data while maintaining the simplicity of the Gaussian assumption, thus resolving the contradiction between model simplicity and adaptability to different data distributions
Solution Approach 2:
The patent introduces an intermediary transformation process using inverse cumulative distribution functions as a mediator between the raw non-normally distributed data and the Gaussian mixture model. This intermediary step converts the data characteristics without changing the fundamental Gaussian model structure, enabling the model to handle diverse data distributions while maintaining its computational simplicity
2Ease of operation
If traditional clustering algorithms are used, then the implementation is straightforward, but the algorithms cannot accurately identify clusters in non-normally distributed healthcare data
Solution Approach 1:
The patent applies parameter transformation through inverse cumulative distribution functions to convert non-normally distributed data into normally distributed data before clustering. This parameter change enables traditional Gaussian-based clustering algorithms to accurately identify clusters in originally non-normally distributed data, simultaneously maintaining implementation simplicity and improving cluster identification accuracy
Solution Approach 2:
The patent performs preliminary data transformation using inverse cumulative distribution functions before applying the clustering algorithm. This preliminary action prepares the data by transforming its distribution characteristics, ensuring that subsequent clustering operations can accurately identify clusters without requiring complex algorithm modifications
Data Source
AI summary
Disclosed are the embodiments for creating a model capable of identifying one or more clusters in a healthcare dataset. An input is received pertaining to a range of numbers. Each number in the range of numbers is representative of a number of clusters in the healthcare dataset. For a cluster, one or more first parameters of a distribution associated with the cluster are estimated. Thereafter, a threshold value is determined based on the one or more first parameters. An inverse cumulative distribution of each of one or more n-dimensional variables in the healthcare dataset is determined. The one or more first parameters are updated to generate one or more second parameters based on the estimated inverse cumulative distribution. A model is created for each number in the range of numbers based on the one or more second parameters.