Modeling for complex outcomes using clustering and machine learning algorithms

Abstract

Described herein are systems and methods for modeling complex outcomes using clustering and machine learning algorithms. Machine learning algorithms and models can be implemented on platforms comprising one or more user interfaces and an insight engine. In these embodiments, insight engine comprises a machine learning software algorithm (or module) configured to ingest data and generate insights.

What is claimed is: 
     
         1 . A computer-implemented system for multi-outcome modeling and input categorization using machine learning, the system comprising:
 (a) one or more processors;   (b) a non-transitory computer readable storage medium encoded with a computer program that causes the one or more processors to:
 (i) receive outcome data comprising a plurality of outcome categories for a plurality of subjects; 
 (ii) process said outcome data to generate a plurality of data points corresponding to said plurality of subjects; 
 (iii) compute distances between each of said plurality of data points based on said plurality of outcome categories, thereby generating a distance matrix; 
 (iv) set each of said plurality of data points as a cluster, thereby providing a plurality of clusters corresponding to said plurality of data points; 
 (v) identify two clusters that are closest within said plurality of clusters using said distance matrix and merge said two clusters that are closest into a single cluster; 
 (vi) update said distance matrix by replacing said two clusters that are closest with said single cluster; 
 (vii) repeat steps v)-vi) until a hierarchical relationship between each of said plurality of data points has been established, wherein said hierarchical relationship comprises two or more clusters defined by said plurality of outcome categories; and 
 (viii) obtain a data set for said plurality of subjects; 
 (ix) extract features from said data set using a natural language processing algorithm to generate a standardized data set; 
 (x) label said standardized data set with said plurality of outcome categories defining said two or more clusters; and 
 (xi) generate a cluster prediction classifier configured to categorize an input into at least one of said two or more clusters defined by said plurality of outcome categories, wherein said classifier is modeled with a machine learning algorithm using said standardized data set labeled with said plurality of outcome categories defining said two or more clusters. 
   
     
     
         2 . The system of  claim 1 , wherein said one or more processors are further caused to generate a dendrogram visually representing said hierarchical relationship between each of said plurality of data points. 
     
     
         3 . The system of  claim 1 , wherein said plurality of data points comprises binary values, nominal values, ordinal values, text or string values, quantitative values, or any combination thereof. 
     
     
         4 . The system of  claim 1 , wherein said distances are computed using simple matching distance, Jaccard's distance, Hamming distance, normalized rank transformation, Spearman distance, footrule distance, Kendall distance, Cayley distance, Ulam distance, Euclidean distance, City block distance, Chebyshev distance, Minkowski distance, Canberra distance, Bray Curtis distance, angular separation, or correlation coefficient. 
     
     
         5 . The system of  claim 1 , wherein said one or more processors are further caused to generate a dendrogram visually representing said two or more clusters. 
     
     
         6 . The system of  claim 1 , wherein said cluster prediction classifier comprises random forest, gradient boosted trees, recurrent neural networks, naïve Bayes classifiers, or penalized multinomial regression. 
     
     
         7 . The system of  claim 1 , wherein said one or more processors are further caused to use a natural language processing algorithm to process said data set, wherein said data set comprises information for said plurality of subjects retrieved from an electronic data record. 
     
     
         8 . The system of  claim 7 , wherein said one or more processors are further caused to use said natural language processing algorithm to extract a plurality of features from said electronic data record for training said cluster prediction classifier. 
     
     
         9 . The system of  claim 7 , wherein said natural language processing algorithm comprises one or more rules for keyword identification, unit conversion, internal consistency, or any combination thereof. 
     
     
         10 . The system of  claim 7 , wherein said natural language processing algorithm comprises a natural language processing model configured to annotate said electronic data record with standard labels. 
     
     
         11 . The system of  claim 7 , wherein said natural language processing algorithm comprises generating rules based on raw data from the electronic data record, and then using the rules to train a model to apply to said raw data to standardize said electronic data record. 
     
     
         12 . The system of  claim 7 , wherein said electronic data record comprises unstructured data, semi-structured data, structured data, or any combination thereof. 
     
     
         13 . The system of  claim 1 , wherein said one or more processors is further caused to generate one or more single-outcome predictive models using a machine learning algorithm. 
     
     
         14 . The system of  claim 13 , wherein said one or more processors is further caused to evaluate said input using said one or more single-outcome predictive models to generate an output comprising one or more predicted outcomes. 
     
     
         15 . The system of  claim 14 , wherein said one or more single-outcome predictive model comprise random forest, gradient boosted trees, penalized linear regression, penalized logistic regression, cox regression, naïve bays classifiers, support vector machines, or recurrent neural network. 
     
     
         16 . The system of  claim 1 , wherein said one or more processors is further caused to generate a visual representation of said input in relation to said hierarchical relationship comprising two or more cohorts. 
     
     
         17 . The system of  claim 1 , wherein said one or more processors are further caused to generate an insight through a user interface, wherein said insight is based on a classification of said input into at least one of said two or more clusters. 
     
     
         18 . The system of  claim 17 , wherein said one or more processors is further caused to display said insight through an interface comprising one or more portals. 
     
     
         19 . The system of  claim 17 , wherein said one or more processors is further caused to calculate a relative contribution of one or more features used to categorize said input into at least one of said two or more clusters. 
     
     
         20 . The system of  claim 1 , wherein said one or more processors is further caused to generate an event timeline comprising said plurality of outcome categories for said plurality of subjects.