Systems and methods for estimating stability of a dataset

Abstract

Disclosed embodiments may provide a framework to measure and leverage the observable attributes that most directly affect the data stability of a customer. In addition, embodiments track the dynamics of the observable components that sustain the data stability of a customer. Embodiments may be used to estimate the stability of a variety of conditions for various contexts, such as the stability of a computing system over time.

What is claimed is: 
     
         1 . A computer-implemented method comprising:
 repeatedly capturing customer data associated with various data sources in real-time as the customer data is generated, wherein the customer data includes dynamically changing financial customer data over a period of time, and wherein the customer data corresponds to a set of customers;   computing non-parametric rates of change corresponding to the dynamically changing financial customer data associated with the various data sources;   generating a set of vectors, wherein the set of vectors are generated in real-time as the customer data is received, and wherein the set of vectors are generated using the non-parametric rates of change according to the dynamically changing financial customer data associated with the various data sources;   mapping the set of vectors to directionally similar template states, wherein mapping includes performing unsupervised clustering of the set of vectors;   generating a time series of the directionally similar template states, wherein the time series is generated using the set of vectors over time and based on the dynamically changing financial customer data associated with the various data sources;   generating one or more features associated with the time series, wherein a feature is generated using a sequence corresponding to one or more directionally similar template states in the time series, and wherein generating features includes running the time series through a classification algorithm; and   determining a trend in the customer data associated with the various data sources, wherein the trend is determined by applying the features of the time series to a pre-defined model, and wherein the trend corresponds to a level of stability associated with the set of customers.