Systems and methods for estimating stability of a dataset

A system, comprising:

• one or more processors;
• and
• memory storing thereon instructions that, as a result of being executed by the one or more processors, cause the system to:
• measure a set of payments and balances associated with one or more accounts corresponding to a customer, wherein the set of payments and balances is measured over a period of time through input data obtained from a set of data sources;
• process the set of payments and balances through a non-parametric algorithm to obtain rates of change corresponding to the set of payments and balances, wherein the rates of change are obtained as the input data is obtained;
• integrate the rates of change into a set of vectors, wherein the set of vectors includes different values corresponding to the input data;
• process the set of vectors through a mapping algorithm to generate a mapping of the set of vectors to a sequence of template states, wherein the mapping algorithm is trained using training data that includes a set of input vectors without any target values, and wherein the mapping algorithm generates the mapping according to a dictionary of template states generated by the mapping algorithm by identifying different clusters from the training data;
• process the mapping of the set of vectors to the sequence of template states over the period of time to generate a time series of the sequence of template states;
• process the time series through a classification algorithm to generate a set of features associated with the time series, wherein the classification algorithm uses a similarity measure to identify the set of features according to a proximity to the sequence of template states;
• process the set of features through a predictive model to identify a set of predictive characteristics associated with the customer, wherein the predictive model is trained using a dataset that includes a set of known trends in different sequences of template states;
• generate an offer for a new account based on the set of predictive characteristics;
• and
• update the predictive model according to feedback associated with the set of predictive characteristics and the offer.

A non-transitory, computer-readable storage medium storing thereon executable instructions that, as a result of being executed by one or more processors of a computer system, cause the computer system to:

• measure a set of payments and balances associated with one or more accounts corresponding to a customer, wherein the set of payments and balances is measured over a period of time through input data obtained from a set of data sources;
• process the set of payments and balances through a non-parametric algorithm to obtain rates of change corresponding to the set of payments and balances, wherein the rates of change are obtained as the input data is obtained;
• integrate the rates of change into a set of vectors, wherein the set of vectors includes different values corresponding to the input data;
• process the set of vectors through a mapping algorithm to generate a mapping of the set of vectors to a sequence of template states, wherein the mapping algorithm is trained using training data that includes a set of input vectors without any target values, and wherein the mapping algorithm generates the mapping according to a dictionary of template states generated by the mapping algorithm by identifying different clusters from the training data;
• process the mapping of the set of vectors to the sequence of template states over the period of time to generate a time series of the sequence of template states;
• process the time series through a classification algorithm to generate a set of features associated with the time series, wherein the classification algorithm uses a similarity measure to identify the set of features according to a proximity to the sequence of template states;
• process the set of features through a predictive model to identify a set of predictive characteristics associated with the customer, wherein the predictive model is trained using a dataset that includes a set of known trends in different sequences of template states;
• generate an offer for a new account based on the set of predictive characteristics;
• and
• update the predictive model according to feedback associated with the set of predictive characteristics and the offer.