System and methods for data evaluation through network sensitivity analysis

A non-transitory, computer-readable storage medium having encoded thereon machine instructions that when executed by a processor, cause the processor to:

• receive a plurality of data sets N, each of the N data sets comprising one or more parameters;
• identify in the plurality of N data sets, a dative contribution of one or more of the one or more parameters;
• apply the plurality of N data sets to a neural network and generate one or more observations O on the plurality of N data sets;
• execute a network sensitivity analysis (NSA) in which the processor:
• transforms the N=i . . . j . . . n data sets into ranked data sets of k distinct input parameters in the N=i . . . j . . . n data sets,
• generates a N NSA curve for each of a plurality k of distinct input parameters in the N data sets by generating an observation ON with the N data sets,
• generates a N−j NSA curve for each of the plurality of k distinct input parameters in which the processor:
• removes the jth data set from the N data sets;
• and
• generates, using the neural network, an observation ON−j with the jth data set removed,
• generates a measure Mj of a contribution of a jth data set based on the k N NSA curves and the k N−j NSA curves;
• generates a relative strength Sj of each of the N data sets as a function of the measure Mj;
• and
• predicts and causes display of one or more most significant parameters of the k distinct input parameters n the N data sets and provides the relative strength Si of one or more of the N data sets.

A method for evaluating the relative contribution of an individual data set Nj in a plurality of data sets N(i . . . j . . . n) to a problem solution O, the data sets N(i . . . j . . . n) processed and applied to a machine learning model, the method comprising:

• a processor executing a network sensitivity analysis (NSA), comprising:
• generating a N NSA curve for each of a plurality distinct input parameters in the data sets N(i . . . j . . . n), comprising computing a solution ON with all of the data sets N(i . . . j . . . n),
• generating a N−j NSA curve for each of the plurality of distinct input parameters, comprising:
• removing the jth data set from the data sets N(i . . . j . . . n);
• and
• generating a solution ON−j with the jth data set removed,
• determining a measure Mj of a contribution of a jth data set based on a difference between the N NSA curves and the N−j NSA curves;
• and
• generating and providing for display, one or more most significant parameters in the data sets, and generating and providing a relative strength Sj of each of the N(i . . . n) data sets as a function of the measure Mi.