System for time based monitoring and improved integrity of machine learning model input data

A computer program product for secured integrity of machine learning model input data, the computer program product comprising at least one non-transitory computer-readable medium having computer-readable program code portions embodied therein, the computer-readable program code portions comprising:

• an executable portion configured access a machine learning model trained using a training dataset to detect patterns associated with one or more users and communication channels, wherein the training dataset comprises malfeasant data correlated with a resource loss threshold associated with the one or more users;
• an executable portion configured to receive one or more data sets for machine learning model analysis that are distinct from the training dataset, wherein the one or more data sets comprise machine learning output, resource transaction information, and device data;
• an executable portion configured to parse the one or more data sets and identify one or more features based on meta data of the one or more data sets;
• an executable portion configured to transmit the one or more data sets to the machine learning model trained on the training dataset that comprises the malfeasant data, wherein the machine learning model comprises an artificial neural network algorithm;
• an executable portion configured to monitor, via a model feedback system, machine learning model analysis output for the one or more data sets;
• an executable portion configured to identify, via the model feedback system and by comparing output of the machine learning model for current data sets with output generated from prior data sets processed at an earlier time, a shift in the machine learning model analysis output of the one or more data sets, wherein the shift comprises corrupted data skewing the machine learning output;
• an executable portion configured to correlate features of the one or more data sets with the identified shift in the machine learning model analysis output to identify one or more volatile features in the one or more data sets, wherein the one or more volatile features correspond to data entries that exist in the one or more data sets and change during a specified time period;
• an executable portion configured to apply, via an automated loop, a weighting to the one or more volatile features in the one or more data sets to offset the shift in the machine learning model analysis output, wherein the weighting further comprises a dampening variable lessening effect of the one or more volatile features on the machine learning model analysis output;
• identify forecast weights to forecast data impacts based on at least one of the one or more data sets and one or more volatile features;
• and
• generate prophylactic actions comprising weighting of data sets to be taken on one or more users and communication channels based on sources, time since creation, and encryption associated with the one or more data sets.

A computer-implemented method for secured integrity of machine learning model input data, the method comprising:

• accessing a machine learning model trained using a training dataset to detect patterns associated with one or more users and communication channels, wherein the training dataset comprises malfeasant data correlated with a resource loss threshold associated with the one or more users;
• receiving one or more data sets for machine learning model analysis that are distinct from the training dataset, wherein the one or more data sets comprise machine learning output, resource transaction information, and device data;
• parsing the one or more data sets and identify one or more features based on meta data of the one or more data sets;
• transmitting the one or more data sets to the machine learning model trained on the training dataset that comprises the malfeasant data, wherein the machine learning model comprises an artificial neural network algorithm;
• monitoring, via a model feedback system, machine learning model analysis output for the one or more data sets;
• identifying, via the model feedback system and by comparing output of the machine learning model for current data sets with output generated from prior data sets processed at an earlier time, a shift in the machine learning model analysis output of the one or more data sets, wherein the shift comprises corrupted data skewing the machine learning output;
• correlating features of the one or more data sets with the identified shift in the machine learning model analysis output to identify one or more volatile features in the one or more data sets, wherein the one or more volatile features correspond to data entries that exist in the one or more data sets and change during a specified time period;
• applying, via an automated loop, a weighting to the one or more volatile features in the one or more data sets to offset the shift in the machine learning model analysis output, wherein the weighting further comprises a dampening variable lessening effect of the one or more volatile features on the machine learning model analysis output;
• identifying forecast weights to forecast data impacts based on at least one of the one or more data sets and one or more volatile features;
• generating prophylactic actions comprising additional weighting of the data sets to be taken on ore or more users and communication channels based on sources, time since creation, and encryption associated with the one or more data sets.