Multi-layer artificial intelligence models for progressive predictions during data pipeline management

A method comprising:

• receiving, for a first predictive output associated with an entity, first input feature data for a decision-making pipeline, wherein the decision-making pipeline provides a plurality of predictive outputs over a time period at different times during the time period using a two-layer machine learning (ML) model framework of the decision-making pipeline;
• extracting, from the first input feature data, static features and dynamic features that are processed by different ML layers of the two-layer ML model framework, wherein the two-layer ML model framework comprises a first layer ML model and a second layer ML model, wherein the first layer ML model processes the static features and the second layer ML model processes the dynamic features, wherein the static features are associated with first data that is unchanged over the time period and the dynamic features are associated with second data that varies over the time period;
• determining, based on the segmented first input feature data, a first layer output of the first layer ML model at a first time for the entity;
• determining, based on the segmented first input feature data and the first layer output, a first predictive output of the two-layer ML model framework at the first time for the entity using the second layer ML model, wherein the second layer ML model includes an input layer having an input node for the first layer output and one or more modes for the segmented first input feature data;
• receiving second input feature data at a second time after the first time for the first predictive output;
• determining, based on the second input feature data and the first layer output, a second predictive output at the second time for the entity using the second layer ML model processing the static feature output with the dynamic features;
• and
• adjusting the first predictive output based on the second predictive output.

A non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations comprising:

• receiving, for a first predictive output associated with an entity, first input feature data for a decision-making pipeline, wherein the decision-making pipeline provides a plurality of predictive outputs over a time period at different times during the time period using a two-layer machine learning (ML) model framework of the decision-making pipeline;
• accessing static features and dynamic features that have been extracted from the first input feature data separately for processing by the two-layer ML model framework, wherein the two-layer ML model framework comprises a first layer ML model and a second layer ML model, wherein the first layer ML model processes the static features and the second layer ML model processes the dynamic features, wherein the static features are associated with first data that is unchanged over the time period and the dynamic features are associated with second data that varies over the time period;
• determining, based on the segmented first input feature data, a first layer output of the first layer ML model at a first time for the entity;
• determining, based on the segmented first input feature data and the first layer output, a first predictive output of the two-layer ML model framework at the first time for the entity using the second layer ML model, wherein the second layer ML model includes an input layer having an input node for the first layer output and one or more modes for the segmented first input feature data;
• receiving second input feature data at a second time after the first time for the first predictive output;
• determining, based on the second input feature data and the first layer output, a second predictive output at the second time for the entity using the second layer ML model processing the static feature output with the dynamic features;
• and
• adjusting the first predictive output based on the second predictive output.