Apparatus, computer program product, and method for predictive data labelling using a dual-prediction model system

An apparatus comprising at least one processor and at least one non-transitory memory, the non-transitory memory including computer-coded instructions thereon, the computer-coded instructions, in execution with the at least one processor, configure the apparatus to:

• generate, based at least in part on applying a candidate selection rule set to a candidate data set, (1) a positive candidate identifier set. (2) a positive candidate index set corresponding to the positive candidate identifier set, (3) a negative candidate identifier set, (4) a negative candidate index set corresponding to the negative candidate identifier set,(5) a neutral candidate identifier set, and (6) a neutral candidate index set corresponding to the neutral candidate identifier set,
• wherein the candidate data set is associated with a candidate pool identifier set comprising the positive candidate identifier set, negative candidate identifier set, and neutral candidate identifier set;
• train a candidate label probabilistic model based at least in part on a candidate label training subset associated with the candidate data set, wherein the candidate label training subset is identified based at least in part on (1) the positive candidate identifier set together with the positive candidate index set and a short-term record threshold, and (2) the negative candidate identifier set together with the negative candidate index set and the short-term record threshold;
• generate a candidate positive-label probability set corresponding to the candidate pool identifier set based at least in part on applying an index-limited candidate data subset to the candidate label probabilistic model,
• wherein the index-limited candidate data subset is based at least in part on (1) the positive candidate identifier set together with the positive candidate index set and the short-term record threshold, (2) the negative candidate identifier set together with the negative candidate index set and the short-term record threshold, and (3) the neutral candidate identifier set together with the neutral candidate index set and the short-term record threshold;
• train a historical record prediction model to predict the candidate positive-label probability set based at least in part on the candidate positive-label probability set and a long-term historical data subset associated with the candidate pool identifier set,
• wherein the long-term historical data subset is based at least in part on (1) the positive candidate identifier set together with the positive candidate index set and a long-term record threshold, (2) the negative candidate identifier set together with the negative candidate index set and the long-term record threshold, and (3) the neutral candidate identifier set together with the neutral candidate index set and the long-term record threshold;
• and
• utilize the historical record prediction model to generate a long-term candidate positive-label probability associated with at least one candidate identifier.

A computer program product comprising at least one non-transitory computer-readable storage medium having computer program code thereon, the computer program code, in execution with at least one processor, configured for:

• generating, based at least in part on applying a candidate selection rule set to a candidate data set, (1) a positive candidate identifier set. (2) a positive candidate index set corresponding to the positive candidate identifier set, (3) a negative candidate identifier set, (4) a negative candidate index set corresponding to the negative candidate identifier set,(5) a neutral candidate identifier set, and (6) a neutral candidate index set corresponding to the neutral candidate identifier set,
• wherein the candidate data set is associated with a candidate pool identifier set comprising the positive candidate identifier set, negative candidate identifier set, and neutral candidate identifier set;
• training a candidate label probabilistic model based at least in part on a candidate label training subset associated with the candidate data set, wherein the candidate label training subset is identified based at least in part on (1) the positive candidate identifier set together with the positive candidate index set and a short-term record threshold, and (2) the negative candidate identifier set together with the negative candidate index set and the short-term record threshold;
• generating a candidate positive-label probability set corresponding to the candidate pool identifier set based at least in part on applying an index-limited candidate data subset to the candidate label probabilistic model,
• wherein the index-limited candidate data subset is based at least in part on (1) the positive candidate identifier set together with the positive candidate index set and the short-term record threshold, (2) the negative candidate identifier set together with the negative candidate index set and the short-term record threshold, and (3) the neutral candidate identifier set together with the neutral candidate index set and the short-term record threshold;
• training a historical record prediction model to predict the candidate positive-label probability set based at least in part on the candidate positive-label probability set and a long-term historical data subset associated with the candidate pool identifier set,
• wherein the long-term historical data subset is based at least in part on (1) the positive candidate identifier set together with the positive candidate index set and a long-term record threshold, (2) the negative candidate identifier set together with the negative candidate index set and the long-term record threshold, and (3) the neutral candidate identifier set together with the neutral candidate index set and the long-term record threshold;
• and
• utilizing the historical record prediction model to generate a long-term candidate positive-label probability associated with at least one candidate identifier.