Method, apparatus, and computer program product for identifying hazardous conditions and predicting policy transaction behavior

A computer program product comprising at least one non-transitory computer readable storage medium storing computer code that, when executed by an apparatus, configure the apparatus to:

• receive, originating from a remote computing device, a plurality of digital image files associated with a property;
• extract, on a pixel by pixel basis and based at least in part on a first trained machine learning model, a plurality of predicted image-based plot properties associated with the property from one or more digital image files of the plurality of digital image files, wherein a predicted image-based plot property of the plurality of predicted image-based plot properties comprises a prediction of an image-based plot property present in the one or more digital image files and a predicted location of the image-based plot property in the one or more digital image files;
• input the plurality of predicted image-based plot properties to a second trained machine learning model;
• generate a plurality of policy transaction predictions by, for each predicted image-based plot property of the plurality of predicted image-based plot properties:
• generating, based at least in part on applying the second trained machine learning model to the plurality of predicted image-based plot properties, a policy transaction prediction comprising a programmatically generated likelihood that the predicted image-based plot property will lead to an occurrence of a policy transaction, wherein the second trained machine learning model evaluates the predicted locations of the image-based plot properties for the plurality of predicted image-based plot properties in relation to one another across the entire plurality of predicted image-based plot properties in generating the policy transaction prediction for a given predicted image-based plot property, and wherein generating the policy transaction prediction comprises:
• determining, by the second trained machine learning model, that the predicted image-based plot property represents one or more hazardous image-based plot properties;
• and
• assigning a severity to each or the one or more hazardous image-based plot properties;
• generating a hazard score based at least in part on a count of the one or more hazardous image-based plot properties represented by the predicted image-based plot property and severities associated therewith;
• and
• generating the policy transaction prediction based at least in part on the hazard score;
• generate, based at least in part on the plurality of policy transaction predictions, an aggregated policy transaction prediction for the property, the aggregated policy transaction prediction comprising a programmatically generated likelihood that a policy transaction will occur associated with the property;
• and
• transmit, to a requesting computing device, the aggregated policy transaction prediction.

A computer-implemented method, comprising:

• receiving, originating from a remote computing device, a plurality of digital image files associated with a property;
• extracting, on a pixel by pixel basis and based at least in part on a first trained machine learning model, a plurality of predicted image-based plot properties associated with the property from one or more digital image files of the plurality of digital Image files, wherein a predicted image-based plot property of the plurality of predicted image-based plot properties comprises a prediction or an image-based plot property present in the one or more digital image files and a predicted location of the image-based plot property in the one or more digital image files;
• inputting the plurality of predicted image-based plot properties to a second trained machine learning model;
• generating a plurality of policy transaction predictions by, for each predicted image-based plot property of the plurality of predicted image-based plot properties:
• generating, based at least in part on applying the second trained machine learning model to the plurality of predicted image-based plot properties, a policy transaction prediction comprising a programmatically generated likelihood that the predicted image-based plot property will lead to an occurrence of a policy transaction, wherein the second trained machine learning model evaluates the predicted locations of the image-based plot properties for the plurality of predicted image-based plot properties in relation to one another across the entire plurality of predicted image-based plot properties in generating the policy transaction prediction for a given predicted image-based plot property, and wherein generating the policy transaction prediction comprises:
• determining, by the second trained machine learning model, that the predicted image-based plot property represents one or more hazardous image-based plot properties;
• and
• assigning a severity to each or the one or more hazardous image-based plot properties;
• generating a hazard score based at least in part on a count of the one or more hazardous image-based plot properties represented by the predicted image-based plot property and severities associated therewith;
• and
• generating the policy transaction prediction based at least in part on the hazard score;
• generating, based at least in part on the plurality of policy transaction predictions, an aggregated policy transaction prediction for the property, the aggregated policy transaction prediction comprising a programmatically generated likelihood that a policy transaction will occur associated with the property;
• and
• transmitting, to a requesting computing device, the aggregated policy transaction prediction.