| CPC G06Q 10/08345 (2013.01) [G06F 18/214 (2023.01); G06N 20/20 (2019.01)] | 19 Claims |
|
1. A system for predicting costs associated with shipping freight, comprising:
at least one processor programmed or configured to:
determine a predicted distance between an origin of a shipment of freight to a destination of the shipment of freight, wherein, when determining the predicted distance, the at least one processor is programmed or configured to:
generate a predicted distance machine learning model to provide a predicted distance between the origin of the shipment of freight and the destination of the shipment of freight; and
use the predicted distance machine learning model to provide an output, wherein the output comprises the predicted distance between the origin of the shipment of freight and the destination of the shipment of freight;
generate a training dataset for a freight production machine learning model, wherein a portion of the training dataset comprises the output of the predicted distance machine learning model;
generate the freight prediction machine learning model according to an ensemble method, to provide a predicted cost associated with shipping the shipment of freight from the origin of the shipment of freight to the destination of the shipment of freight base on the training dataset, wherein the freight prediction machine learning model comprises a random forest algorithm that provides a continuous variable as an output, wherein the continuous variable comprises the predicted cost associated with shipping the shipment of freight, wherein the training dataset comprises data associated with shipments of freight conducted during a first time interval and data associated with shipments of freight conducted during a second time interval, and wherein the first time interval is at least a year and wherein the second time interval is shorter than the first time interval, wherein, when generating the freight prediction machine learning model, the at least one processor is programmed or configured to:
train the freight prediction machine learning model based on the training dataset that comprises one or more outputs of the predicted distance machine learning model;
determine a predicted cost associated with shipping a specified shipment of freight from an origin of the specified shipment of freight to a destination of the specified shipment of freight using the freight predicting machine learning model;
generate an updated training dataset at a predetermined time interval, wherein the updated training dataset comprises data associated with shipments of freight conducted during an updated first time interval and data associated with shipments of freight conducted during an updated second time interval, and wherein the updated second time interval is shorter than the updated first time interval; and
retain the freight prediction machine learning model according to the predetermined time interval based on the updated training dataset;
wherein, when generating the freight prediction machine learning model, the at least one processor is programmed or configured to:
generate the freight prediction machine learning model such that at each step the random forest algorithm provides decision trees at random and keeps a decision tree that minimizes a remaining mean squared error (MSE) according to the formula:
 where y is an actual cost for a shipment of freight, ŷ is a predicted cost for a shipment of freight, and n is a number of observations in the training dataset.
|