Machine learning models for predicting time in traffic

1. A computerized method for using machine learning methods for modeling for time in traffic for a vehicle on a delivery route comprising:

collecting a set of traffic feature values from a database;

normalizing the set of traffic feature values;

providing a machine learning model;

inputting the set of normalized traffic features into the machine learning model;

training the machine learning model with the set of normalized traffic features;

determining a target time for the vehicle on the delivery route;

removing distances where the average historical time is less than one hundred (100) seconds, wherein the machine learning model comprises a deep neural network (DNN) model;

training the DNN model with a five (5) fold cross validation with a mean squared error as a loss function, wherein the machine learning model comprises a random forest (RF) model,

wherein the set of traffic feature values comprises a road distance, an air distance, a weekday, and an hour of day,

wherein the set of normalizing the set of traffic feature values are normalized to lie with a range of negative one (−1) and to positive one (+1),

wherein the set of traffic feature values comprises an average speed computed as the road distance divided by an average historical time to travel the road distance,

wherein the average speed is normalized to lie with the range of negative one (−1) and to positive one (+1),

wherein the target time comprises a normalized time that is a fractional deviation of a time in traffic time from the average historical time,

wherein the DNN model is four (4) layers deep with hyperbolic tangent activation function, and

wherein a set the number of neurons of the DNN are set at 32, 16, 8, and a linear layer at an end regression.

2. The computerized method of claim 1 wherein the RF model comprises an ensemble model that is set with RF one hundred (100) trees.