Automatically predicting shipper behavior using machine learning models

Abstract

Embodiments are disclosed for autonomously predicting shipper behavior. An example method includes the following operations. One or more learning models are generated. Shipper behavior data for at least one shipper is extracted. The shipper behavior data includes a plurality of features associated with the at least one shipper scheduled to ship one or more parcels. It is predicted whether one or more shipments will be sent or arrive at a particular time based at least in part on running the plurality of features of the at least one shipper through the one or more learning models.

A method comprising: 
     
         1 . receiving, by a shipper behavioral data management tool, raw shipper behavioral data associated with a shipper scheduled to ship at least one parcel;
 parsing, by the shipper behavioral data management tool, the raw shipper behavior data into a shipper information unit, wherein the shipper information unit comprises a subset of the shipper behavioral data;   accessing, by one or more computer processors from the shipper behavioral data management tool, the shipper information unit;   extracting, by the one or more computer processors, a set of features from the shipper information unit, wherein each feature of the set of features represents a characteristic of at least one of the shipper or the at least one parcel; and   processing, by the one or more computer processors using a shipper behavior learning model, the set of features to generate an output comprising a set of probability measures, wherein each probability measure of the set of probability measures represents a probability of the shipper being a member of a corresponding shipper class from a set of shipper classes.   
     
     
         2 . The method of  claim 1 , wherein extracting the set of features comprises generating at least one feature in the set of features by categorizing an element found in the shipper information unit so that the at least one feature represents a category of the characteristic of the at least one of the shipper or the at least one parcel. 
     
     
         3 . The method of  claim 1 , wherein extracting the set of features comprises generating at least two features in the set of features by separating an element present in the shipper information unit. 
     
     
         4 . The method of  claim 1 , wherein the shipper behavioral data management tool is configured to centrally collect and manage shipper behavior data provided by at least one of different service points, different vehicles, or different mobile computing entities. 
     
     
         5 . The method of  claim 1 , wherein the shipper behavior learning model has been trained to recognize patterns within a set of shipper information units derived from historical shipper behavioral data with respect to the set of shipper classes. 
     
     
         6 . The method of  claim 1  further comprising normalizing, by the shipper behavioral data management tool, the subset of the shipping behavioral data found in the shipper information unit. 
     
     
         7 . The method of  claim 1 , wherein the set of shipper classes comprises a timely shipper class, an early shipper class, and a late shipper class. 
     
     
         8 . A system comprising:
 a shipper behavioral data management tool configured to:   receive raw shipper behavioral data associated with a shipper scheduled to ship at least one parcel; and   parse the raw shipper behavior data into a shipper information unit, wherein the shipper information unit comprises a subset of the shipper behavioral data; and computing hardware configured to:   access, from the shipper behavioral data management tool, the shipper information unit;   extract a set of features from the shipper information unit, wherein each feature of the set of features represents a characteristic of at least one of the shipper or the at least one parcel; and   process, using a shipper behavior learning model, the set of features to generate an output comprising a probability measure, wherein the probability measure of represents a probability of a timeliness of the at least one parcel arriving at a sort.   
     
     
         9 . The system of  claim 8 , wherein the computing hardware is configured to extract the set of features to generate at least one feature in the set of features by categorizing an element found in the shipper information unit so that the at least one feature represents a category of the characteristic of the at least one of the shipper or the at least one parcel. 
     
     
         10 . The system of  claim 8 , wherein the computing hardware is configured to extract the set of features to generate at least two features in the set of features by separating an element present in the shipper information unit. 
     
     
         11 . The system of  claim 8 , wherein the shipper behavioral data management tool is configured to centrally collect and manage shipper behavior data provided by at least one of different service points, different vehicles, or different mobile computing entities. 
     
     
         12 . The system of  claim 8 , wherein the shipper behavior learning model has been trained to recognize patterns within a set of shipper information units derived from historical shipper behavioral data with respect to timeliness of parcels arriving at sorts. 
     
     
         13 . The system of  claim 8 , wherein the shipper behavioral data management tool is further configured to normalize the subset of the shipping behavioral data found in the shipper information unit. 
     
     
         14 . A method comprising:
 receiving, by a shipper behavioral data management tool, raw shipper behavioral data associated with a shipper scheduled to ship at least one parcel;   parsing, by the shipper behavioral data management tool, the raw shipper behavior data into a shipper information unit, wherein the shipper information unit comprises a subset of the shipper behavioral data;   accessing, by one or more computer processors from the shipper behavioral data management tool, the shipper information unit;   extracting, by the one or more computer processors, a set of features from the shipper information unit, wherein each feature of the set of features represents a characteristic of at least one of the shipper or the at least one parcel; and   processing, by the one or more computer processors using a shipper behavior learning model, the set of features to generate a prediction output corresponding to a size of the at least one parcel.   
     
     
         15 . The method of  claim 14 , wherein extracting the set of features comprises generating at least one feature in the set of features by categorizing an element found in the shipper information unit so that the at least one feature represents a category of the characteristic of the at least one of the shipper or the at least one parcel. 
     
     
         16 . The method of  claim 14 , wherein extracting the set of features comprises generating at least two features in the set of features by separating an element present in the shipper information unit. 
     
     
         17 . The method of  claim 14 , wherein the shipper behavioral data management tool is configured to centrally collect and manage shipper behavior data provided by at least one of different service points, different vehicles, or different mobile computing entities. 
     
     
         18 . The method of  claim 1 , wherein the shipper behavior learning model has been trained to recognize patterns within a set of shipper information units derived from historical shipper behavioral data with respect to various sizes of parcels. 
     
     
         19 . The method of  claim 1  further comprising normalizing, by the shipper behavioral data management tool, the subset of the shipping behavioral data found in the shipper information unit. 
     
     
         20 . The method of  claim 14 , wherein the shipper behavior learning model comprises at least one of a random forest based machine learning model or a gradient boosting based machine learning model.