Method of training machine learning models for making simulated estimations

1. A computer-implemented method of training machine learning models for making simulated estimations, the method comprising:

collecting, from a database, a set of historical business data;

applying one or more transformations to the set of historical business data to create a set of model features;

separating the set of model features into three or more pools based at least in part on a weather forecast, each pool comprising:

one or more model features of the set that are homogeneous with respect to a common value, and

a different type of feature from the other of the three or more pools;

for each of the three or more pools, dynamically creating a corresponding training set comprising the one or more model features of the pool and at least some of the historical business data;

for each of the three or more pools, creating a corresponding machine learning model;

for each training set, training the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set based at least in part on gradient boosting to predict a retail volume;

generating, a predicted labor demand by predicting, via at least one of the machine learning models, a labor demand value for a gap between an end date of the historical business data and a start date of a prediction time period;

generating a worker schedule based at least in part on the predicted labor demand; and

updating the machine learning models via at least:

deleting and recreating the machine learning models, and

for each training set, retraining the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set to improve the prediction of the retail volume via improving the correlation of each model to the features within the corresponding pool;

wherein:

each of the set of model features corresponds to a store; and

each of the set of model features corresponds to two or more departments of a same type and each department is from a different store.

2. A system comprising:

at least one processor; and

a memory device that stores an application that adapts the at least one processor to:

collect a set of historical business data from a database;

apply one or more transformations to the set of historical business data to create a set of model features;

separate the set of model features into three or more pools based at least in part on a weather forecast, each pool comprising:

one or more model features of the set that are homogeneous with respect to a common value, and

a different type of feature from the other of the three or more pools;

for each of the three or more pools, dynamically create a corresponding training set comprising the one or more model features of the pool and at least some of the historical business data;

for each of the three or more pools, create a corresponding machine learning model;

for each training set, train the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set based at least in part on gradient boosting to predict a retail volume;

generate a predicted labor demand by predicting, via at least one of the machine learning models, at labor demand value for a gap between an end date of the historical business data and a start date of a prediction time period;

generate a worker schedule based at least in part on the predicted labor demand; and

update the machine learning models via at least:

deleting and recreating the machine learning models as part of updating the machine learning models; and

for each training set, retraining the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set to improve the prediction of the retail volume via improving the correlation of each model to the features within the corresponding pool;

wherein:

each of the set of model features corresponds to a store; and

each of the set of model features corresponds to two or more departments of a same type and each department is from a different store.

3. A non-transitory computer-readable medium storing instructions that adapt at least one processor to:

collect a set of historical business data from a database;

apply one or more transformations to the set of historical business data to create a set of model features;

separate the set of model features into three or more pools based at least in part on a weather forecast, each pool comprising:

one or more model features of the set that are homogeneous with respect to a common value, and

a different type of feature from the other of the three or more pools;

for each of the three or more pools, dynamically create a corresponding training set comprising the one or more model features of the pool and at least some of the historical business data;

for each of the three or more pools, create a corresponding machine learning model;

for each training set, train the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set based at least in part on gradient boosting to predict a retail volume;

generate a predicted labor demand by predicting, via at least one of the machine learning models, a labor demand value for a gap between an end date of the historical business data and a start date of a prediction time period;

generate a worker schedule based at least in part on the predicted labor demand; and

update the machine learning models via at least:

deleting and recreating the machine learning models, and

for each training set, retraining the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set to improve the prediction of the retail volume via improving the correlation of each model to the features within the corresponding pool;

wherein:

each of the set of model features corresponds to a store; and

each of the set of model features corresponds to two or more departments of a same type and each department is from a different store.

4. The computer-implemented method of claim 1 , wherein the model features are separated into the three or more pools further based at least in part on a seasonality.

5. The system of claim 2 , wherein the model features are separated into the three or more pools further based at least in part on a seasonality.

6. The non-transitory computer-readable medium of claim 3 , wherein the model features are separated into the three or more pools further based at least in part on a seasonality.

7. A computer-implemented method of training machine learning models for making simulated estimations, the method comprising:

collecting, from a database, a set of historical business data;

applying one or more transformations to the set of historical business data to create a set of model features;

separating the set of model features into three or more pools based at least in part on a seasonality, each pool comprising:

one or more model features of the set that are homogeneous with respect to a common value, and

a different type of feature from the other of the three or more pools;

for each of the three or more pools, dynamically creating a corresponding training set comprising the one or more model features of the pool and at least some of the historical business data;

for each of the three or more pools, creating a corresponding machine learning model;

for each training set, training the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set based at least in part on gradient boosting to predict a retail volume;

generating, a predicted labor demand by predicting, via at least one of the machine learning models, a labor demand value for a gap between an end date of the historical business data and a start date of a prediction time period;

generating a worker schedule based at least in part on the predicted labor demand; and

updating the machine learning models via at least:

deleting and recreating the machine learning models, and

for each training set, retraining the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set to improve the prediction of the retail volume via improving the correlation of each model to the features within the corresponding pool;

wherein:

each of the set of model features corresponds to a store; and

each of the set of model features corresponds to two or more departments of a same type and each department is from a different store.

8. The computer-implemented method of training machine learning models for making simulated estimations of claim 7 further comprising:

for each machine learning model, testing the machine learning model to determine an accuracy of the machine learning model.

9. A system comprising:

at least one processor; and

a memory device that stores an application that adapts the at least one processor to:

collect a set of historical business data from a database;

apply one or more transformations to the set of historical business data to create a set of model features;

separate the set of model features into three or more pools based at least in part on a seasonality, each pool comprising:

one or more model features of the set that are homogeneous with respect to a common value, and

a different type of feature from the other of the three or more pools;

for each of the three or more pools, dynamically create a corresponding training set comprising the one or more model features of the pool and at least some of the historical business data;

for each of the three or more pools, create a corresponding machine learning model;

for each training set, train the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set based at least in part on gradient boosting to predict a retail volume;

generate a predicted labor demand by predicting, via at least one of the machine learning models, at labor demand value for a gap between an end date of the historical business data and a start date of a prediction time period;

generate a worker schedule based at least in part on the predicted labor demand; and

update the machine learning models via at least:

deleting and recreating the machine learning models as part of updating the machine learning models; and

for each training set, retraining the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set to improve the prediction of the retail volume via improving the correlation of each model to the features within the corresponding pool;

wherein:

each of the set of model features corresponds to a store; and

each of the set of model features corresponds to two or more departments of a same type and each department is from a different store.

10. The system of claim 9 , wherein the application further adapts the at least one processor to:

for each machine learning model, test the machine learning model to determine an accuracy of the machine learning model.

11. A non-transitory computer-readable medium storing instructions that adapt at least one processor to:

collect a set of historical business data from a database;

apply one or more transformations to the set of historical business data to create a set of model features;

separate the set of model features into three or more pools based at least in part on a seasonality, each pool comprising:

one or more model features of the set that are homogeneous with respect to a common value, and

a different type of feature from the other of the three or more pools;

for each of the three or more pools, dynamically create a corresponding training set comprising the one or more model features of the pool and at least some of the historical business data;

for each of the three or more pools, create a corresponding machine learning model;

for each training set, train the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set based at least in part on gradient boosting to predict a retail volume;

generate a predicted labor demand by predicting, via at least one of the machine learning models, a labor demand value for a gap between an end date of the historical business data and a start date of a prediction time period;

generate a worker schedule based at least in part on the predicted labor demand; and

update the machine learning models via at least:

deleting and recreating the machine learning models, and

for each training set, retraining the machine learning model, that corresponds to the pool that corresponds to the training set, on the training set to improve the prediction of the retail volume via improving the correlation of each model to the features within the corresponding pool;

wherein:

each of the set of model features corresponds to a store; and

each of the set of model features corresponds to two or more departments of a same type and each department is from a different store.

12. The non-transitory computer-readable medium of claim 11 , wherein the instructions further adapt at least one processor to:

for each machine learning model, test the machine learning model to determine an accuracy of the machine learning model.