Causal inference machine learning with statistical background subtraction

1. A computer-implemented method, comprising:

receiving, with a server comprising a processor and memory, historical sales data for one or more past time periods and corresponding historical data for one or more causal variables;

deconfounding the cause-effect relationship of historical sales data and historical data on the one or more causal variables by conducting one or more randomized controlled AB group trials that reduce an effect of one or more confounders on one or more variables;

defining one or more sample weights for statistical background subtraction of the historical data;

performing statistical background subtraction on the historical data; and

training, by an iterative approach comprising cyclic boosting in additive regression mode, a first machine learning model to predict an absolute individual causal effect on a considered demand quantity in relation to the one or more causal variables and one or more sample weights.

2. The computer-implemented method of claim 1 , further comprising:

predicting, with the first machine learning model, an absolute individual causal effect on one or more considered demand quantities during a prediction period by training a second machine learning model on historical data, wherein a training period is earlier than the prediction period; and

rendering, for display on a user interface, a prediction feature explanation visualization comprising a predicted absolute individual causal effect and one or more features identified during the training of the second machine learning model that influence the predicted value.

3. The computer-implemented method of claim 2 , wherein the one or more randomized controlled AB group trials comprise the server issuing one or more discount coupons to a subset of a population of customers.

4. The computer-implemented method of claim 2 , further comprising:

performing, with the server, one or more additional training and prediction steps to calculate and apply independence weights by means of inverse propensity scores, estimated by a third machine learning model, to predicting the values of one or more causal variables by including one or more confounding variables as features.

5. The computer-implemented method of claim 2 , further comprising:

defining, with the server, one or more specific feature sequences in combination with a coordinate descent optimization; and

performing, with the server, one or more regularization and smoothing techniques, incorporating the defined one or more specific feature sequences, during the training of the first machine learning model.

6. A system, comprising:

a server comprising a processor and memory and configured to:

receive historical sales data for one or more past time periods and corresponding historical data for one or more causal variables;

deconfound the cause-effect relationship of historical sales data and historical data on the one or more causal variables by conducting one or more randomized controlled A/B group trials that reduce an effect of one or more confounders on one or more variables;

define one or more sample weights for statistical background subtraction of the historical data;

perform statistical background subtraction on the historical data; and

train, by an iterative approach comprising cyclic boosting in additive regression mode, a first machine learning model to predict an absolute individual causal effect on a considered demand quantity in relation to the one or more causal variables and one or more sample weights.

7. The system of claim 6 , further comprising the server:

predicting, with the first machine learning model, an absolute individual causal effect on one or more considered demand quantities during a prediction period by training a second machine learning model on historical data, wherein a training period is earlier than the prediction period; and

rendering, for display on a user interface, a prediction feature explanation visualization comprising a predicted absolute individual causal effect and one or more features identified during the training of the second machine learning model that influence the predicted value.

8. The system of claim 7 , wherein the one or more randomized controlled AB group trials comprise the server issuing one or more discount coupons to a subset of a population of customers.

9. The system of claim 7 , further comprising the server:

performing one or more additional training and prediction steps to calculate and apply independence weights by means of inverse propensity scores, estimated by a third machine learning model, to predicting the values of one or more causal variables by including one or more confounding variables as features.

10. The system of claim 7 , further comprising the server:

defining one or more specific feature sequences in combination with a coordinate descent optimization; and

performing one or more regularization and smoothing techniques, incorporating the defined one or more specific feature sequences, during the training of the first machine learning model.

11. A non-transitory computer-readable storage medium embodied with software, the software when executed configured to:

receive, with a server comprising a processor and memory, historical sales data for one or more past time periods and corresponding historical data for one or more causal variables;

deconfound the cause-effect relationship of historical sales data and historical data on the one or more causal variables by conducting one or more randomized controlled A/B group trials that reduce an effect of one or more confounders on one or more variables;

define one or more sample weights for statistical background subtraction of the historical data;

perform statistical background subtraction on the historical data; and

train, by an iterative approach comprising cyclic boosting in additive regression mode, a first machine learning model to predict an absolute individual causal effect on a considered demand quantity in relation to the one or more causal variables and one or more sample weights.

12. The non-transitory computer-readable storage medium of claim 11 , wherein the software when executed is further configured to:

predict, with the first machine learning model, an absolute individual causal effect on one or more considered demand quantities during a prediction period by training a second machine learning model on historical data, wherein a training period is earlier than the prediction period; and

render, for display on a user interface, a prediction feature explanation visualization comprising a predicted absolute individual causal effect and one or more features identified during the training of the second machine learning model that influence the predicted value.

13. The non-transitory computer-readable storage medium of claim 12 , wherein the one or more randomized controlled AB group trials comprise the server issuing one or more discount coupons to a subset of a population of customers.

14. The non-transitory computer-readable storage medium of claim 12 , wherein the software when executed is further configured to:

perform one or more additional training and prediction steps to calculate and apply independence weights by means of inverse propensity scores, estimated by a third machine learning model, to predicting the values of one or more causal variables by including one or more confounding variables as features.

15. The non-transitory computer-readable storage medium of claim 12 , wherein the software when executed is further configured to:

define one or more specific feature sequences in combination with a coordinate descent optimization; and

perform one or more regularization and smoothing techniques, incorporating the defined one or more specific feature sequences, during the training of the first machine learning model.