Multi-model prediction and resolution of order issues
Techniques are provided for multi-model prediction and resolution of order issues. An exemplary method for predicting issues that will delay an order comprises: obtaining features of the order; processing the features using a multi-model predictor to predict the issues that will delay the order, wherein the multi-model predictor comprises a first model that determines a probability that the order will be delayed and a second model that determines a probability that the order will be delayed for each of a plurality of different predefined order issue types; and aggregating an output of the first and second models to provide a likelihood of the issues that will delay the order. The multi-model predictor also optionally comprises a prediction model that predicts a number of the delay issues for a future time period and/or a cost of the delay issues for different predefined order issue types based on historical data.
1. A method for predicting one or more issues that will delay an electronic commerce order, comprising:
obtaining a plurality of features of said electronic commerce order;
processing, using at least one processing device, said plurality of features using a multi-model predictor to predict said one or more issues that will delay said electronic commerce order, wherein said multi-model predictor comprises a first model that determines a probability that said electronic commerce order will be delayed and a second model that determines a probability that said electronic commerce order will be delayed for each of a plurality of different predefined order issue types; and
aggregating an output of said first and second models to provide a likelihood of said one or more issues that will delay said electronic commerce order.
2. The method of claim 1 , wherein said multi-model predictor further comprises a prediction model that predicts one or more of a number of delay issues for a future time period and a cost of said delay issues for said plurality of different predefined order issue types based on historical data.
3. The method of claim 2 , wherein said prediction model comprises an autoregressive integrated moving average model.
4. The method of claim 1 , wherein said multi-model predictor further comprises a support vector machine model that calculates a mean of said plurality of features for a given customer, determines a distance of the mean from a plurality of orders of the given customer, and compares the distances to one or more predefined thresholds.
5. The method of claim 1 , wherein said first model comprises a logistic regression model.
6. The method of claim 1 , wherein said second model comprises one or more of an XG Boost algorithm and a gradient Boost algorithm.
7. The method of claim 1 , further comprising the step of providing feedback to one or more of a customer, a sales agent and an order processing device regarding one or more orders that are delayed.
8. The method of claim 1 , further comprising the step of recommending one or more resolutions of the one or more issues that will delay the electronic commerce order based on an analysis of historical transaction data.
9. The method of claim 1 , further comprising the step of identifying one or more order processing channels to be optimized.
10. A system for predicting one or more issues that will delay an electronic commerce order, comprising:
a memory; and
at least one processing device, coupled to the memory, operative to implement the following steps:
obtaining a plurality of features of said electronic commerce order;
processing, using at least one processing device, said plurality of features using a multi-model predictor to predict said one or more issues that will delay said electronic commerce order, wherein said multi-model predictor comprises a first model that determines a probability that said electronic commerce order will be delayed and a second model that determines a probability that said electronic commerce order will be delayed for each of a plurality of different predefined order issue types; and
aggregating an output of said first and second models to provide a likelihood of said one or more issues that will delay said electronic commerce order.
11. The system of claim 10 , wherein said multi-model predictor further comprises a prediction model that predicts one or more of a number of delay issues for a future time period and a cost of said delay issues for said plurality of different predefined order issue types based on historical data.
12. The system of claim 10 , wherein said multi-model predictor further comprises a support vector machine model that calculates a mean of said plurality of features for a given customer, determines a distance of the mean from a plurality of orders of the given customer, and compares the distances to one or more predefined thresholds.
13. The system of claim 10 , wherein said first model comprises a logistic regression model and wherein said second model comprises one or more of an XG Boost algorithm and a gradient Boost algorithm.
14. The system of claim 10 , further comprising the step of recommending one or more resolutions of the one or more issues that will delay the electronic commerce order based on an analysis of historical transaction data.
15. The system of claim 10 , further comprising the step of identifying one or more order processing channels to be optimized.
16. A computer program product for predicting one or more issues that will delay an electronic commerce order, comprising a tangible machine-readable storage medium having encoded therein executable code of one or more software programs, wherein the one or more software programs when executed by at least one processing device perform the following steps:
obtaining a plurality of features of said electronic commerce order;
processing, using at least one processing device, said plurality of features using a multi-model predictor to predict said one or more issues that will delay said electronic commerce order, wherein said multi-model predictor comprises a first model that determines a probability that said electronic commerce order will be delayed and a second model that determines a probability that said electronic commerce order will be delayed for each of a plurality of different predefined order issue types; and
aggregating an output of said first and second models to provide a likelihood of said one or more issues that will delay said electronic commerce order.
17. The computer program product of claim 16 , wherein said multi-model predictor further comprises a prediction model that predicts one or more of a number of delay issues for a future time period and a cost of said delay issues for said plurality of different predefined order issue types based on historical data.
18. The computer program product of claim 16 , wherein said multi-model predictor further comprises a support vector machine model that calculates a mean of said plurality of features for a given customer, determines a distance of the mean from a plurality of orders of the given customer, and compares the distances to one or more predefined thresholds.
19. The computer program product of claim 16 , wherein said first model comprises a logistic regression model and wherein said second model comprises one or more of an XG Boost algorithm and a gradient Boost algorithm.
20. The computer program product of claim 16 , further comprising the step of recommending one or more resolutions of the one or more issues that will delay the electronic commerce order based on an analysis of historical transaction data.