Real-time and computationally efficient prediction of values for a quote variable in a pricing application
The present disclosure describes a system, method, and computer program for real-time and computationally efficient calculation of a recommended value range for a quote variable, such as price, discount, volume, or closing time. The system uses the highest-density interval (HDI) of probability density function (PDF) as a recommended or suggested value range for a quote variable. PDFs for the quote variable are precomputed for groups of related inputs, and each PDF is summarized as an array of discrete points. A dimension reduction technique is applied to the PDF inputs in both the training and real-time (non-training) phases to reduce the number of possible combinations of PDFs. During a quote-creation process, a PDF look-up table enables the system to efficiently identify an applicable PDF from the group of precomputed PDFs based on reduced input values.
1. A method, performed by a pricing system having a customer relationship management (CRM) system running a customer-specific instance of a quote-creation application and a machine-learning server that executes a machine-learning module, for real-time and computationally-efficient calculation of a recommended value range for a quote variable, the method comprising:
in response to a user of the CRM pricing system engaging in a quote-creation process with the quote variable, the machine-learning module performing the following steps:
retrieving attributes of a customer from a customer database associated with the CRM system for which the quote is being created;
creating inputs in an input array for a probability density function (PDF) for the quote variable from the attributes;
reducing the inputs using a dimension reduction algorithm, wherein reducing the inputs using a dimension reduction algorithm comprises transforming the input array into a lower dimensional array and wherein transforming the input array into a lower dimensional array reduces the number of PDFs stored in a storage device of the machine-learning server;
identifying an applicable PDF for the quote variable from a plurality of precomputed PDFs, wherein the applicable PDF is identified using the reduced inputs and a PDF look-up table stored on the storage device of the machine-learning server that associates input values with a PDF and wherein, in the PDF look-up table, similar inputs are grouped together and associated with a precomputed PDF and wherein grouping similar inputs together and associating them with a precomputed PDF further reduces the number of PDFs stored in the storage device of the machine-learning server;
retrieving a stored array from the storage device of the machine-learning server that estimates the identified PDF;
calculating a highest-density interval for values for the quote variable using the stored array, wherein the highest-density interval is calculated for a given probability; and
displaying the calculated highest-density interval in a user interface of the customer-specific instance of the quote-creation application, wherein the highest-density interval is a prediction of values for the quote variable that the customer will accept with a given probability.
2. The method of claim 1 , wherein the quote variable is one of the following: price, closing time, product volume, or discount.
3. The method of claim 1 , wherein the highest-density interval is calculated and displayed in substantially real time as the user navigates to a user interface page related to the quote variable.
4. The method of claim 1 , wherein the dimension reduction algorithm is principal component analysis (PCA).
5. The method of claim 1 , wherein creating the PDF arrays comprises the following steps:
retrieving attributes of customers in a set of training data (“training customers”);
for each training customer, creating inputs for a PDF for the quote variable from the attributes and reducing the inputs using a dimension reduction algorithm;
grouping training customers with similar reduced input values;
creating a PDF for each group;
summarizing each PDF as an array of data points; and
storing the arrays.
6. The method of claim 5 , wherein the grouping training customers and creating PDFs steps comprise:
segmenting training input values into a grid; and
creating a PDF for each cell in the grid.
7. The method of claim 6 , wherein the PDF look-up table is based on the grid.
8. The method of claim 1 , wherein the probability is configurable by an administrator of the pricing system.
9. A non-transitory computer-readable medium comprising a computer program, that, when executed by a computer system having a customer relationship management (CRM) system running a customer-specific instance of a quote-creation application and a machine-learning server that executes a machine-learning module, enables the computer system to perform the following method for real-time and computationally-efficient calculation of a recommended value range for a quote variable, the method comprising:
in response to a user of the CRM system engaging in a quote-creation process with the quote variable, the machine-learning module performing the following steps:
retrieving attributes of a customer from a customer database associated with the CRM system for which the quote is being created;
creating inputs in an input array for a probability density function (PDF) for the quote variable from the attributes;
reducing the inputs using a dimension reduction algorithm, wherein reducing the inputs using a dimension reduction algorithm comprises transforming the input array into a lower dimensional array and wherein transforming the input array into a lower dimensional array reduces the number of PDFs stored in a storage device of the machine-learning server;
identifying an applicable PDF for the quote variable from a plurality of precomputed PDFs, wherein the applicable PDF is identified using the reduced inputs and a PDF look-up table stored on the storage device of the machine-learning server that associates input values with a PDF and wherein, in the PDF look-up table, similar inputs are grouped together and associated with a precomputed PDF and wherein grouping similar inputs together and associating them with a precomputed PDF further reduces the number of PDFs stored in the storage device of the machine-learning server;
retrieving a stored array from the storage device of the machine-learning server that estimates the identified PDF;
calculating a highest-density interval for values for the quote variable using the stored array, wherein the highest-density interval is calculated for a given probability; and
displaying the calculated highest-density interval in a user interface of the customer-specific instance of the quote-creation application, wherein the highest-density interval is a prediction of values for the quote variable that the customer will accept with a given probability.
10. The non-transitory computer-readable medium of claim 9 , wherein the quote variable is one of the following: price, closing time, product volume, or discount.
11. The non-transitory computer-readable medium of claim 9 , wherein the highest-density interval is calculated and displayed in substantially real time as the user navigates to a user interface page related to the quote variable.
12. The non-transitory computer-readable medium of claim 9 , wherein the dimension reduction algorithm is principal component analysis (PCA).
13. The non-transitory computer-readable medium of claim 9 , wherein creating the PDF arrays comprises the following steps:
retrieving attributes of customers in a set of training data (“training customers”);
for each training customer, creating inputs for a PDF for the quote variable from the attributes and reducing the inputs using a dimension reduction algorithm;
grouping training customers with similar reduced input values;
creating a PDF for each group;
summarizing each PDF as an array of data points; and
storing the arrays.
14. The non-transitory computer-readable medium of claim 13 , wherein the grouping training customers and creating PDFs steps comprise:
segmenting training input values into a grid; and
creating a PDF for each cell in the grid.
15. The non-transitory computer-readable medium of claim 14 , wherein the PDF look-up table is based on the grid.
16. The non-transitory computer-readable medium of claim 9 , wherein the probability is configurable by an administrator of the pricing system.
17. A computer system having a customer relationship management (CRM) system running a customer-specific instance of a quote-creation application and a machine-learning server that executes a machine-learning module for real-time and computationally-efficient calculation of a recommended value range for a quote variable, the computer system comprising:
one or more processors;
one or more memory units coupled to the one or more processors, wherein the one or more memory units store instructions that, when executed by the one or more processors, cause the computer system to perform the operations of:
in response to a user of the CRM system engaging in a quote-creation process with the quote variable, the machine-learning module performing the following steps:
retrieving attributes of a customer from a customer database associated with the CRM system for which the quote is being created;
creating inputs in an input array for a probability density function (PDF) for the quote variable from the attributes;
reducing the inputs using a dimension reduction algorithm, wherein reducing the inputs using a dimension reduction algorithm comprises transforming the input array into a lower dimensional array and wherein transforming the input array into a lower dimensional array reduces the number of PDFs stored in a storage device of the machine-learning server;
identifying an applicable PDF for the quote variable from a plurality of precomputed PDFs, wherein the applicable PDF is identified using the reduced inputs and a PDF look-up table stored on the storage device of the machine-learning server that associates input values with a PDF and wherein, in the PDF look-up table, similar inputs are grouped together and associated with a precomputed PDF and wherein grouping similar inputs together and associating them with a precomputed PDF further reduces the number of PDFs stored in the storage device of the machine-learning server;
retrieving a stored array from the storage device of the machine-learning server that estimates the identified PDF;
calculating a highest-density interval for values for the quote variable using the stored array, wherein the highest-density interval is calculated for a given probability; and
displaying the calculated highest-density interval in a user interface of the customer-specific instance of the quote-creation application, wherein the highest-density interval is a prediction of values for the quote variable that the customer will accept with a given probability.
18. The computer system of claim 17 , wherein the quote variable is one of the following: price, closing time, product volume, or discount.
19. The computer system of claim 17 , wherein the highest-density interval is calculated and displayed in substantially real time as the user navigates to a user interface page related to the quote variable.
20. The computer system of claim 17 , wherein the dimension reduction algorithm is principal component analysis (PCA).
21. The computer system of claim 17 , wherein creating the PDF arrays comprises the following steps:
retrieving attributes of customers in a set of training data (“training customers”);
for each training customer, creating inputs for a PDF for the quote variable from the attributes and reducing the inputs using a dimension reduction algorithm;
grouping training customers with similar reduced input values;
creating a PDF for each group;
summarizing each PDF as an array of data points; and
storing the arrays.
22. The computer system of claim 21 , wherein the grouping training customers and creating PDFs steps comprise:
segmenting training input values into a grid; and
creating a PDF for each cell in the grid.
23. The computer system of claim 22 , wherein the PDF look-up table is based on the grid.
24. The computer system of claim 17 , wherein the probability is configurable by an administrator of the pricing system.