Oracle International Corporation

Massachusetts Institute of Technology

Oracle International Corpration

Kiran V Panchamgam

Su-Ming Wu

Aswin Kannan

Georgia Perakis

Sajith Vijayan

Setareh Borjian

Leann Thayaparan

Andrew Vakhutinsky

Setareh Borjian Boroujeni

Sandhya Lonial

Mengzhenyu Zhang

Bernard Griffiths

Maxime Cohen

Jeremy Kalas

Santosh Bai Reddy

Embodiments generate and train an optimized demand model for predicting a demand of an item. Embodiments receive a plurality of trees, each of the plurality of trees including one or more levels of splits and a plurality of nodes, each of the plurality of nodes corresponding to a demand feature that influences demand for the item. Embodiments store a first bound as a current bound for each of the plurality of trees. Starting at a top split of each of the plurality of trees, embodiments select a first demand feature that a greatest number of the plurality of trees split on. Embodiments optimize the first demand feature using the stored current bound to generate a second bound. Embodiments store the second bound as the current bound for each of the plurality of trees and move down each of the plurality of trees to a next level of splits.

Optimized tree ensemble based demand model

Embodiments generate an optimized demand model for a retail item. Embodiments train a tree ensemble machine learning model comprising a plurality of trees, the training comprising storing upper bounds for each of the trees, the trees comprising levels and branches that correspond to the demand features that influence demand for the item. Embodiments generate an objective function for the demand model. At a top split of each tree, embodiments determine optimal child nodes using the stored upper bounds and calculate a new feasible region for each tree. Using bounds on the new feasible region, embodiments move down each tree to a next level of splits and generate the optimized demand model when a leaf node of every tree has been reached.

Embodiments provide systems and methods for navigating a dialogue flow using a trained intelligence bot. Upon initiation of a chat session between a user and a trained intelligence bot, one or more utterances can be received. The utterances can be processed using the trained intelligence bot to resolve an intent from among a plurality of predefined intents, where the intelligence bot is trained to resolve predefined intents based on training data associated with the predefined intents. A predefined dialogue flow associated with the resolved intent can be navigated using the intelligence bot, where the intelligence bot guides the user through the dialogue flow using context variables that are associated with the user or the chat session. The user can be provided enterprise data retrieved by the intelligence bot using a retrieval request generated based on one or more of the navigation of the dialogue flow or the context variables.

Machine learning tool for navigating a dialogue flow

Systems, methods, and other embodiments associated with generating a price schedule are described. In one embodiment, for each customer segment of a plurality of customer segments, a per-segment value of an approximate objective function for the customer segment is determined by an optimizer, and a ratio of the per-segment value to a sum of all per-segment values for the customer segments is computed. The inventory quantity is allocated amongst the customer segments according to the ratio for each customer segment to form an inventory quantity for each customer segment. For each customer segment, a promotion portion of the price schedule that maximizes the objective function by the optimizer is determined. A quantity of remaining inventory allocated to the plurality of customer segments at an end of the regular season is aggregated. A markdown portion of the price schedule for the item that maximizes the objective function is determined by the optimizer. The promotion portion and the markdown portion are combined to create the price schedule for the item.

Computerized promotion and markdown price scheduling

One example of computerized inventory redistribution control includes, for each location inventory record in a set of location inventory records, calculating a quantity change that will bring a current item quantity to a different item quantity for the location inventory record. Determining a cost of a minimum-cost redistribution among the physical locations to effect the quantity changes. Determining a scaling factor that maximizes total revenue when the quantity changes are scaled by the scaling factor after deducting the cost scaled by the scaling factor. Generating transfer instructions for a redistribution of the item by scaling the transfer quantities of the minimum-cost redistribution by the scaling factor. Transmitting each transfer instruction to a computing device associated with a physical location indicated in the transfer instruction.

Computerized inventory redistribution control system

Embodiments determine a price schedule for an item by, for each item, receiving a set of prices for the item, an inventory quantity for the item, a per-segment demand model for the item, and an objective function that is a function of the per-segment demand model and maximizes revenue based at least on a probability of a return of the item and a cost of the return. Embodiments allocate the inventory quantity among a plurality of customer segments based at least on a predicted contribution of each customer segment to the objective function. Embodiments determine a markdown portion of the price schedule for the item that maximizes the objective function, where the markdown portion assigns a series of prices selected from the set of prices for respective time periods during a clearance season for the item.

Price optimization system

Systems, methods, and other embodiments associated with generating a price schedule are described. An inventory quantity for the item is allocated amongst a plurality of customer segments based on a predicted contribution of each customer segment to the objective function. For each customer segment, based on a quantity of inventory allocated to the customer segment, a promotion portion of the price schedule is determined that maximizes the objective function. Remaining inventory allocated to the plurality of customer segments at the end of the regular season is aggregated. Based on the aggregated inventory, a markdown portion of the price schedule for the item is determined that maximizes the objective function. The promotion portion and the markdown portion are combined to create a price schedule for the item. In one embodiment, a price schedule may be generated that includes promotions on top of markdown prices during the clearance season.

Systems, methods, and other embodiments associated with determining a promotion price schedule for each item in a group are described. In one embodiment, a method includes computing an item coefficient that corresponds to a change in a value of an objective function when the item is priced at the promotion price. The objective function is based on a multiple product demand model. An item coefficient is computed for each item, each time period in the price schedule, and each promotion price in a price ladder for the item. An approximate objective function is formulated that includes products of item coefficients and binary decision variables. The item coefficients, the approximate objective function, and constraints are provided to an optimizer that determines values of the decision variables that maximize the approximate objective function. A promotion price schedule is created for each item based on values of the decision variables.

Computerized promotion price scheduling utilizing multiple product demand model

Systems, methods, and other embodiments associated with generating a price schedule are described. A inventory quantity for the item is allocated amongst a plurality of customer segments based at least on a predicted contribution of each customer segment to the objective function. For each customer segment, based at least on a quantity of inventory allocated to the customer segment, a promotion portion of the price schedule is determined that maximizes the objective function. A quantity of remaining inventory allocated to the plurality of customer segments at the end of the regular season is aggregated. Based at least on the aggregated inventory, a markdown portion of the price schedule for the item is determined that maximizes the objective function. The promotion portion and the markdown portion are combined to create a price schedule for the item.

Systems, methods, and other embodiments associated with forecasting customer channel choice using cross-channel loyalty are described. In one embodiment, a method includes accessing historical values for each of one or more loyalty variables for respective customers. The method also includes determining respective loyalty variable predictors for each of the one or more loyalty variables for each customer based on the historical values. In response to a trigger event associated with a given customer, the loyalty variable predictors for the customer are used to estimate a present value of each of the one or more loyalty variables for the customer. The present value of each of the loyalty variables is input to a forecast model that calculates, for each channel, a probability that the customer will make a purchase using the channel. The purchase probabilities are provided for use in selecting a marketing message for the customer.

Forecasting customer channel choice using cross-channel loyalty

Systems, methods, and other embodiments associated with assortment and display space optimization are described. In one embodiment, a method creates an optimal planogram. The example method includes receiving data describing i) a set of items described by item dimensions, ii) display space dimensions; iii) business rules, and iv) a key performance indicator. A set of possible shelf positions is identified for each item. An expected sales volume is calculated for each item and shelf position pair based, at least in part, on a selected demand model. The method includes providing i) the expected sales volume for the item and shelf position pairs, ii) a set of constraints that embody the business rules, and iii) an objective function to an optimization problem solver that computes a solution. Based on the solution, a planogram is output that specifies the assortment of items and respective optimal shelf positions of the items.

Growing community of inventors

Burlington, MA, United States of America

Kiran V Panchamgam