Oracle International Corporation

Catalin Popescu

Ming Lei

Brent Li

Lin He

Jianwu Xu

Lawrence Leo Mesquita

Wendy L Etkind

Embodiments forecast demand of an item by receiving historical sales data for the item for a plurality of past time periods including a plurality of features that define one or more feature sets. Embodiments use the feature sets as inputs to one or more different algorithms to generate a plurality of different models. Embodiments train each of the different models. Embodiments use each of the trained models to generate a plurality of past demand forecasts for each of some or all of the past time periods. Embodiments determine a root-mean-square error ('RMSE') for each of the past demand forecasts and, based on the RMSE, determine a weight for each of the trained models and normalize each weight. Embodiments then generate a final demand forecast for the item for each future time period by combining a weighted value for each trained model.

Demand forecasting using weighted mixed machine learning models

Embodiments generate a model of demand of a product that includes an optimized feature set. Embodiments receive sales history for the product and receive a set of relevant features for the product and designate a subset of the relevant features as mandatory features. From the sales history, embodiments form a training dataset and a validation dataset and randomly select from the set of relevant features one or more optional features. Embodiments include the selected optional features with the mandatory features to create a feature test set. Embodiments train an algorithm using the training dataset and the feature test set to generate a trained algorithm and calculate an early stopping metric using the trained algorithm and the validation dataset. When the early stopping metric is below a predefined threshold, the feature test set is the optimized feature set.

Dynamic feature selection for model generation

Embodiments predict future demand for a first product by receiving historical sales data for an aggregate products/locations level, the historical sales data including a plurality of sales data points, including sales data points for the first product at each of a plurality of locations. Embodiments extract a plurality of different types of features related to sales of each of the products and generate a plurality of clusters of sales data points based on the plurality of different types of features. Embodiments train each of the clusters to generate a plurality of trained cluster models including promotion effects per cluster. For a particular time period, a particular location and the first product, embodiment identify the features for the time period and map to one of the trained cluster models to fetch the promotion effects for the time period. Embodiments then use the promotion effects to forecast demand for the first product.

Auto clustering prediction models

Embodiments generate a short life cycle sales curve for a short life cycle item. Embodiments generate a plurality of similar sales curves corresponding to at least one similar item that is similar to the short life cycle item. Embodiments parametrize each of the similar sales curves, including estimating, for each similar sales curve, a coefficient of innovation parameter, a coefficient of imitation parameter, and an error parameter, and determine a weight for each error parameter. Embodiments combine the coefficient of innovation parameters and the coefficient of imitation parameters using the weights, and generate the short life cycle sales curve using the combined coefficient of innovation parameters and the combined coefficient of imitation parameters.

Short life cycle sales curve estimation

Embodiments select demand forecast parameters for a demand model for one or more items, receive historical sales data for the items on a per store basis and receive a plurality of seasonality curves for a first item. Embodiments determine a repeatability of each of the seasonality curves using a correlation of each seasonality curve over year to year demand and retain a first seasonality curve based on the repeatability. Embodiments determine a smoothness of the first seasonality curve and determine a sparsity of the first seasonality curve. Based on the determined repeatability, smoothness and sparsity, embodiments determine that the first seasonality curve is reliable and repeat the receiving the plurality of seasonality curves, determining the repeatability, determining the smoothness, and determining the sparsity to determine a plurality of reliable seasonality curves. Embodiments use the demand model and the reliable seasonality curves and determine a demand forecast for the first item.

Optimization of demand forecast parameters

Systems, methods, and other embodiments are disclosed that are configured to characterize an effect on sales of a retail item due to a sales promotion. In one embodiment, first sales data for the retail item is retrieved from a plurality of stores that have applied the sales promotion for the retail item. Second sales data for the retail item is retrieved from a single store that has applied the sales promotion for the retail item. A combined promotion effect value is generated based on the first sales data and the second sales data. The combined promotion effect value characterizes an effect on sales of the retail item as sold by the single store due to the sales promotion.

System and method for blending promotion effects based on statistical relevance

Embodiments forecasting future demand for an item. Embodiments receive a regression based demand algorithm for the item that includes the set of features as regression variables and split the data points into a training set and a testing set. Embodiments assign each of the features of the set of features into one of a plurality of regularization categories and assign a penalty parameter to each of the features subject to regularization. Embodiments train the demand algorithm using the training set, the penalty parameters and the features to generate a trained demand model. Embodiments evaluate the trained demand model using the testing set to determine an early drop metric and repeat the assigning each of the features, the assigning the penalty parameter, the training the demand algorithm and the evaluating the trained demand model until the early drop metric meets a threshold.

Flexible feature regularization for demand model generation

Systems, methods, and other embodiments are disclosed that are configured to generate replenishment parameters for use by an external replenishment system. In one embodiment, sales statistics are generated for an item based at least in part on historical sales data for the item. A determination is made as to if demand forecast data is available for the item. If demand forecast data is not available, an order-point value for the item is generated based at least in part on the sales statistics. If demand forecast data is available, demand forecast statistics are generated and the order-point value is generated based at least in part on the sales statistics and the demand forecast statistics. An order-up-to-level is generated based at least in part on the order-point value.

Growing community of inventors

Atlanta, GA, United States of America

Catalin Popescu