Back to results
US 11,379,865 B2Granted
Machine learned models for item price planning
AssigneeSap SE
InventorsRajat Agrawal, Uday Guntupalli
Life of this patent
Estimating the term…
Abstract
In an example embodiment, multiple machine learned models are used to continually learn from data to update various prediction models. Prediction of item prices, at the production and distribution point level, and utilizing this information along with other item information, such as crop yield, operational cost, and storage cost in the case of arming, may be used to solve the long- and short-term planning problems of farmers and help in decision making based on daily and the most up-to-date forecasts.
Claims20
A system comprising:
- at least one hardware processor;
- and
- a non-transitory computer-readable medium storing instructions that, when executed by the at least one hardware processor, cause the at least one hardware processor to perform operations comprising:
- receiving an identification of an item and an identification of a distribution point for the item from a graphical user interface;
- obtaining price history data of prices paid for the item by the distribution point on a plurality of past dates;
- feeding the price history data into a first basis prediction machine learned model trained to forecast a basis price for a plurality of future dates based upon the price history data, the basis price being a difference between a national price for the item and a price paid by the distribution point for the item;
- obtaining a location for the distribution point;
- identifying one or more distribution points nearby the distribution point, based on criteria maintained by a rules framework;
- obtaining price history data of prices paid for the item by the one or more distribution points nearby the distribution point;
- feeding the price history data for the distribution point and the price history data of prices paid for the item by the one or more distribution points nearby the distribution point to a second basis prediction machine learned model trained to forecast the basis price for the plurality of future dates;
- calculating a root mean square error for the first basis prediction machine learned model and the second basis prediction machine learned model;
- training a selection machine learned model using the root mean square error for the first basis prediction machine learned model, the root mean square error for the second basis prediction machine learned model, and a historical trend of a winning model between the first basis price prediction model and the second basis price prediction machine learned model;
- automatically selecting either the first basis prediction machine learned model or the second basis prediction machine learned model based on a prediction made by the selection machine learned model;
- and
- computing an item earnings forecast by adding the basis price predicted by the selected machine learned model to a forecast item price and subtracting one or more costs.
12 more elements
A method comprising:
- receiving an identification of an item and an identification of a distribution point for the item from a graphical user interface;
- obtaining price history data of prices paid for the item by the distribution point on a plurality of past dates;
- feeding the price history data into a first basis prediction machine learned model trained to forecast a basis price for a plurality of future dates based upon the price history data, the basis price being a difference between a national price for the item and a price paid by the distribution point for the item;
- obtaining a location for the distribution point;
- identifying one or more distribution points nearby the distribution point, based on criteria maintained by a rules framework;
- obtaining price history data of prices paid for the item by the one or more distribution points nearby the distribution point;
- feeding the price history data for the distribution point and the price history data of prices paid for the item by the one or more distribution points nearby the distribution point to a second basis prediction machine learned model trained to forecast the basis price for the plurality of future dates;
- calculating a root mean square error for the first basis prediction machine learned model and the second basis prediction machine learned model;
- training a selection machine learned model using the root mean square error for the first basis prediction machine learned model, the root mean square error for the second basis prediction machine learned model, and a historical trend of a winning model between the first basis price prediction model and the second basis price prediction machine learned model;
- automatically selecting either the first basis prediction machine learned model or the second basis prediction machine learned model based on a prediction made by the selection machine learned model;
- and
- computing an item earnings forecast by adding the basis price predicted by the selected machine learned model to a forecast item price and subtracting one or more costs.
9 more elements
A non-transitory machine-readable medium storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations comprising:
- receiving an identification of an item and an identification of a distribution point for the item from a graphical user interface;
- obtaining price history data of prices paid for the item by the distribution point on a plurality of past dates;
- feeding the price history data into a first basis prediction machine learned model trained to forecast a basis price for a plurality of future dates based upon the price history data, the basis price being a difference between a national price for the item and a price paid by the distribution point for the item;
- obtaining a location for the distribution point;
- identifying one or more distribution points nearby the distribution point, based on criteria maintained by a rules framework;
- obtaining price history data of prices paid for the item by the one or more distribution points nearby the distribution point;
- feeding the price history data for the distribution point and the price history data of prices paid for the item by the one or more distribution points nearby the distribution point to a second basis prediction machine learned model trained to forecast the basis price for the plurality of future dates;
- calculating a root mean square error for the first basis prediction machine learned model and the second basis prediction machine learned model;
- training a selection machine learned model using the root mean square error for the first basis prediction machine learned model, the root mean square error for the second basis prediction machine learned model, and a historical trend of a winning model between the first basis price prediction model and the second basis price prediction machine learned model;
- automatically selecting either the first basis prediction machine learned model or the second basis prediction machine learned model based on a prediction made by the selection machine learned model;
- and
- computing an item earnings forecast by adding the basis price predicted by the selected machine learned model to a forecast item price and subtracting one or more costs.
9 more elements
Dependencies are inferred from the claim text.