Systems for predicting activity distribution across mediums and methods thereof

A method comprising:

• receiving, by at least one processor, a plurality of activity records associated with a plurality of entities;
• wherein each activity record of the plurality of activity records comprise:
• i) an activity entity identifier associated with an entity of the plurality of entities,
• ii) a date indicator identifying a date associated with each activity record, iii) an activity location identifier associated with an entity location associated with each activity record, and
• iv) an activity value;
• accessing, by the at least one processor, via an external service, enriched entity data associated with the at least one entity;
• wherein the enriched entity data comprises at least one attribute of the at least one entity to add to each activity record;
• wherein the at least one attribute is associated with at least one physical location at which the at least one entity has a presence;
• formatting, by the at least one processor, based on the at least one attribute of the enriched data, the enriched entity data and the plurality of activity records to standardize the enriched entity data and the plurality of activity records into a common format;
• generating, by the at least one processor, a plurality of enriched activity records comprising the plurality of activity records modified with the enriched entity data in the common format;
• determining, by the at least one processor, enriched first channel activity records of the plurality of enriched activity records based on the at least one attribute of the enriched entity data and the activity location identifier of each activity record that identifies a first channel;
• determining, by the at least one processor, the at least one entity associated with the enriched first channel activity records based on the activity entity identifier of each enriched activity record;
• determining, by the at least one processor, a first channel opening date associated with each first channel based on the date indicator of each enriched activity record;
• determining, by the at least one processor, enriched pre-first channel activity records of the plurality of enriched activity records based on:
• i) the activity location identifier of each activity record associated with the at least one entity that identifies a second channel,
• ii) the date indication of each activity record associated with the at least one entity that identifies a pre-first channel opening date occurring before the first channel opening date, and
• iii) the at least one attribute of the enriched entity data;
• determining, by the at least one processor, enriched post-first channel activity records of the plurality of enriched activity records based on:
• i) the activity location identifier of each activity record associated with the at least one entity that identifies the second channel,
• ii) the date indication of each activity record associated with the at least one entity that identifies a post-first channel opening date occurring on or after the first channel opening date, and
• iii) the at least one attribute of the enriched entity data;
• determining, by the at least one processor, at least one pre-first channel metric through time associated with the enriched pre-first channel activity records of the plurality of enriched activity records based at least in part on the activity value of each enriched pre-first channel activity record of the enriched pre-first channel activity records;
• determining, by the at least one processor, at least one post-first channel metric associated with the enriched post-first channel activity records of the plurality of enriched activity records based at least in part on the activity value of each enriched post—first channel activity record of the enriched post-first channel activity records;
• determining, by the at least one processor, a combination of channel attributes for the second channel based at least in part on the enriched post-first channel activity records;
• utilizing, by the at least one processor, an optimizer to train a plurality of cannibalization forecasting neural network parameters of at least one cannibalization forecasting neural network so as to enable improved predictions of at least one predicted channel metric of future activities upon creating a new channel, wherein the training comprises:
• generating a predicted post-first channel activity metric based at least in part on:
• the at least one attribute of the at least one entity,
• the at least one pre-first channel activity metric, and
• the combination of channel attributes for the second channel;
• wherein the predicted post-first channel activity metric represents the at least one predicted channel metric as a function of time;
• determining, using a loss function, an error between:
• the predicted post-first channel activity metric, and
• the at least one post-first channel activity metric;
• and
• wherein the plurality of cannibalization forecasting neural network parameters comprise a plurality of node configurations having learned weights to correlate an input of the activity data of the plurality of activity records and the at least one attribute of the enriched entity data to an output of the predicted post-first channel activity metric for the second time period for the at least one entity;
• updating, using at least one optimization algorithm, the plurality of cannibalization forecasting neural network parameters of the cannibalization forecasting neural network based at least in part on the error by backpropagating the error to the plurality of cannibalization forecasting neural network parameters so as to train the cannibalization forecasting neural network to forecast the at least one predicted channel metric of the future activities based on the combination of channel attributes so as to update the learned weights of the plurality of node configurations to more accurately correlate an input of the activity data of the plurality of activity records and the at least one attribute of the enriched entity data to an output of the at least one forecasted activity metric for the at least one entity;
• and
• deploying, by the at least one processor, the at least one cannibalization forecasting neural network for access by at least one user computing device associated with at least one user to forecast the at least one channel metric of the future activities by the new channel based on the activity value and the activity location.

A system comprising:

• at least one processor configured to implement software instructions causing the at least one processor to perform steps to:
• receive an activity record history comprising a plurality of activity records across at least one activity channel of at least one medium of activity;
• wherein each activity record of the plurality of activity records comprise:
• i) an activity entity identifier associated with an entity of the plurality of entities,
• ii) a date indicator identifying a date associated with each activity record,
• iii) an activity location identifier associated with an entity location associated with each activity record, and
• iv) an activity value;
• access, via an external service, enriched entity data associated with the at least one entity;
• wherein the enriched entity data comprises at least one attribute of the at least one entity to add to each activity record;
• wherein the at least one attribute is associated with at least one physical location at which the at least one entity has a presence;
• format, based on the at least one attribute of the enriched data, the enriched entity data and the plurality of activity records to standardize the enriched entity data and the plurality of activity records into a common format;
• generate a plurality of enriched activity records comprising the plurality of activity records modified with the enriched entity data in the common format;
• determine at least one channel metric through time associated with the plurality of enriched activity records in a first time period based at least in part on the activity value of each activity record;
• receive, from at least one computing device associated with a user, a user selection via a dynamic channel forecasting user interface, the user selection comprising at least one new activity channel attribute associated with a future new activity channel different form the at least one activity channel;
• wherein the dynamic channel forecasting user interface comprises:
• at least one user selectable filter element that enables the user to perform the user selection, and
• at least one dynamic visualization for depicting the at least one channel metric through time;
• input the plurality of enriched activity records into at least one cannibalization forecasting model to output a prediction of at least one forecasted activity metric for a second time period after the first time period based on an input of the at least one channel metric and the at least one new activity channel attribute;
• wherein the at least one forecasted activity metric represents at least one predicted channel metric as a function of time for a particular combination of channel attributes;
• wherein the at least one cannibalization forecasting model comprises a plurality of cannibalization forecasting neural network parameters of a cannibalization forecasting neural network;
• wherein the plurality of cannibalization forecasting neural network parameters is configured to output the at least one forecasted activity metric based on an input of the at least one channel metric;
• wherein the plurality of cannibalization forecasting neural network parameters comprise a plurality of node configurations having learned weights trained on historical enriched activity records to correlate an input of the activity data of the plurality of activity records and the at least one attribute of the enriched entity data to an output of the at least one forecasted activity metric for the second time period for the at least one entity;
• causing to display, one the at least one computing device, at least one update to the at least one dynamic visualization in response to the at least one forecasted activity metric;
• wherein the at least one update depicts at least one change to the at least one channel metric through time based at least in part on the at least one forecasted activity metric associated with the user selection of at least one new activity channel attribute;
• receive at least one subsequent activity record associated with the at least one activity channel of the at least one medium of activity;
• access, via the external service, the enriched entity data associated with the at least one entity;
• format the enriched entity data and the at least one subsequent activity record to standardize the enriched entity data and the at least one subsequent activity record to the common format;
• generate at least one subsequent enriched activity record comprising the at least one subsequent activity record modified with the enriched entity data in the common format;
• determine at least one subsequent channel metric associated with the at least one subsequent enriched activity record based at least in part on at least one subsequent activity value of the at least one subsequent activity record;
• and
• utilize an optimizer to re-train the plurality of cannibalization forecasting neural network parameters of the cannibalization forecasting neural network so as to improve a fit of predictions by the cannibalization forecasting neural network, wherein the re-training comprises:
• determining an error between the at least one subsequent channel metric and the at least one forecasted activity metric based at least in part on a loss function, and
• updating the plurality of cannibalization forecasting neural network parameters by backpropagating the error to the plurality of cannibalization forecasting neural network parameters using at least one optimization algorithm so as to update the learned weights of the plurality of node configurations to more accurately correlate an input of the activity data of the plurality of activity records and the at least one attribute of the enriched entity data to an output of the at least one forecasted activity metric for the at least one entity.