Item transaction modeling engine systems and methods

A method for item transactions comprising:

• receiving, from a structured transaction data source, structured transaction data indicative of values of an item over a time interval of interest;
• generating, based on the structured transaction data, a historical structured transaction dataset corresponding to the time interval of interest, the historical structured transaction dataset comprising timeseries data comprising a value for the item for each of a plurality of discrete points in time across the time interval of interest;
• determining, based on the historical structured dataset, a set of historical directional values for the time interval of interest, the set of historical directional values comprising, for each of the plurality of discrete points in time, a directional value that is indicative of a trend of the value of the item for a time interval that extends from a given leading time before the discrete point in time to a given trailing time after the discrete point in time;
• receiving, from an unstructured transaction data source, unstructured transaction data comprising datasets published over the time interval of interest;
• generating, based on the unstructured transaction data, a historical unstructured transaction dataset corresponding to the time interval of interest, the historical unstructured transaction dataset comprising datasets published over the time interval of interest and associated with one or more of the discrete points in time, the datasets being associated with respective ones of the discrete points in time;
• determining numerical representations of the historical unstructured transaction dataset, wherein determining the numerical representations comprises:
• for each dataset of the datasets, vectorizing content of the dataset to generate a dataset vector representative of the content of the dataset;
• and
• for each of the respective ones of the discrete points in time,
• identifying a set of dataset vectors corresponding to one or more datasets associated with the discrete point in time;
• and
• generating, based on the set of dataset vectors, a point-in-time vector representation for the discrete point in time;
• determining, based on the set of historical directional values and the numerical representations of the historical unstructured transaction datasets, a transaction model configured to determine a predicted direction of value for the item based on current structured transaction data and a current numerical representation generated from current unstructured transaction data, the determining comprising training the transaction model using, for the respective ones of the discrete points in time, the point-in-time vector representation for the discrete point in time and a corresponding directional value of the set of historical directional values;
• obtaining a current transaction dataset comprising:
• a current structured transaction dataset associated with a given point in time;
• and
• a current unstructured transaction dataset associated with the given point in time, the current unstructured transaction dataset comprising one or more current datasets associated with the given point in time;
• generating a current numerical representation of the current unstructured transaction dataset based on vectorization of content of the one or more current datasets;
• and
• determining, based on application of the current structured transaction dataset and the current numerical representation of the current unstructured transaction dataset to the transaction model, a predicted direction of value for the item for the given point in time.

A non-transitory computer readable storage medium comprising program instructions stored thereon that are executable by a processor to perform the following operations for item transactions:

• receiving, from a structured transaction data source, structured transaction data indicative of values of an item over a time interval of interest;
• generating, based on the structured transaction data, a historical structured transaction dataset corresponding to the time interval of interest, the historical structured transaction dataset comprising timeseries data comprising a value for the item for each of a plurality of discrete points in time across the time interval of interest;
• determining, based on the historical structured dataset, a set of historical directional values for the time interval of interest, the set of historical directional values comprising, for each of the plurality of discrete points in time, a directional value that is indicative of a trend of the value of the item for a time interval that extends from a given leading time before the discrete point in time to a given trailing time after the discrete point in time;
• receiving, from an unstructured transaction data source, unstructured transaction data comprising datasets published over the time interval of interest;
• generating, based on the unstructured transaction data, a historical unstructured transaction dataset corresponding to the time interval of interest, the historical unstructured transaction dataset comprising datasets published over the time interval of interest and associated with one or more of the discrete points in time, the datasets being associated with respective ones of the discrete points in time;
• determining numerical representations of the historical unstructured transaction dataset, wherein determining the numerical representations comprises:
• for each dataset of the datasets, vectorizing content of the dataset to generate a dataset vector representative of the content of the dataset;
• and
• for each of the respective ones of the discrete points in time:
• identifying a set of dataset vectors corresponding to one or more datasets associated with the discrete point in time;
• and
• generating, based on the set of dataset vectors, a point-in-time vector representation for the discrete point in time;
• determining, based on the set of historical directional values and the numerical representations of the historical unstructured transaction datasets, a transaction model configured to determine a predicted direction of value for the item based on current structured transaction data and a current numerical representation generated from current unstructured transaction data, the determining comprising training the transaction model using, for the respective ones of the discrete points in time, the point-in-time vector representation for the discrete point in time and a corresponding directional value of the set of historical directional values;
• obtaining a current transaction dataset comprising:
• a current structured transaction dataset associated with a given point in time;
• and
• a current unstructured transaction dataset associated with the given point in time, the current unstructured transaction dataset comprising one or more current datasets associated with the given point in time;
• generating a current numerical representation of the current unstructured transaction dataset based on vectorization of content of the one or more current datasets;
• and
• determining, based on application of the current structured transaction dataset and the current numerical representation of the current unstructured transaction dataset to the transaction model, a predicted direction of value for the item for the given point in time.

A method for item transactions comprising:

• obtaining, from a structured transaction data source, structured transaction data indicative of values of an item over a time interval;
• generating, based on the structured transaction data, a historical structured transaction dataset corresponding to the time interval, the historical structured transaction dataset comprising timeseries data comprising values of the item for a plurality of discrete points in time across the time interval;
• determining, based on the historical structured dataset, a set of historical directional values for the time interval, the set of historical directional values comprising directional values corresponding to respective ones of the plurality of discrete points in time;
• receiving, from an unstructured transaction data source, unstructured transaction data comprising datasets published over the time interval;
• generating, based on the unstructured transaction data, a historical unstructured transaction dataset corresponding to the time interval, the historical unstructured transaction dataset comprising the datasets published over the time interval, the datasets being associated with respective ones of the discrete points in time;
• determining numerical representations of the historical unstructured transaction dataset, wherein determining the numerical representations comprises:
• for each dataset of the datasets, vectorizing content of the dataset to generate a dataset vector representative of the content of the dataset;
• and
• for each discrete point in time of the respective ones of the discrete points in time:
• identifying a set of dataset vectors corresponding to one or more datasets associated with the discrete point in time;
• and
• generating, based on the set of dataset vectors, a point-in-time vector representation for the discrete point in time;
• determining, based on the set of historical directional values and the numerical representations of the historical unstructured transaction datasets, a transaction model configured to determine a predicted direction of value for the item based on current structured transaction data and a current numerical representation generated from current unstructured transaction data the determining comprising training the transaction model using, for each discrete point in time of the respective ones of the discrete points in time, the point-in-time vector representation for the discrete point in time and a corresponding directional value of the set of historical directional values:
• generating the current numerical representation of the current unstructured transaction data, the current unstructured transaction data comprising one or more current datasets associated with a given point in time, based on vectorization of content of the one or more current datasets;
• and
• determining, based on application of current structured transaction data and the current numerical representation of the current unstructured transaction data to the transaction model, a predicted direction of value for the item for the given point in time.