Data learning and analytics apparatuses, methods and systems
The DATA LEARNING AND ANALYTICS APPARATUSES, METHODS AND SYSTEMS (“DLA”) provides a data learning platform that analyzes and incorporates data knowledge from new data event updates (e.g., in a real-time stream, in a batch, etc.). In one implementation, the DLA provides a real-time data mining mechanism that accommodates an ever increasing data load instantaneously with reduced complexity, e.g., without re-building the data analytical model in its entirety whenever new data is received. In this way, the DLA provides an incremental learning mechanism for data analytics and thus the latency for real-time data assessment is reduced and the data processing efficiency is improved.
1. A processor-implemented, real-time learning machine component, streaming data analytical structure building method, comprising:
obtaining an analytical data structure building request including a target analytical metric from a requestor, wherein the target analytical metric includes any of: ad click through rate (CTR), ad cost per mile (CPM), ad cost per click (CPC), ad cost per action (CPA), wherein the analytical data structure building request includes the target analytical metric within the analytical data structure building request;
determining an analytical structure type requiring a plurality of coefficients based on the target analytical metric;
obtaining an incoming training stream of data variables associated with one or more real-time events, each data variable belonging to a data attribute that describes the one or more real-time events;
formatting the incoming training stream of data variables in compliance with a pre-defined data format of a real-time learning component;
determining an encoder type based on the target analytical metric;
encoding, via the determined encoder based on the target analytical metric, at least one training data variables into numeric representations by mapping the formatted incoming training stream of data variables to the target analytical metric;
compositing the training data variables including the encoded at least one training data variables creating an additive and reusable basic element data structure indicative of joint occurrence counts of pairs of data attributes based on the composited training data variables;
calculating the plurality of coefficients required by the analytical structure type based on cell values of the additive and reusable basic element data structure; and
establishing an analytical data structure with the calculated plurality of coefficients and the additive and reusable basic element data structure;
in which establishing the analytical data structure includes:
querying for a reusable analytical data structure matching the target analytical metric, and upon determination that the reusable analytical data structure results from the query, updating the reusable analytical data structure with the composited training data variables, and
building a new analytical data structure with the composited training data variables based on the target analytic metrics upon determination that no reusable analytical data structure results from the query; and
providing the established analytical data structure to the requestor.
2. A real-time learning machine component and streaming data analytical structure building system, comprising:
means for obtaining an analytical data structure building request including a target analytical metric from a requestor, wherein the target analytical metric includes any of: ad click through rate (CTR), ad cost per mile (CPM), ad cost per click (CPC), ad cost per action (CPA)), wherein the analytical data structure building request includes the target analytical metric within the analytical data structure building request;
means for determining an analytical structure type requiring a plurality of coefficients based on the target analytical metric;
means for obtaining an incoming training stream of data variables associated with one or more real-time events,
each data variable belonging to a data attribute that describes the one or more real-time events;
means for formatting the incoming training stream of data variables in compliance with a pre-defined data format of a real-time learning component;
means for determining an encoder type based on the target analytical metric;
means for encoding, via the determined encoder based on the target analytical metric, at least one training data variables into numeric representations by mapping the formatted incoming training stream of data variables to the target analytical metric;
means for compositing the training data variables including the encoded at least one training data variables creating an additive and reusable basic element data structure indicative of joint occurrence counts of pairs of data attributes based on the composited training data variables;
means for calculating the plurality of coefficient required by the analytical structure type based on cell values of the additive and reusable basic element data structure; and
means for establishing an analytical data structure with the calculated plurality of coefficients and the additive and reusable basic element data structure;
in which means for establishing the analytical data structure includes:
means for querying for a reusable analytical data structure matching the target analytical metric, and upon determination that the reusable analytical data structure results from the query, updating the reusable analytical data structure with the composited training data variables, and
means for building a new analytical data structure with the composited training data variables based on the target analytic metrics upon determination that no reusable analytical data structure results from the query; and
means for providing the established analytical data structure to the requestor.
3. A real-time learning machine component and streaming data analytical structure building apparatus, comprising:
a processor; and
a memory disposed in communication with the processor and storing processor-issuable instructions to:
obtain an analytical data structure building request including a target analytical metric from a requestor, wherein the target analytical metric includes any of: ad click through rate (CTR), ad cost per mile (CPM), ad cost per click (CPC), ad cost per action (CPA)), wherein the analytical data structure building request includes the target analytical metric within the analytical data structure building request;
determine an analytical structure type requiring a plurality of coefficients based on the target analytical metric;
obtain an incoming training stream of data variables associated with one or more real-time events, each data variable belonging to a data attribute that describes the one or more real-time events;
format the incoming training stream of data variables in compliance with a pre-defined data format of a real-time learning component;
determine an encoder type based on the target analytical metric;
encode, via the determined encoder based on the target analytical metric, at least one training data variables into numeric representations by mapping the formatted incoming training stream of data variables to the target analytical metric;
composite the training data variables including the encoded at least one training data variables creating an additive and reusable basic element data structure indicative of joint occurrence counts of pairs of data attributes based on the composited training data variables;
calculate the plurality of coefficients required by the analytical structure type based on cell values of the additive and reusable basic element data structure, and
establish an analytical data structure with the calculated plurality of coefficients and the additive and reusable basic element data structure;
in which establishing the analytical data structure includes:
query for a reusable analytical data structure matching the target analytical metric, and upon determination that the reusable analytical data structure results from the query, updating the reusable analytical data structure with the composited training data variables, and
build a new analytical data structure with the composited training data variables based on the target analytic metrics upon determination that no reusable analytical data structure results from the query; and
provide the established analytical data structure to the requestor.
4. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , wherein the target analytical metric is identified by a user.
5. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , wherein the target analytical metric comprises any of a target click-through-rate, a target cost per click, a target cost per action, and a target cost per mile.
6. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , wherein the analytical data structure building request comprises a request to generate a numeric evaluative value reflecting an estimate of the target analytical metric.
7. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , wherein the analytical structure type requiring a plurality of coefficients comprises a linear regression formula.
8. The real-time learning machine component and streaming data analytical structure building apparatus of claim 7 , wherein the linear regression formula employs one or data attributes as independent regression variables.
9. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , wherein the incoming stream of data variables is received in any of the following manner: one-by-one, periodic, continuous, and in a batch.
10. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , wherein the encoding comprises mapping textual values of the one or more data variables to numeric values.
11. The real-time learning machine component and streaming data analytical structure building apparatus of claim 10 , wherein the numeric values are obtained based on a data attribute value associated with a real-time event,
said data attribute value indicative of the target analytical metric.
12. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , wherein the additive and reusable basic element data structure is an orthogonal matrix.
13. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , wherein each cell of the additive and reusable basic element data structure is calculated by taking a summation of a first order and a second order power series of two data variables.
14. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , wherein the additive and reusable basic element data structure is updated by calculating new power series of new pairs of data variables and adding the calculated new power series to each cell without re-building the basic element data structure.
15. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , is further configured to calculate a covariance data structure based on the additive and reusable basic element data structure.
16. The real-time learning machine component and streaming data analytical structure building apparatus of claim 15 , further configured to calculate a correlation data structure based on the covariance data structure.
17. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , further configured to:
obtain a request to update the analytical data structure with a new data variable;
retrieve the additive and reusable basic element data structure;
calculate a new joint occurrence count for the new data variable;
add the new joint occurrence count to the additive and reusable basic element data structure; and
obtain an updated basic element data structure without re-building the additive and reusable basic element data structure.
18. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , further configured to:
obtain a new data attribute;
retrieve the additive and reusable basic element data structure;
add a new column and a new row to the additive and reusable basic element data structure;
calculate new joint occurrence counts between existing data attributes and the new data attribute;
add the new joint occurrence counts to the new column and the new row; and
obtain an undated additive and reusable basic element data structure without re-building the additive and reusable basic element data structure.
19. The real-time learning machine component and streaming data analytical structure building apparatus of claim 3 , further configured to:
retrieve a first analytical data structure built upon a first additive and reusable basic element data structure;
retrieve a second analytical data structure built upon a second additive and reusable basic element data structure;
create a new basic element data structure by adding the first additive and reusable basic element data structure to the second basic element data structure; and
calculate new coefficients for a new analytical data structure based on the new additive and reusable basic element data structure.
20. A real-time learning machine component and streaming data analytical structure building non-transitory storage processor-readable medium storing processor-executable instructions executable by a processor to:
obtain an analytical data structure building request including a target analytical metric from a requestor, wherein the target analytical metric includes any of: ad click through rate (CTR), ad cost per mile (CPM), ad cost per click (CPC), ad cost per action (CPA)), wherein the analytical data structure building request includes the target analytical metric within the analytical data structure building request;
determine an analytical structure type requiring a plurality of coefficients based on the target analytical metric;
obtain an incoming training stream of data variables associated with one or more real-time events, each data variable belonging to a data attribute that describes the one or more real-time events;
format the incoming training stream of data variables in compliance with a pre-defined data format of a real-time learning component;
determine an encoder type based on the target analytical metric;
encode, via the determined encoder based on the target analytical metric, at least one training data variables into numeric representations by mapping the formatted incoming training stream of data variables to the target analytical metric;
composite the training data variables including the encoded at least one training data variables creating an additive and reusable basic element data structure indicative of joint occurrence counts of pairs of data attributes based on the composited training data variables;
calculate the plurality of coefficients required by the analytical structure type based on cell values of the additive and reusable basic element data structure; and
establish an analytical data structure with the calculated plurality of coefficients and the additive and reusable basic element data structure;
in which establishing the analytical data structure includes:
query for a reusable analytical data structure matching the target analytical metric, and upon determination that the reusable analytical data structure results from the query, updating the reusable analytical data structure with the composited training data variables, and
build a new analytical data structure with the composited training data variables based on the target analytic metrics upon determination that no reusable analytical data structure results from the query; and
provide the established analytical data structure to the requestor.