Data stream analytics

An apparatus for use in data stream analytics, the apparatus comprising:

• a hash generator to generate a plurality of sets of hashes, each set of hashes generated for a sample data object of sample data objects of a data stream, the data stream originating from a plurality of streaming, each of which transmit sample data objects over a network independent of the other streaming devices, and the data stream including different data types, each of the set of hashes generated based on a real-valued vector of a corresponding sample data object, wherein the hash generator generates the sets of hashes by determining a plurality of keys for the real-valued vector of the corresponding sample data object and using partitioning to select a top set of the keys as the hashes;
• an indexer to index each set of hashes with an object identifier of the corresponding sample data object in an index of a database;
• a list generator to generate, for each sample data object, a list of previously received data objects from the index based on the sets of hashes, the index comprising sets of hashes of real-valued vectors corresponding to the previously received data objects and object identifiers of the previously received data objects, the list generator identifying data objects in the index having a set of hashes that share the most number of hashes generated by the hash generator;
• and
• a stream analytics manager to update, for each sample data object, a data structure, corresponding to a respective sample data object, based on the list of previously received data objects, for the respective sample data object, and to apply machine learning to the data structure to provide real-time data stream analytics based on the respective sample data object.

A non-transitory machine readable storage medium comprising instructions that, when executed, cause a machine to at least:

• receive real-valued vectors, each corresponding to a data object of data objects of data streams, the data streams originating from a plurality of streaming devices, each of which transmit sample data objects over a network independent of the other streaming devices, and the data streams including different data types;
• and
• provide real-time stream analytics corresponding to the data objects of the data streams by:
• for each newly received real-valued vector,
• generate a set of hashes based on the newly received real-valued vector, wherein the set of hashes are generated by determining a plurality of keys for the newly received real-valued vector and using partitioning to select a top set of the keys as the hashes;
• index the set of hashes in an index of a database with an object identifier corresponding to a data object of the newly received real-valued vector;
• generate a list of nearest neighbor data objects based an amount of overlap between the set of hashes of the newly received real-valued vector and each of sets of hashes corresponding to previously received real-valued vectors for corresponding data objects;
• and
• update data structure, corresponding to the data object of the newly received real-valued vector, the corresponding data structure to provide real-time stream analytics associated with the data object of the newly received real-valued vector.