Utilizing one hash permutation and populated-value-slot-based densification for generating audience segment trait recommendations

In a digital medium environment for distributing digital content to client devices over a computer network, a computer-implemented method for generating distribution segment trait recommendations, comprising:

• training a distribution segment trait recommendation model by:
• utilizing a one permutation hashing model to generate a plurality of sketch vectors from training samples, wherein the training samples comprise a plurality of distribution segment traits and each sketch vector corresponds to a distribution segment trait from the training samples;
• generating a plurality of densified sketch vectors from the plurality of sketch vectors by populating unpopulated value slots of the plurality of sketch vectors based on analyzing populated value slots of the plurality of sketch vectors;
• and
• utilizing a singular value decomposition model to generate a plurality of trait embeddings based on the plurality of densified sketch vectors;
• and
• identifying an input distribution segment trait;
• and
• utilizing the distribution segment trait recommendation model to generate a distribution segment trait recommendation by comparing a first trait embedding corresponding to the input distribution segment trait with a second trait embedding from the plurality of trait embeddings.

A system comprising:

• at least one processor;
• and
• at least one non-transitory computer-readable medium storing instructions that, when executed by the at least one processor, cause the system to:
• train a distribution segment trait recommendation model by:
• utilizing a one permutation hashing model to generate a plurality of sketch vectors from training samples, wherein the training samples comprise a plurality of distribution segment traits and each sketch vector corresponds to a distribution segment trait from the training samples;
• generating a plurality of densified sketch vectors from the plurality of sketch vectors by populating unpopulated value slots of the plurality of sketch vectors based on analyzing populated value slots of the plurality of sketch vectors;
• and
• utilizing a singular value decomposition model to generate a plurality of trait embeddings based on the plurality of densified sketch vectors;
• and
• identify an input distribution segment trait;
• and
• utilize the distribution segment trait recommendation model to generate a distribution segment trait recommendation by comparing a first trait embedding corresponding to the input distribution segment trait with a second trait embedding from the plurality of trait embeddings.