Prediction of cacheable queries

Patent metadata

Field	Value
Publication number	US-2025315431-A1
Application number	US-202519245897-A
Country	US
Kind code	A1
Filing date	Jun 23, 2025
Priority date	Nov 27, 2023
Publication date	Oct 9, 2025
Grant date	—

Abstract

Predicting cacheable queries is provided. For example, a system can one or more processors, coupled with memory, to generate a plurality of predicted requests based on one or more requests received from a client device. The system can compute a metric associated with retrieving data for a respective predicted request from a database. The system can select a subset of predicted requests from the plurality of predicted requests based on the metric corresponding to each predicted request and a threshold metric. The system can generate labels classifying the subset of predicted requests. The system can store, using the labels, data from the database for the subset of predicted requests in a cache store. The system can transmit, responsive to receiving a client request matching a predicted request of the subset of predicted requests, corresponding data from the cache store.

First claim

What is claimed is:

1 . A system comprising:
one or more processors, coupled with memory, to: generate, using a machine learning model, a plurality of predicted requests based on one or more requests received from a client device; compute, for each predicted request of the plurality of predicted requests, a metric associated with retrieving data for a respective predicted request from a database; select a subset of predicted requests from the plurality of predicted requests based on the metric corresponding to each predicted request and a threshold metric; generate, using the machine learning model, labels classifying the subset of predicted requests; store, using the labels as cache keys, data from the database for the subset of predicted requests in a cache store; and transmit, responsive to receiving a client request matching a predicted request of the subset of predicted requests, corresponding data from the cache store.

2 . The system of claim 1 , wherein the one or more processors further:
generate a plurality of predicted request identifiers associated with the subset of predicted requests; and construct, using the machine learning model and based on the plurality of predicted request identifiers, the labels comprising classifications for the subset of predicted requests.

3 . The system of claim 1 , wherein the one or more processors:
receive one or more identifiers and key-value pairs associated with the one or more requests; and generate, using the machine learning model, the plurality of predicted requests based on the one or more identifiers and key-value pairs.

4 . The system of claim 1 , wherein the one or more requests comprise at least a first request, and wherein the one or more processors further:
predict, responsive to receiving a second request, using the machine learning model, a second plurality of predicted requests based on the first request and the second request; identify, using the machine learning model and based on a comparison with the threshold metric, a second subset of predicted requests from the second plurality of predicted requests indicative of one or more subsequent requests; and transmit, responsive to receipt of a subsequent request that matches one of a group of predicted requests comprising the second subset of predicted requests and one or more of the subset of predicted requests that are below a cache store threshold, a cache value from the cache store that corresponds to the subsequent request.

5 . The system of claim 1 , wherein the metric associated with retrieving data for the respective predicted request comprises at least one of a latency of the plurality of predicted requests and a frequency of the plurality of predicted requests.

6 . The system of claim 1 , wherein to generate the plurality of predicted requests, the one or more processors further:
generate a hash from data associated with each of the one or more requests based on a hash function; concatenate each hash with a value derived from the data associated with each of the one or more requests to generate input feature encoding; and generate the plurality of predicted requests by identifying requests associated with the input feature encoding.

7 . The system of claim 1 , wherein the one or more processors further:
configure, in the cache store, the labels as the cache keys and the data from the database as cache values for the subset of predicted requests.

8 . The system of claim 1 , wherein the one or more processors further:
generate a ranking of the plurality of predicted requests based on the metric corresponding to each predicted request; and select the subset of predicted requests based on the ranking.

9 . The system of claim 1 , wherein the one or more processors further:
determine the metric for each of the plurality of predicted requests by calculating a product of a latency associated with each of the plurality of predicted requests and a frequency of receiving each of the plurality of predicted requests.

10 . The system of claim 1 , wherein the one or more processors further:
train the machine learning model based on historical data associated with a plurality of profiles received from a plurality of entities; receive an entity identifier and a profile associated with each of the one or more requests; and use the entity identifier and the profile associated with each of the one or more requests to generate the plurality of predicted requests.

11 . A method comprising:
generating, by one or more processors, coupled with memory, using a machine learning model, a plurality of predicted requests based on one or more requests received from a client device; computing, by the one or more processors, for each predicted request of the plurality of predicted requests, a metric associated with retrieving data for a respective predicted request from a database; selecting, by the one or more processors, a subset of predicted requests from the plurality of predicted requests based on the metric corresponding to each predicted request and a threshold metric; generating, by the one or more processors, using the machine learning model, labels classifying the subset of predicted requests; storing, by the one or more processors, using the labels as cache keys, data from the database for the subset of predicted requests in a cache store; and transmitting, by the one or more processors, responsive to receiving a client request matching a predicted request of the subset of predicted requests, corresponding data from the cache store.

12 . The method of claim 11 , further comprising:
generating, by the one or more processors, a plurality of predicted request identifiers associated with the subset of predicted requests; and constructing, by the one or more processors, using the machine learning model and based on the plurality of predicted request identifiers, the labels comprising classifications for the subset of predicted requests.

13 . The method of claim 11 , further comprising:
receiving, by the one or more processors, one or more identifiers and key-value pairs associated with the one or more requests; and generating, by the one or more processors, using the machine learning model, the plurality of predicted requests based on the one or more identifiers and key-value pairs.

14 . The method of claim 11 , wherein the one or more requests comprise at least a first request, and the method further comprising:
predicting, by the one or more processors, responsive to receiving a second request, using the machine learning model, a second plurality of predicted requests based on the first request and the second request; identifying, by the one or more processors, using the machine learning model and based on a comparison with the threshold metric, a second subset of predicted requests from the second plurality of predicted requests indicative of one or more subsequent requests; and transmitting, by the one or more processors, responsive to receipt of a subsequent request that matches one of a group of predicted requests comprising the second subset of predicted requests and one or more of the subset of predicted requests that are below a cache store threshold, a cache value from the cache store that corresponds to the subsequent request.

15 . The method of claim 11 , wherein the metric associated with retrieving data for the respective predicted request comprises at least one of a latency of the plurality of predicted requests and a frequency of the plurality of predicted requests.

16 . The method of claim 11 , wherei

Assignees

• Adp Inc

Inventors

• Roy Kuntal
• Liao Xinpeng
• Jain Ruchi Jinendra

Classifications

• G06N20/00

  Machine learning · CPC title

• G06F16/3349

  Reuse of stored results of previous queries · CPC title

• G06F16/2453 Primary

  Query optimisation · CPC title

• G06F16/24552 Primary

  Database cache management · CPC title