Machine learning with partial inversion

Patent metadata

Field	Value
Publication number	US-10339441-B2
Application number	US-201715850395-A
Country	US
Kind code	B2
Filing date	Dec 21, 2017
Priority date	Oct 17, 2016
Publication date	Jul 2, 2019
Grant date	Jul 2, 2019

Abstract

An example embodiment may involve a machine learning model representing relationships between a dependent variable and a plurality of n independent variables. The dependent variable may be a function of the n independent variables, where the n independent variables are measurable characteristics of computing devices, and where the dependent variable is a predicted behavior of the computing devices. The embodiment may also involve obtaining a target value of the dependent variable, and separating the n independent variables into n−1 independent variables with fixed values and a particular independent variable with an unfixed value. The embodiment may also involve performing a partial inversion of the function to produce a value of the particular independent variable such that, when the function is applied to the value of the particular independent variable and the n−1 independent variables with fixed values, the dependent variable is within a pre-defined range of the target value.

First claim

What is claimed is:

1. A system comprising:
a machine learning model representing relationships between a dependent variable and a plurality of independent variables, including a first independent variable and a second independent variable, wherein the dependent variable is a function of the plurality of independent variables, wherein the first independent variable comprises an identifying characteristic of users of one or more computer-based services offered by a managed network, wherein the second independent variable comprises a type of communication directed to the users of the one or more computer-based services offered by a managed network, and wherein the dependent variable is a likelihood of a respective user performing a specific action in response to the communication when interacting with the one or more computer-based services; and
a computing device including a processor and memory, wherein execution, by the processor, of program instructions stored in the memory causes the computing device to perform operations comprising:
training the machine learning model to output the function of the dependent variable using a training data set, wherein the training data set comprises historical data for the plurality of independent variables and the dependent variable;
obtaining a target value of the dependent variable;
fixing the first independent variable at a fixed value;
performing a partial inversion of the function, using the fixed value of the first independent value and the target value of the dependent variable as inputs, to solve for the second independent variable to obtain a partially inverted function to produce one or more values of the second independent variable such that, when the partially inverted function is applied to the first independent variable having the fixed value and any of the one or more values of the second independent variable, the dependent variable is within a pre-defined range of the target value of the dependent variable, wherein solving for the second independent variable to obtain the partially inverted function comprises:
determining a plurality of candidate values of the second independent variable:
determining, for each of the plurality of candidate values of the second independent variable, a respective output value of the dependent variable when the fixed value of the first independent variable and the candidate value of the second independent variable are applied as inputs to the function and comparing the respective output value of the dependent variable to the target value of the dependent variable; and
outputting one or more of the plurality of candidate values of the second independent variable that, when input to the function with the fixed value of the first independent variable, result in the respective output value of the dependent variable being within the pre-defined range of the target value of the dependent variable;

generating a notification based on the produced one or more values of the second independent variable; and
communicating the notification to one or more of the users of the one or more computer-based services offered by a managed network.

2. The system of claim 1 , wherein comparing the expected value of the dependent variable to the target value of the dependent variable comprises:
determining that the respective output value that is within the pre-defined range of the target value of the dependent variable.

3. The system of claim 1 , wherein determining the plurality of candidate values of the second independent variable comprises:
determining the plurality of candidate values in accordance with a binary search over a range of the candidate values.

4. The system of claim 1 , wherein determining the plurality of candidate values of the second independent variable comprises:
randomly selecting the plurality of candidate values.

5. The system of claim 1 , wherein performing the partial inversion of the function comprises:
using a non-linear solver software module to determine the one or more values of the second independent variable.

6. The system of claim 1 , wherein the operations further comprise:
monitoring the measurable characteristics of the users of the one or more computer-based services corresponding to the second independent variable.

7. The system of claim 6 , wherein the operations further comprise:
determining that the measurable characteristics of the users of the one or more computer-based services is within a threshold range of at least one of the one or more values of the second independent variable.

8. The system of claim 1 , wherein communicating the notification to one or more of the users of the one or more computer-based services offered by a managed network comprises:
transmitting the notification by way of email, voice call, or text message.

9. The system of claim 1 , wherein the specific action is a desirable behavior of the users of the one or more computer-based services.

10. The system of claim 1 , wherein the operations further comprise:
performing a second partial inversion of the function to produce one or more values of the first independent variable such that, when the partially inverted function is applied to any of the one or more values of the first independent variable and the second independent variable with a second fixed value, the dependent variable is within the pre-defined range of the target value of the dependent variable.

11. The system of claim 1 , wherein the specific action comprises performing a backup of a computer.

12. The system of claim 1 , wherein the type of communication comprises an email, a text message, or a phone call.

13. The system of claim 1 , wherein the machine learning model comprises a linear regression-based model, a polynomial regression-based model, a decision tree, a random forest of decision trees, or a neural network.

14. The system of claim 1 , wherein the managed network is managed by a single entity.

15. The system of claim 1 , wherein the first independent variable comprises a user's department or the user's rank within an organization.

16. A method comprising:
obtaining, by a computing system, a machine learning model representing relationships between a dependent variable and a plurality of independent variables, including a first independent variable and a second independent variable, wherein the dependent variable is a function of the plurality of independent variables, wherein the first independent variable comprises an identifying characteristic of users of one or more computer-based services offered by a managed network, wherein the second independent variable comprises a type of communication directed to the users of the one or more computer-based services offered by a managed network, and wherein the dependent variable is a likelihood of a respective user performing a specific action in response to the communication when interacting with the one or more computer-based services;
training the machine learning model to output the function of the dependent variable using a training data set, wherein the training data set comprises historical data for the plurality of independent variables and the dependent variable;
obtaining, by the computing system, a target value of the dependent variable;
fixing, by the computing system, the first independent variable at a fixed value;
performing, by the computing system, a partial inversion of the function, using the fixed value of the first independent value and the target value of the dependent variable as inputs, to solve for the second independent variable to obtain a partiall

Assignees

• Servicenow Inc

Inventors

• Jayaraman Baskar
• Thakur Aniruddha
• Govindarajan Kannan

Classifications

• G06F17/11

  for solving equations {, e.g. nonlinear equations, general mathematical optimization problems (optimization specially adapted for a specific administrative, business or logistic context G06Q10/04)} · CPC title

• G06F11/3684

  for test design, e.g. generating new test cases · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06F17/10

  Complex mathematical operations {(function generation by table look-up G06F1/03; evaluation of elementary functions by calculation G06F7/544)} · CPC title

• H04L41/16 Primary

  using machine learning or artificial intelligence · CPC title