Machine Learning Unit Test Generation System

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The generation of unit tests for software code is a tedious and time-consuming process, necessitating a mechanism for automated testing to improve efficiency in software development.

Innovation Solution

A method using machine learning in conjunction with existing unit tests to automatically generate new tests, where a processor accesses existing unit tests, determines the intended objective, selects relevant tests, and generates new tests using a machine learning algorithm, potentially with a recurrent neural network, and validates the tests using a PIT testing protocol.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If automated test generation using machine learning is implemented, then productivity is improved, but device complexity increases

Engineering Contradiction:

Improvetest generation efficiencyVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

A machine learning model is introduced as an intermediary between existing unit tests and new test generation. The model learns patterns from existing tests and applies them to generate new tests automatically, reducing the need for manual test creation while managing system complexity through a dedicated AI component.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system copies patterns and structures from existing unit tests to generate new tests. By analyzing the characteristics of existing tests and replicating their effective patterns, the system can automatically produce new tests that maintain quality standards without requiring manual intervention for each test case.

Inventive Principle:

Principle #26Copying

2Reliability

If manual test generation is performed, then test coverage and reliability are improved, but loss of time increases

Engineering Contradiction:

Improvetest validityVSAvoidtest generation time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The machine learning model performs preliminary learning from existing unit tests before new test generation is needed. By pre-processing and extracting patterns from existing tests in advance, the system is prepared to quickly generate new tests that maintain high validity and coverage without time-consuming manual analysis.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system incorporates feedback mechanisms where the validity and performance of generated tests are evaluated, and this information is fed back to improve the machine learning model. This continuous learning process ensures that automated tests maintain high reliability while reducing generation time over the system's operational lifecycle.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11074163B2Method and system for generating unit tests using machine learning

Publication Date: 2021.07.27 JPMORGAN CHASE BANK NA

AI summary

A method for generating a new test for a set of software code is provided. The method includes accessing a plurality of existing unit tests; implementing a machine learning algorithm; determining intended objectives of the set of software code; selecting a subset of the plurality of existing unit tests based on the determined objectives and an output of the machine learning algorithm; and using the selected unit tests to automatically generate the new test.