AI for QA Tests in Telecom

Qualitest provides telecom companies with a sophisticated predictions model solution that optimizes QA tests. A mobile carrier, as part of a leading telecom company, used Qualitest’s solution to predict which set of tests are more likely to have faults along with estimating the time needed for fault resolution. After using Qualitest’s solution, the company found that they can efficiently direct employees and resources to fix these faults before they occur.

Use Case Steps

  • The company data analysts were guided by the Qualitest team to plan and test data preprocessing. The team explored the data and created a unified panel of exploratory variables.
  • Our running on-premise Qualitest solution, with more than 40 machine learning algorithms on offer, evaluated more than 60k tests to find the best variable combinations and prediction models.
  • Our solution delivered an accurate model that proved to be stable over time and transparent ready for deployment in production.
  • Our team implemented the Qualitest solution based on a source listener mechanism that allowed for easy connectivity with the production tests database. The model ran constantly on operational testing data to provide predictions.

The Problem

QA software tests were based on human gut feeling of weak development points. There was no proactive process for discovering and preventing faults on different platforms in place.

The Mission

Qualitest’s machine learning technology sought to turn the tests and QA process into a foreseeing process. Our solution predicted which QA tests were going to result with a high number of faults along with identifying the reasons for such failures.

The Data Citizen Solution

Qualitest is guided by the Data Citizen vision; enabling data analysts with no statistical or math expertise to create accurate prediction models easily and efficiently. Qualitest provided the technology that runs 40 machine learning algorithms on the targeted QA software tests database. The solution included an automatic fine-tuning mechanism that found the best exploratory variables combination as well as selected the best suitable algorithm for the task in question.

The solution’s ability to crunch data and create prediction models automatically with no need to write a single line of code proved to be highly efficient. A stable and transparent prediction model was ready to be deployed in 30 working days.

Implementation

Qualitest’s solution provided a variety of methods for implementing the chosen model in the production environment. The implementation process with database connectivity was performed by the client’s IT department within 3 working days.

Project Results

The best prediction model was implemented in the production environment. Based on the software QA tests data, the company can predict which QA test is going to result in a high rate of issues. The company targets the faults and directs its employees to resolve them.

Endnotes

The Qualitest machine learning software tool for non-Data Scientist personnel delivered effective results. The solution proved to be accurate, efficient and powerful. The solution decreased costs and improved service by detecting and predicting QA and testing faults ahead of time.