CYPRESS PERFORMANCE INSIGHTS PREDICTING UI TEST EXECUTION TIME USING COMPLEXITY METRICS

Abstract

This study presents a comparative analysis of three machine learning regression algorithms—Linear Regression (LR), Multi-layer Perceptron (MLP), and Gaussian Process Regression (GPR)—for predicting software execution time based on critical testrelated input parameters. The aim is to evaluate model performance in accurately estimating execution time using features such as Test Script Complexity, DOM Element Count, and Test Data Size (in KB). A robust evaluation was carried out using multiple metrics, including R² score, Explained Variance Score (EVS), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Squared Log Error (MSLE), and others. The results indicate that MLP outperforms the other models on test Cypress Performance Insights: Predicting UI Test Execution Time Using Complexity Metrics https://iaeme.com/home/journal/IJRCAIT 168 editor@iaeme.com data, showcasing high accuracy and generalization capability, making it the most suitable model for predictive automation scenarios.

Research Significance: With increasing demands for automation in software testing, accurately predicting test execution time has become crucial for optimizing resource allocation, managing testing pipelines, and improving CI/CD efficiency. Traditional heuristic-based estimations often fail to capture complex nonlinear dependencies between test attributes and execution performance. This research contributes to the field by applying and benchmarking advanced machine learning models to quantify these relationships, offering a data-driven framework for execution time prediction in test automation environments. Methodology: Algorithm Analysis The study utilizes three supervised learning regression algorithms: Linear Regression (LR): A baseline model to understand linear relationships.

Multi-layer Perceptron (MLP): A neural network-based model capable of capturing nonlinear patterns. Gaussian Process Regression (GPR): A probabilistic model that estimates uncertainty along with predictions. The input dataset includes the following features: Test Script Complexity, DOM Element Count, Test Data Size (in KB). The target variable is: Execution Time (in seconds). Each model was trained on a structured dataset and evaluated using a comprehensive set of regression metrics on both training and test data. Alternative: Input Parameter The predictive model was built using the following input features that influence execution time:

Test_Script_Complexity: Indicates the logical and structural difficulty of the test script.

DOM_Element_Count: Number of Document Object Model elements present, representing UI complexity.

Test_Data_Size_KB: Volume of input test data in kilobytes, which impacts processing time.

Evaluation Parameter: Output The output parameter being predicted is: Execution_Time_Seconds: The total time (in seconds) taken to execute a test case/script from start to finish. Result: The evaluation results highlight significant differences in performance across the three models. MLP achieved the highest R² (≈ 0.98) and the lowest error metrics (MSE ≈ 0.019, RMSE ≈ 0.139) on the test data, indicating excellent generalization and predictive accuracy. GPR performed moderately, while Linear Regression, though simpler, lagged behind in accuracy and error consistency. The findings support the use of MLP as the most robust and reliable model for predicting execution time in dynamic test environments