Software Testing based on Random Forest and Adaboost Model

Abstract

In this paper, Adaboost is integrated with Random Forest to form a new machine learning model, known as the Adaboost-Random Forest model.

Aims and Objective: To use Random Forest as a base model, and Adaboost as a meta-classifier for an adjustable and adaptable software testing process. 

Methodology: The integrated model utilized 350 datasets from Kaggle.com, which were randomly split into two groups in an 80:20 ratio. The model was implemented using the Python programming language, with the training and testing of the models performed several times to achieve accurate results. 

Result: Accuracy of 80%, a precision of 0.88, a recall of 0.68, and an F1-score of 0.71 were recorded for Random Forest. The integrated model (Adaboost-Random Forest) also had 85.5% accuracy, 0.86 precision, 0.81 recall, and 0.87 F1-score.

Conclusion: These results justified that the integrated Adaboost-Random Forest model is a more suitable model for software testing with its higher accuracy, recall, and F1-score, while the Random Forest only had higher precision. The Adaboost-Random Forest model achieved an increased accuracy of 5.5% compared to Random Forest, which is significant and demonstrates its suitability for high-accuracy testing in software engineering.