A novel methodological approach to SaaS churn prediction using whale optimization algorithm.

Area of Science:

• Computer Science
• Machine Learning
• Cloud Computing

Background:

• Customer churn poses a significant threat to Software as a Service (SaaS) companies, impacting sustained growth in the cloud computing sector.
• Limited research exists on SaaS-specific churn models, particularly concerning feature selection and predictive algorithm efficacy.
• Effective churn prediction is crucial for informed managerial strategies and academic understanding.

Purpose of the Study:

• To introduce a novel approach for predicting customer churn in SaaS using the Whale Optimization Algorithm (WOA) for feature selection.
• To evaluate the performance of WOA-reduced datasets against full-variable and chi-squared-derived datasets in SaaS churn prediction.
• To compare various machine learning algorithms on these datasets using standard performance metrics.

Main Methods:

• Feature selection was performed using the Whale Optimization Algorithm (WOA).
• Three datasets were created: WOA-reduced, full-variable, and chi-squared-derived, from a multinational SaaS company's user data (>1,000 users).
• Machine learning models including k-nearest neighbor, Decision Trees, Naïve Bayes, Random Forests, and Neural Networks were applied and evaluated using AUC, Accuracy, Precision, Recall, and F1 Score.

Main Results:

• The WOA-reduced dataset demonstrated superior predictive performance compared to the full-variable and chi-squared-derived datasets.
• Optimizing feature selection via WOA enhanced processing efficiency.
• All tested machine learning algorithms performed better on the WOA-reduced dataset.

Conclusions:

• The Whale Optimization Algorithm (WOA) is an effective method for feature selection in SaaS churn prediction models.
• WOA-based feature reduction leads to improved predictive accuracy and efficiency in SaaS environments.
• This approach offers valuable insights for SaaS businesses aiming to mitigate customer churn.