Authors
Keywords
Abstract
This study focuses on developing a predictive framework for estimating cloud migration time from Netezza to the Azure Cloud environment. As organizations increasingly adopt cloud-based infrastructure to enhance scalability and performance, accurate migration time estimation becomes a critical planning factor. The study leverages machine learning regression techniques—Gradient Boosting Regression (GBR) and Hist Gradient Boosting Regression (HGBR)—to model migration complexity and duration based on key system attributes.
Research Significance: Cloud migration, particularly from legacy systems such as Netezza, presents significant challenges in terms of estimating time, cost, and resource allocation. Accurate prediction of migration time is essential for minimizing downtime and optimizing operational efficiency. This research holds practical significance by offering a data-driven decision support model that enhances forecasting accuracy.
Methodology: The study employed a supervised machine learning approach using regression-based algorithms. A dataset comprising 500 instances was generated, containing three input parameters—DataSize_GB, NumTables, and ComplexityScore—and one output parameter, MigrationTime_Hours.
Data preprocessing steps included normalization, feature correlation analysis, and outlier treatment to ensure consistency. Two regression models—Gradient Boosting Regression (GBR) and HistGradientBoosting Regression (HGBR)—were trained and evaluated. Alternative: Input Parameters The input parameters used in this study represent key factors influencing the migration process: DataSize_GB – Denotes the total volume of data to be migrated from the Netezza system to the Azure cloud, measured in gigabytes. Larger datasets generally lead to longer migration times.
Evaluation Parameter: Output Parameter The output parameter in the model is: MigrationTime_Hours – Represents the total estimated time required for the migration process, including data transfer, validation, and post-migration optimization. The value is predicted using the trained regression models based on the input parameters.
Results: The results indicated that both models performed effectively, but HistGradientBoosting Regression (HGBR) achieved superior generalization on test data. While Gradient Boosting Regression achieved nearly perfect training accuracy (R² ≈ 0.9997), it showed signs of overfitting with a lower test performance (R² ≈ 0.685). In contrast, HGBR maintained a more balanced performance with R² ≈ 0.922 for training and R² ≈ 0.751 for testing. Additionally, HGBR exhibited lower MSE and MAE, confirming its robustness and predictive consistency.
Conclusion: This research successfully demonstrates a machine learning-based framework for predicting migration duration in Netezza to Azure Cloud migration projects. The models highlight the importance of considering multiple technical parameters—data volume, schema complexity, and table count—to achieve reliable predictions. While Gradient Boosting Regression shows high fitting accuracy, the HistGradientBoosting Regression model provides superior generalization and practical applicability.
Keywords: Cloud Migration, Netezza, Azure Cloud, Gradient Boosting Regression, HistGradientBoosting Regression, Migration Time Prediction, Machine Learning, Decision Support System, Cloud Analytics
