Azure cloud services rate forecast using artificial intelligence
Ashraf, Noman (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025060621192
https://urn.fi/URN:NBN:fi:amk-2025060621192
Tiivistelmä
Cloud has become an important practice for IT, as the reliance grows on cloud services, and Microsoft Azure is a key cloud service provider. As organizations continue to embrace Azure services as they meet dynamic and flexible needs, quantifying expenditures associated with cloud services in advance is vital for budgeting and control. In this thesis, the focus was on using a time series forecasting method called ARIMA to estimate monthly costs for the Azure services, such as virtual machines, storage, and bandwidth. Using historical usage data from May 2023 to April 2024, the research also sought to develop precise expense figures that organizations can use to make informed decisions regarding the acquisition of resources.
The research comprised key issues of the Azure billing scheme, which is based on different and variable prices for each customer and usage metric. The thesis assessed the effects of seasonal variation and trends on expenditure predictions and demonstrated how organizations can efficiently use these predictions in their financial planning. Data preparation steps included dealing with missing data, finding out whether the dataset is stationary by performing the ADF test, and proceeding with differencing where stationarity is not achieved. p, d, and q values were attained through ACF and PACF analysis and by comparing graphical plots of actual and forecasted cost variables.
The results showed the skills of the ARIMA model to make feasible forecasts for the Azure service expense for the period May 2024 to June 2025. This makes it easier for organizations to plan for cost, avoid cases of emergent budgetary overruns, and improve decision-making. When it comes to cost estimation in cloud environments, the study emphasized the necessity to work more on the accuracy of cost forecasting models and present several opportunities for further investigation of better cost prediction tools to take into account other parameters, with the in-volvement of the cloud environment. Evaluation of further research in light of this, more complex quantum cost models can be developed in the future, along with other specifications of the cloud environment.
The research comprised key issues of the Azure billing scheme, which is based on different and variable prices for each customer and usage metric. The thesis assessed the effects of seasonal variation and trends on expenditure predictions and demonstrated how organizations can efficiently use these predictions in their financial planning. Data preparation steps included dealing with missing data, finding out whether the dataset is stationary by performing the ADF test, and proceeding with differencing where stationarity is not achieved. p, d, and q values were attained through ACF and PACF analysis and by comparing graphical plots of actual and forecasted cost variables.
The results showed the skills of the ARIMA model to make feasible forecasts for the Azure service expense for the period May 2024 to June 2025. This makes it easier for organizations to plan for cost, avoid cases of emergent budgetary overruns, and improve decision-making. When it comes to cost estimation in cloud environments, the study emphasized the necessity to work more on the accuracy of cost forecasting models and present several opportunities for further investigation of better cost prediction tools to take into account other parameters, with the in-volvement of the cloud environment. Evaluation of further research in light of this, more complex quantum cost models can be developed in the future, along with other specifications of the cloud environment.