Operationally, it is increasingly burdensome for firms dealing with huge amounts of data to pay for cloud storage. Studies have revealed that a large part of the objects stored in the cloud are never or rarely accessed after the first upload, but due to bad placement policies, they continue to remain in the high-cost storage tiers. In this paper, we present a hybrid machine learning approach that combines Long Short-Term Memory (LSTM) networks with a Random Forest (RF) classifier for the automation of storage tier assignment in cloud object storage systems. The LSTM modelled temporal access patterns to predict a seven-day future access count for each object. This prediction was then used as an attribute along with eight other engineered attributes from raw access logs, and finally, a Random Forest classifier was trained to map this attribute set to one of the three storage tiers.

We conduct our experiments on the Microsoft AzureFunctionsBlobDataset2020, which is publicly available. This dataset consists of around 10.7 million blob access events over 287,432 unique objects, collected for a duration of 14 days. The proposed RF+LSTM pipeline obtains classification accuracy of 91.2% and weighted F1-score of 0.903. Based on the Azure Blob Storage pricing, our cost simulation framework estimates a 34.2% cost savings over a static baseline Hot-tier policy, and a 19.4% savings over a time-based threshold policy. An ablation study demonstrates that the access feature contribution to overall accuracy and estimated cost savings by the LSTM is 5.9% and 7.4%, respectively. Scope of dataset, assumptions made in labelling and the difference between simulated and actual deployment savings are discussed openly as limitations.

1. D. Reinsel, J. Gantz, and J. Rydning, “The digitization of the world: From edge to core,” IDC, Framingham, MA, USA, White Paper, 2018. [Online]. Available: https://www.seagate.com/files/www-content/our-story/trends/files/idc-seagate-dataage-whitepaper.pdf.
2. G. Savva, E. Kakoulli, and H. Herodotou, “Machine learning-based caching and tiering in modern data storage systems: A survey,” ACM Computing Surveys, vol. 58, no. 11, Art. no. 295, pp. 1–36, Apr. 2026, doi: 10.1145/3803857.
3. H. Zhang et al., “Cost optimization for cloud storage from user perspectives: Recent advances, taxonomy, and survey,” ACM Computing Surveys, vol. 55, no. 13s, Art. no. 266, 2023, doi: 10.1145/3582883.
4. L. Ruan, Y. Bai, S. Li, S. Hu, and L. Xiao, “Workload time series prediction in storage systems: A deep learning based approach,” Cluster Computing, vol. 26, pp. 25–35, 2023, doi: 10.1007/s10586-020-03214-y.
5. F. Romero et al., “Faa$T: A transparent auto-scaling cache for serverless applications,” in Proc. ACM Symp. Cloud Computing (SoCC), 2021, pp. 151–164, doi: 10.1145/3472883.3487013.
6. Y. Li, S. Tong, P. Zhou, Y. Liu, and J. Wu, “Efficient storage management for social network events based on clustering and hot/cold data classification,” IEEE Transactions on Computational Social Systems, vol. 10, no. 1, pp. 120–130, 2022, doi: 10.1109/TCSS.2021.3061506.
7. S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997, doi: 10.1162/neco.1997.9.8.1735.
8. L. Breiman, “Random forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001, doi: 10.1023/A:1010933404324.
9. H. Yuan, J. Bi, S. Li, J. Zhang, and M. Zhou, “An improved LSTM-based prediction approach for resources and workload in large-scale data centers,” IEEE Internet of Things Journal, vol. 11, no. 12, pp. 22816–22829, 2024, doi: 10.1109/JIOT.2024.3360085.
10. L. Zhang, X. Zhong, B. Chen, T. Guo, Z. Zhu, and Y. Zhu, “A novel hybrid model for Docker container workload prediction,” IEEE Transactions on Network and Service Management, vol. 20, no. 3, pp. 2726–2743, 2023, doi: 10.1109/TNSM.2022.3228668.
11. B. Feng, Z. Ding, and C. Jiang, “FAST: A forecasting model with adaptive sliding window and time locality integration for dynamic cloud workloads,” IEEE Transactions on Services Computing, vol. 16, no. 2, pp. 1184–1197, 2022, doi: 10.1109/TSC.2021.3098793.
12. Y.-F. Hsu, R. Irie, S. Murata, and M. Matsuoka, “A novel automated cloud storage tiering system through hot-cold data classification,” in Proc. IEEE Int. Conf. Parallel Distrib. Process. Appl., Ubiquitous Comput. Commun., Big Data Cloud Comput., Social Comput. Netw., Sustainable Comput. Commun. (ISPA/IUCC/BDCloud/SocialCom/SustainCom), 2018, pp. 306–313, doi: 10.1109/BDCloud.2018.00057.
13. T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in Proc. 22nd ACM SIGKDD Int. Conf. Knowledge Discovery and Data Mining, 2016, pp. 785–794, doi: 10.1145/2939672.2939785.
14. M. Armbrust et al., “Delta Lake: High-performance ACID table storage over cloud object stores,” Proceedings of the VLDB Endowment, vol. 13, no. 12, pp. 3411–3424, 2020, doi: 10.14778/3415478.3415560.
15. R. Wali, A. S. Imran, Z. Kastrati, and M. I. Ghafoor, “Towards optimizing storage costs on the cloud,” arXiv preprint arXiv:2305.14818, 2023.
16. Y. Mansouri, M. A. Babar, and R. Buyya, “A survey of multi-cloud management platforms,” ACM Computing Surveys, vol. 55, no. 9, Art. no. 190, 2022, doi: 10.1145/3540508.
17. Y. Guan, Z. Ma, and L. Li, “HDFS optimization strategy based on hierarchical storage of hot and cold data,” Future Generation Computer Systems, vol. 132, pp. 33–44, 2022, doi: 10.1016/j.future.2022.02.002.
18. B. Jeong, S. Baek, S. Park, J. Jeon, and Y. S. Jeong, “Stable and efficient resource management using deep neural network on cloud computing,” Neurocomputing, vol. 521, pp. 99–112, 2023, doi: 10.1016/j.neucom.2022.11.033.
19. N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “SMOTE: Synthetic minority over-sampling technique,” Journal of Artificial Intelligence Research, vol. 16, pp. 321–357, 2002, doi: 10.1613/jair.953.
20. H. Y. Tran and J. Hu, “Comprehensive analysis of random forest and XGBoost performance with SMOTE, ADASYN, and GNUS under varying imbalance levels,” Technologies, vol. 13, no. 3, Art. no. 88, 2025, doi: 10.3390/technologies13030088.
21. Microsoft Azure, “Azure Functions blob access dataset 2020,” Azure Public Dataset, GitHub, 2020. [Online]. Available: https://github.com/Azure/AzurePublicDataset.
22. N. Bender, F. Tschirner, and M. Peinl, “Cloud storage cost: A taxonomy and survey,” World Wide Web, 2024, doi: 10.1007/s11280-024-01273-4.
23. A. N. Ayesh, “Optimizing of cloud storage performance by using enhanced clustering technology,” International Journal of Computational and Electronic Aspects in Engineering, vol. 5, no. 1, pp. 16–24, 2024, doi: 10.26706/ijceae.5.1.20240503.
24. H. F. Qasim, “Intent-aware software performance optimization based on application behavioral learning,” International Journal of Computational and Electronic Aspects in Engineering, vol. 7, no. 1, pp. 82–90, 2026, doi: 10.26706/ijceae.7.1.20260109.
25. M. F. Mahdi, “Enhancing cloud security through artificial intelligence: Detecting advanced cyber attacks and analyzing anomalous patterns,” International Journal of Computational and Electronic Aspects in Engineering, vol. 6, no. 3, pp. 108–120, 2025, doi: 10.26706/ijceae.6.3.20250601.
26. A. B. Salih, S. J. Hasan, F. A. K. M. Alhelli, and M. A. Noor, “Cloud-based real-time traffic management using IoT and AI,” in Proc. 2025 3rd International Conference on Business Analytics for Technology and Security (ICBATS), 2025, pp. 1–8.