The dataset contains the print screens of the execution of following benchmark: Vellamo Browser, Vellamo Metal, Vellamo Multicore, and AnTuTu General. The benchmark was repeated until the smartphone reaches a battery level below 20%. For each benchmark, we recorded the numerical results for each microbenchmark. Each benchmark was evaluated using different approaches for CPU frequency scaling: Ondemand, Performance, Interactive, HS, ZT, OUR-G-0.1, OUR-G-0.5 and OUR-G-0.9. Ondemand, Performance, and Interactive are native from Android OS. OUR-G-0.1, OUR-G-0.5 and OUR-G-0.9 referrers to our proposed method to save energy, executed in smartphone Xiaomi Redmi Note 4. The approaches HS is from: @article{hshen2013, author = "H. Shen and Y. Tan and J. Lu and Q. Wu and Q. Qiu", title = "Achieving Autonomous Power Management Using Reinforcement Learning", journal = "ACM Transactions on Design Automation of Electronic Systems (TODAES)", volume = "18", month = "", number = "2", year = "2013", pages = "24-32" } and ZT in from: @inproceedings{ztian2018, author = "Zhongyuan Tian and Zhe Wang and Haoran Li and Peng Yang and Rafael Kioji Vivas Maeda and Jiang Xu", title = "Multi-device collaborative management through knowledge sharing", booktitle = "Proc. 23rd Asia and South Pacific Design Automation Conference (ASP-DAC)", month = "" , year = "2018", pages = "22-27" } The energy.txt file contains the acquisition of voltage and current from smartphone fuel gauge. The other files are self-explained.