lavanya251104/AI_BATTERY_OPTIMIZER

Dataset Card for AI Battery Optimizer

The AI Battery Optimizer Dataset contains synthetic smartphone battery usage logs created during the development of the AI Battery Optimizer App.
It is intended for research and experimentation on battery prediction, app usage forecasting, and adaptive resource management.

---

Dataset Details

This dataset logs:

• Battery percentage over time
• Power usage (mW)
• Estimated time remaining
• Predicted app usage with confidence score
• Screen brightness level
• Frame rate (FPS)

---

Dataset Sources

• Repository: Hugging Face Dataset Repo – AI Battery Optimizer
• Related Project: AI Battery Optimizer App

---

Uses

Direct Use

• Training time-series models (Chronos, TBATS, PatchTSMixer) for predicting battery drain
• Evaluating ML-based app usage predictions
• Research on energy optimization in smartphones
• Simulating adaptive energy-saving systems

Out-of-Scope Use

• Real-world personal battery health monitoring
• Any application requiring sensitive/private user data (dataset is synthetic)

---

Dataset Structure

Format: CSV / JSON

Fields:

• timestamp → Log time (UTC)
• battery_percentage → Battery level (%)
• power_usage_mw → Power consumption in milliwatts
• time_remaining_min → Estimated time left (minutes)
• predicted_app → Next app predicted (e.g. Instagram, YouTube)
• confidence → ML prediction confidence score (0–1)
• brightness → Screen brightness (%)
• fps → Frame rate setting

Example Row:

timestamp	battery_percentage	power_usage_mw	time_remaining_min	predicted_app	confidence	brightness	fps
2025-08-28 12:30:00	85	850	272	Instagram	0.87	75	60

---

Dataset Creation

Curation Rationale

Battery drain is influenced by app usage, FPS, brightness, and background processes.
This dataset was created to simulate realistic smartphone usage patterns for developing an ML-driven energy optimization system.

Source Data

• Synthetic logs generated during AI Battery Optimizer app simulations
• Inspired by real smartphone usage, but fully anonymized

Data Collection and Processing

• Battery drain simulated every 30s via backend API
• App predictions generated every 15s with probabilistic ML logic
• Logs normalized into CSV format for training

---

Annotations

• Predictions contain confidence scores
• Users can validate predictions inside the app (feedback loop)
• Dataset can be extended with these feedback labels

---

Personal and Sensitive Information

• Dataset is synthetic
• No personal or sensitive user data included

---

Bias, Risks, and Limitations

• Synthetic dataset may not capture all real-world battery usage variability
• Predictions are approximations, not exact reflections of real device usage
• Should be treated as a benchmark/simulation dataset

Recommendations

• Use this dataset for prototyping and model training
• Fine-tune with real anonymized battery logs for production apps

---

Citation

BibTeX: