README.md
2.7 KB · 96 lines · markdown Raw
1 ---
2 license: mit
3 language:
4 - en
5 tags:
6 - battery
7 - state-of-health
8 - remaining-useful-life
9 - time-series
10 - regression
11 - lstm
12 - transformer
13 - xgboost
14 - lightgbm
15 - random-forest
16 - ensemble
17 datasets:
18 - NASA-PCoE-Battery
19 metrics:
20 - r2
21 - mae
22 - rmse
23 pipeline_tag: tabular-regression
24 ---
25
26 # AI Battery Lifecycle — Model Repository
27
28 Trained model artifacts for the [aiBatteryLifeCycle](https://huggingface.co/spaces/NeerajCodz/aiBatteryLifeCycle) project.
29
30 SOH (State-of-Health) and RUL (Remaining Useful Life) prediction for lithium-ion batteries
31 trained on the NASA PCoE Battery Dataset.
32
33 ## Repository Layout
34
35 ```
36 artifacts/
37 ├── v1/
38 │ ├── models/
39 │ │ ├── classical/ # Ridge, Lasso, ElasticNet, KNN ×3, SVR, XGBoost, LightGBM, RF
40 │ │ └── deep/ # Vanilla LSTM, Bi-LSTM, GRU, Attention-LSTM, TFT,
41 │ │ # BatteryGPT, iTransformer, Physics-iTransformer,
42 │ │ # DG-iTransformer, VAE-LSTM
43 │ └── scalers/ # MinMax, Standard, Linear, Sequence scalers
44 └── v2/
45 ├── models/
46 │ ├── classical/ # Same family + Extra Trees, Gradient Boosting, best_rul_model
47 │ └── deep/ # Same deep models re-trained on v2 feature set
48 ├── scalers/ # Per-model feature scalers
49 └── results/ # Validation JSONs
50 ```
51
52 ## Model Performance Summary (v3)
53
54 | Rank | Model | R² | MAE | Family |
55 |------|-------|----|-----|--------|
56 | 1 | XGBoost | 0.9866 | 1.58 | Classical |
57 | 2 | GradientBoosting | 0.9860 | 1.38 | Classical |
58 | 3 | LightGBM | 0.9826 | 1.98 | Classical |
59 | 4 | RandomForest | 0.9814 | 1.83 | Classical |
60 | 5 | ExtraTrees | 0.9701 | 3.20 | Classical |
61 | 6 | TFT | 0.8751 | 3.88 | Transformer |
62 | 7 | Weighted Avg Ensemble | 0.8991 | 3.51 | Ensemble |
63
64 ## Usage
65
66 These artifacts are automatically downloaded by the Space on startup via
67 `scripts/download_models.py`. You can also use them directly:
68
69 ```python
70 from huggingface_hub import snapshot_download
71
72 local = snapshot_download(
73 repo_id="NeerajCodz/aiBatteryLifeCycle",
74 repo_type="model",
75 local_dir="artifacts",
76 token="<your-token>", # only needed if private
77 )
78 ```
79
80 ## Framework
81
82 - **Classical models:** scikit-learn / XGBoost / LightGBM `.joblib`
83 - **Deep models (PyTorch):** `.pt` state-dicts (CPU weights)
84 - **Deep models (Keras):** `.keras` SavedModel format
85 - **Scalers:** scikit-learn `.joblib`
86
87 ## Citation
88
89 ```bibtex
90 @misc{aiBatteryLifeCycle2025,
91 author = {Neeraj},
92 title = {AI Battery Lifecycle — SOH/RUL Prediction},
93 year = {2025},
94 url = {https://huggingface.co/spaces/NeerajCodz/aiBatteryLifeCycle}
95 }
96 ```