Duplicate from NifferLi/Cold-Chain-Transportation-Strawberry

Co-authored-by: Hu Li <NifferLi@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,66 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text
+article_release/ALL_benchmark_W60.xlsx filter=lfs diff=lfs merge=lfs -text
+Cold-Chain-Transportation-Strawberry-Release/ALL_benchmark_W60.xlsx filter=lfs diff=lfs merge=lfs -text
+article_release/ALL_benchmark_W60/ALL_benchmark_W60.xlsx filter=lfs diff=lfs merge=lfs -text
+article_release/_xet_repair_test/ALL_benchmark_W60.xlsx filter=lfs diff=lfs merge=lfs -text
+article_release/_fallback_test/ALL_benchmark_W60.xlsx filter=lfs diff=lfs merge=lfs -text
+article_release/ALL_benchmark_W60_release_v1.xlsx filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,837 @@
+---
+license: apache-2.0
+language:
+  - en
+pretty_name: Cold-Chain Transportation Strawberry Dataset for ADVEI Article Release
+tags:
+  - time-series
+  - cold-chain
+  - early-warning
+  - risk-prediction
+  - human-centric-ai
+  - supply-chain
+  - explainability
+  - prescriptive-analytics
+task_categories:
+  - tabular-classification
+size_categories:
+  - 10K<n<100K
+---
+
+# Cold-Chain Transportation Strawberry Dataset — ADVEI Article Release
+
+This repository provides the processed dataset used in the accepted *Advanced Engineering Informatics* article:
+
+**A Human-Centric Edge-Oriented Decision Support System for Cold Chain Transportation: Early Warning, Trigger-Time Explanation, and Prescriptive Action Ranking**
+
+To be published in: *Advanced Engineering Informatics*.
+
+## Final Article Release
+
+The finalized article-release dataset is hosted directly in this Hugging Face repository and can be viewed or downloaded using the links below:
+
+| File | Hugging Face |
+|---|---|
+| `ALL_benchmark_W60.parquet` | [View or download](article_release/ALL_benchmark_W60.parquet) |
+| `ALL_benchmark_W60.xlsx` | [View or download](article_release/ALL_benchmark_W60.xlsx) |
+
+The Parquet file is recommended for programmatic use. The Excel file is provided for convenient inspection.
+
+If the Hugging Face preview or download is temporarily unavailable, the same files can be downloaded from the following public Google Drive backup folder:
+
+[Download from the Google Drive backup mirror](https://drive.google.com/drive/folders/1nGwz-wM6gM68djXA60qpG-73-kPFidKW?usp=sharing)
+
+Detailed download instructions are available at:
+
+[`article_release/DOWNLOAD_DATA.md`](article_release/DOWNLOAD_DATA.md)
+
+The final article-release table contains the integrated W60 processed benchmark used in the study, including shipment identifiers, timestamps, resampled multi-sensor temperature records, engineered W60 features, risk labels, future severe-risk prediction targets, cause flags for explanation consistency checking, and evaluation/audit-related fields.
+
+---
+
+## Dataset Summary
+
+- **Domain:** Cold-chain logistics for strawberry transportation
+- **Source:** Public strawberry cold-chain transportation dataset
+- **Entities:** 6 shipments (`S1`–`S6`)
+- **Sensors:** 9 temperature probe positions per timestamp
+- **Sensor layout:** Front / Middle / Rear × Top / Middle / Bottom
+- **Sampling interval after processing:** 10 minutes
+- **Feature window:** W60, using the past 60 minutes
+- **Primary prediction horizon:** 120 minutes
+- **Primary target:** `y_next_120_R2`
+- **Main evaluation setting:** leave-one-shipment-out (LOSO) generalization
+
+The primary task is to predict, at time `t`, whether the shipment will enter a severe-risk state (`R2`) within the next 120 minutes, using 10-minute sampled multi-sensor temperature data and engineered past-window statistics.
+
+This dataset supports research on:
+
+- cold-chain early-warning prediction;
+- deployment-like generalization across unseen shipments;
+- event-level alerting evaluation rather than point-wise classification only;
+- trigger-time explanation and weak-supervision consistency checking;
+- human-centric decision support and prescriptive action ranking.
+
+---
+
+## Repository Structure
+
+```text
+Cold-Chain-Transportation-Strawberry/
+├── article_release/
+│   ├── ALL_benchmark_W60.parquet
+│   ├── ALL_benchmark_W60.xlsx
+│   └── DOWNLOAD_DATA.md
+├── data/
+│   ├── w60_S1.parquet
+│   ├── w60_S2.parquet
+│   ├── w60_S3.parquet
+│   ├── w60_S4.parquet
+│   ├── w60_S5.parquet
+│   ├── w60_S6.parquet
+│   └── w60_all.parquet
+├── folds/
+├── splits/
+├── benchmark_v2/
+├── benchmark_v2_pca/
+└── README.md
+```
+
+The two finalized files under `article_release/` are the primary files for reproducing or inspecting the dataset used in the accepted ADVEI article. A public Google Drive backup mirror is provided in `article_release/DOWNLOAD_DATA.md`.
+
+## Recommended Files
+
+The authoritative processed dataset for the accepted ADVEI article is:
+
+```text
+article_release/ALL_benchmark_W60.parquet
+```
+
+The corresponding Excel file is:
+
+```text
+article_release/ALL_benchmark_W60.xlsx
+```
+
+The Parquet file is recommended for programmatic analysis. The Excel file contains the same article-release dataset in a format suitable for convenient inspection.
+
+If either Hugging Face file is temporarily unavailable, use the public Google Drive backup mirror:
+
+[Open the Google Drive backup folder](https://drive.google.com/drive/folders/1nGwz-wM6gM68djXA60qpG-73-kPFidKW?usp=sharing)
+
+The six shipment-level files under `data/` are retained for shipment-level inspection:
+
+```text
+data/w60_S1.parquet
+data/w60_S2.parquet
+data/w60_S3.parquet
+data/w60_S4.parquet
+data/w60_S5.parquet
+data/w60_S6.parquet
+```
+
+The all-shipment file under `data/` is retained for convenience:
+
+```text
+data/w60_all.parquet
+```
+
+The folders `benchmark_v2/` and `benchmark_v2_pca/` are earlier or auxiliary processed releases. They are retained for transparency but are not the primary files for reproducing the accepted ADVEI article.
+
+For the final article release, use the files under `article_release/`.
+
+---
+
+## Primary Prediction Task
+
+### Target Label
+
+```text
+y_next_120_R2
+```
+
+### Meaning
+
+At time `t`, predict whether the shipment will enter the severe-risk state `R2` within the next 120 minutes.
+
+- `1` = the shipment will enter R2 within the prediction horizon
+- `0` = the shipment will not enter R2 within the prediction horizon
+
+Future labels such as `y_next_*` and time-to-event fields such as `eta_to_R2_*` are provided for ground truth and evaluation only. They must not be used as predictive input features.
+
+---
+
+## How to Load
+
+### Load the final article-release Parquet file
+
+```python
+from huggingface_hub import hf_hub_download
+import pandas as pd
+
+repo_id = "NifferLi/Cold-Chain-Transportation-Strawberry"
+
+path = hf_hub_download(
+    repo_id=repo_id,
+    filename="article_release/ALL_benchmark_W60.parquet",
+    repo_type="dataset"
+)
+
+df = pd.read_parquet(path)
+
+print(df.shape)
+print(df.head())
+```
+
+### Load the Excel version
+
+```python
+from huggingface_hub import hf_hub_download
+import pandas as pd
+
+repo_id = "NifferLi/Cold-Chain-Transportation-Strawberry"
+
+path = hf_hub_download(
+    repo_id=repo_id,
+    filename="article_release/ALL_benchmark_W60.xlsx",
+    repo_type="dataset"
+)
+
+df = pd.read_excel(path)
+
+print(df.shape)
+print(df.head())
+```
+
+### Backup Download
+
+If the Hugging Face preview or download is temporarily unavailable, download the same files from the public Google Drive backup folder:
+
+[Google Drive backup folder](https://drive.google.com/drive/folders/1nGwz-wM6gM68djXA60qpG-73-kPFidKW?usp=sharing)
+
+After downloading, the files can be loaded locally:
+
+```python
+import pandas as pd
+
+df_parquet = pd.read_parquet("ALL_benchmark_W60.parquet")
+df_excel = pd.read_excel("ALL_benchmark_W60.xlsx")
+
+print(df_parquet.shape)
+print(df_excel.shape)
+```
+
+### Load a shipment-level Parquet file
+
+```python
+from huggingface_hub import hf_hub_download
+import pandas as pd
+
+repo_id = "NifferLi/Cold-Chain-Transportation-Strawberry"
+
+path = hf_hub_download(
+    repo_id=repo_id,
+    filename="data/w60_S1.parquet",
+    repo_type="dataset"
+)
+
+df_s1 = pd.read_parquet(path)
+
+print(df_s1.shape)
+print(df_s1.head())
+```
+
+---
+
+## Column Groups
+
+Each row corresponds to one W60 window snapshot for one shipment at one timestamp.
+
+### Identifiers and Time
+
+```text
+Time
+window_id
+window_start_time
+window_end_time
+shipment_id
+```
+
+### Raw Sensor Readings at Time t
+
+```text
+Front_Top
+Front_Middle
+Front_Bottom
+Middle_Top
+Middle_Middle
+Middle_Bottom
+Rear_Top
+Rear_Middle
+Rear_Bottom
+```
+
+Missing readings are recorded as `NaN`.
+
+### Missing Masks at Time t
+
+```text
+mask_Front_Top
+mask_Front_Middle
+mask_Front_Bottom
+mask_Middle_Top
+mask_Middle_Middle
+mask_Middle_Bottom
+mask_Rear_Top
+mask_Rear_Middle
+mask_Rear_Bottom
+```
+
+### Data Quality and Guardrail Fields
+
+```text
+N_valid
+coverage_points
+N_active_t
+coverage_time
+sconf
+conf_band
+conf_level
+is_incomplete
+is_fail_safe
+is_soft_guardrail
+is_guardrail
+is_trainable
+mask_ratio_t
+```
+
+### Current Rule-Based Risk Stage
+
+```text
+risk_level
+label_R0
+label_R1
+label_R2
+```
+
+Risk level definitions:
+
+- `0` = R0, normal
+- `1` = R1, warning
+- `2` = R2, severe risk
+
+### Cause Flags for Explanation Consistency Checking
+
+```text
+cause_high_peak
+cause_high_duration
+cause_low_peak
+cause_low_duration
+```
+
+These cause flags are current-time rule-derived indicators based on sensor readings. They are retained for weak-supervision consistency checking and audit purposes.
+
+### Rule Primitives and Current-State Statistics
+
+```text
+T_max_window
+T_min_window
+T_mean
+T_std
+dur_gt4
+dur_lt0
+dur_lt_minus1
+has_over10
+spatial_range_t
+spatial_std_t
+T_mean_t
+hot_ratio_t
+cold_ratio_t
+```
+
+### Future Labels and Time-to-Event Fields
+
+```text
+y_next_60_R2
+eta_to_R2_60
+y_next_120_R2
+eta_to_R2_120
+```
+
+These are target or evaluation fields and must not be used as model input features.
+
+### W60 Engineered Features
+
+Examples include:
+
+```text
+W60_T_mean
+W60_T_std
+W60_T_min
+W60_T_max
+W60_T_range
+W60_delta
+W60_slope
+W60_spatial_range_mean
+W60_spatial_range_max
+W60_spatial_std_mean
+W60_hot_ratio_mean
+W60_hot_ratio_max
+W60_over_auc_mean
+W60_over_auc_max
+W60_under_auc_mean
+W60_under_auc_max
+W60_over_dur_mean
+W60_under_dur_mean
+W60_active_ratio_mean
+W60_mask_ratio_mean
+W60_runlen_hot_any_min
+W60_runlen_cold_any_min
+W60_runlen_hot_mean_min
+W60_runlen_cold_mean_min
+W60_runlen_hot_any_ratio
+W60_runlen_cold_any_ratio
+W60_runlen_hot_mean_ratio
+W60_runlen_cold_mean_ratio
+```
+
+### v4 Engineered Features
+
+Examples include:
+
+```text
+v4_over_auc_t
+v4_under_auc_t
+v4_over_max_t
+v4_under_max_t
+v4_hot_ratio_t
+v4_cold_ratio_t
+v4_spatial_range_t
+v4_spatial_std_t
+v4_median_t
+v4_iqr_t
+v4_p90_t
+v4_p95_t
+v4_shock_t
+v4_slope_short_t
+v4_slope_long_t
+v4_accel_t
+v4_active_ratio_t
+v4_missing_streak_t
+```
+
+---
+
+## Leakage Policy
+
+To ensure deployment-realistic evaluation, future labels and evaluation-only fields must be excluded from predictive model inputs.
+
+### Must Exclude from Predictive Inputs
+
+```text
+y_next_60_R2
+eta_to_R2_60
+y_next_120_R2
+eta_to_R2_120
+risk_level
+label_R0
+label_R1
+label_R2
+```
+
+The target column for the main task is:
+
+```text
+y_next_120_R2
+```
+
+### Cause Flags
+
+```text
+cause_high_peak
+cause_high_duration
+cause_low_peak
+cause_low_duration
+```
+
+These cause flags are retained for explanation consistency checking and audit purposes. If users train alternative models, they should clearly report whether these fields are included or excluded.
+
+For reproducing the article protocol, users should follow the feature exclusion rules described in the associated article and use the event-level early-warning evaluation protocol.
+
+---
+
+## Evaluation Protocol
+
+### Outer Validation
+
+Use leave-one-shipment-out (LOSO) validation:
+
+- train on 5 shipments;
+- test on the held-out shipment;
+- repeat for all six shipments;
+- report mean and standard deviation across `S1`–`S6`.
+
+### Metrics
+
+Report both point-wise and event-level metrics.
+
+Point-wise metrics may include:
+
+```text
+Precision
+Recall
+F1
+```
+
+Event-level metrics may include:
+
+```text
+EVENT_F1
+EVENT_TP
+EVENT_FN
+EVENT_FP_OUTSIDE
+EVENT_PRED_TOTAL
+LEAD_mean
+LEAD_median
+```
+
+### Event-Level Alerting
+
+Point-wise predictions can be converted into alert events using a persistence-plus-cooldown policy.
+
+Typical operational parameters used in the article pipeline are:
+
+```text
+PERSIST_K = 1
+COOLDOWN_MIN = 120
+```
+
+Event-level evaluation should focus on early warning rather than within-crisis identification. Detections after the shipment is already in R2 should not be rewarded as valid early-warning detections.
+
+---
+
+## Suggested Baselines
+
+### Model Baselines
+
+- ExtraTrees (ET)
+- RandomForest (RF)
+- Logistic Regression (LOGIT)
+- Gradient boosting models such as LightGBM or XGBoost as optional comparisons
+
+### Rule Baselines
+
+Deterministic threshold baselines can be constructed using rule-related primitives such as:
+
+```text
+T_max_window
+T_min_window
+dur_gt4
+dur_lt0
+dur_lt_minus1
+has_over10
+```
+
+These rule baselines are useful for sanity checks and interpretability comparisons.
+
+---
+
+## Human-Centric Decision Support Outputs
+
+The dataset was used in a human-centric edge-oriented decision support pipeline including:
+
+- predictive early warning;
+- trigger-time local explanation;
+- trigger-type probability representation;
+- prescriptive action ranking;
+- operator-facing structured messages;
+- explanation and message audit.
+
+The cause flags and risk-stage fields support weak-supervision consistency checking and audit analysis. The article evaluates the complete system through event-level prediction, explanation consistency, prescriptive action ranking, message audit, and a controlled human-subject decision-support experiment.
+
+---
+
+## Source Dataset
+
+The processed benchmark in this repository is derived from a publicly available strawberry cold-chain transportation dataset:
+
+```text
+Abdella, A., Brecht, J. K., & Uysal, I.
+A time-temperature dataset for the strawberry cold chain across multiple shipments and locations.
+arXiv preprint arXiv:2103.12895.
+```
+
+The processed files in this repository provide the article-specific W60 benchmark used for early-warning prediction, explanation, and decision-support evaluation.
+
+---
+
+## Citation
+
+If you use this dataset, please cite the associated article:
+
+```text
+Li, H., Uygun, Ö., Yu, X., Zhou, Y., Chang, X., & Chen, C.-H.
+A Human-Centric Edge-Oriented Decision Support System for Cold Chain Transportation:
+Early Warning, Trigger-Time Explanation, and Prescriptive Action Ranking.
+Advanced Engineering Informatics, forthcoming.
+```
+
+The DOI and final bibliographic details will be added once available.
+
+You may also cite this dataset repository as:
+
+```bibtex
+@dataset{li_coldchain_transportation_strawberry_advei,
+  author    = {Li, Hu},
+  title     = {Cold-Chain Transportation Strawberry Dataset for ADVEI Article Release},
+  publisher = {Hugging Face},
+  year      = {2026},
+  note      = {Processed dataset for the accepted Advanced Engineering Informatics article}
+}
+```
+
+---
+
+## Contact
+
+For questions regarding this dataset, please open an issue in this repository or contact the corresponding author listed in the associated article.
+
+---
+
+# Appendix A — Formal Label and Risk Definitions
+
+This appendix summarises the rule-stage labels and future labels used in the processed benchmark.
+
+## A.1 Notation
+
+- Sampling interval: `Δt = 10 minutes`
+- Window length: `W = 60 minutes`
+- Number of time points in each W60 window: 6
+- Number of temperature sensors: 9
+- Let `x_{t,s}` denote the temperature at time `t` for sensor `s`.
+
+## A.2 Rule Primitives Computed on the W60 Window
+
+Define per-time-step maxima and minima across sensors:
+
+```text
+Tmax_j = max_s x_{j,s}
+Tmin_j = min_s x_{j,s}
+```
+
+Rule primitives include:
+
+```text
+dur_gt4(t)
+dur_lt0(t)
+dur_lt_minus1(t)
+has_over10(t)
+T_min_window(t)
+T_max_window(t)
+```
+
+where:
+
+- `dur_gt4(t)` measures cumulative exposure above 4°C within the W60 window;
+- `dur_lt0(t)` measures cumulative exposure below 0°C within the W60 window;
+- `dur_lt_minus1(t)` measures cumulative exposure below -1°C within the W60 window;
+- `has_over10(t)` indicates whether temperature above 10°C occurs within the W60 window;
+- `T_min_window(t)` and `T_max_window(t)` are the minimum and maximum observed temperatures within the W60 window.
+
+## A.3 Cause Indicators
+
+The four cause indicators are:
+
+```text
+cause_high_peak
+cause_high_duration
+cause_low_peak
+cause_low_duration
+```
+
+They correspond to:
+
+- high-temperature peak excursion;
+- sustained high-temperature exposure;
+- low-temperature peak excursion;
+- sustained low-temperature exposure.
+
+## A.4 Current Rule Risk Stage
+
+The processed benchmark contains:
+
+```text
+risk_level
+label_R0
+label_R1
+label_R2
+```
+
+The risk levels are:
+
+- `R0`: normal
+- `R1`: warning
+- `R2`: severe risk
+
+## A.5 Future Labels
+
+The released future-label columns are:
+
+```text
+y_next_60_R2
+y_next_120_R2
+eta_to_R2_60
+eta_to_R2_120
+```
+
+The primary article task uses:
+
+```text
+y_next_120_R2
+```
+
+For the article protocol, timestamps already in R2 are included during model training, but detections after the shipment is already in R2 are not rewarded as valid early-warning detections during event-level evaluation. Therefore, users should use the released target columns as provided and apply the event-level early-warning masking rule when reproducing article-level early-warning evaluation.
+
+All future checks are performed within the same shipment.
+
+---
+
+# Appendix B — Data Quality and Guardrails
+
+## B.1 Coverage
+
+At each time `t`:
+
+```text
+N_valid(t) = number of observed sensors at time t
+coverage_points(t) = N_valid(t) / 9
+```
+
+Within the W60 window:
+
+```text
+N_active_t(t) = number of active time points in the W60 window
+coverage_time(t) = N_active_t(t) / 6
+```
+
+## B.2 Confidence Score and Banding
+
+The sensor confidence score is:
+
+```text
+sconf(t) = (coverage_points(t) + coverage_time(t)) / 2
+```
+
+Confidence bands are encoded in:
+
+```text
+conf_band
+conf_level
+```
+
+The corresponding guardrail fields are:
+
+```text
+is_incomplete
+is_fail_safe
+is_soft_guardrail
+is_guardrail
+is_trainable
+```
+
+These fields are used to distinguish full, partial, and zero-observability regimes and to support audit and reliability handling in the decision-support pipeline.
+
+---
+
+# Appendix C — Practical Feature Grouping
+
+## C.1 Raw Sensors
+
+```text
+Front_Top
+Front_Middle
+Front_Bottom
+Middle_Top
+Middle_Middle
+Middle_Bottom
+Rear_Top
+Rear_Middle
+Rear_Bottom
+```
+
+## C.2 Sensor Masks
+
+```text
+mask_Front_Top
+mask_Front_Middle
+mask_Front_Bottom
+mask_Middle_Top
+mask_Middle_Middle
+mask_Middle_Bottom
+mask_Rear_Top
+mask_Rear_Middle
+mask_Rear_Bottom
+```
+
+## C.3 Data Quality and Guardrails
+
+```text
+N_valid
+coverage_points
+N_active_t
+coverage_time
+sconf
+conf_band
+conf_level
+is_incomplete
+is_fail_safe
+is_soft_guardrail
+is_guardrail
+is_trainable
+mask_ratio_t
+```
+
+## C.4 Rule and Audit Fields
+
+```text
+risk_level
+label_R0
+label_R1
+label_R2
+cause_high_peak
+cause_high_duration
+cause_low_peak
+cause_low_duration
+```
+
+## C.5 Future Labels and Evaluation Fields
+
+```text
+y_next_60_R2
+eta_to_R2_60
+y_next_120_R2
+eta_to_R2_120
+```
+
+## C.6 Window and Engineered Features
+
+Feature families include:
+
+```text
+W60_*
+v4_*
+spatial_*
+T_*
+dur_*
+```
+
+Users should inspect the column names in `article_release/ALL_benchmark_W60.parquet` for the complete feature list.
+
+---
+
+## Changelog
+
+- `article_release`: Final processed benchmark files and download instructions for the accepted ADVEI article.
+- `ALL_benchmark_W60.parquet` and `ALL_benchmark_W60.xlsx` are hosted directly in the Hugging Face repository.
+- A public Google Drive folder is maintained as a backup mirror in case Hugging Face preview or download is temporarily unavailable.
+- Earlier folders such as `benchmark_v2/` and `benchmark_v2_pca/` are retained as legacy or auxiliary processed releases.

article_release/ALL_benchmark_W60.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84638d948e04ca1eb810d2f46848eeadcacbb9d719b40aa647a4d68e1c9fff88
+size 1446284

article_release/ALL_benchmark_W60.xlsx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d35ec35f55fbe602486b6548c921327de0f1b7b2c8673ecd838dc0e3236d8d36
+size 6640576

article_release/DOWNLOAD_ALL_benchmark_W60_PARQUET.md

@@ -0,0 +1 @@
+https://drive.google.com/drive/folders/1nGwz-wM6gM68djXA60qpG-73-kPFidKW?usp=sharing

article_release/DOWNLOAD_ALL_benchmark_W60_XLSX.md

@@ -0,0 +1 @@
+https://drive.google.com/drive/folders/1nGwz-wM6gM68djXA60qpG-73-kPFidKW?usp=sharing

article_release/README.md

@@ -0,0 +1,159 @@
+# Article-Release Dataset Downloads
+
+The finalized article-release dataset is hosted directly in this Hugging Face repository.
+
+## Primary Download — Hugging Face
+
+| File | Format | View on Hugging Face | Direct download |
+|---|---|---|---|
+| `ALL_benchmark_W60.parquet` | Apache Parquet | [View file](https://huggingface.co/datasets/NifferLi/Cold-Chain-Transportation-Strawberry/blob/main/article_release/ALL_benchmark_W60.parquet) | [Download](https://huggingface.co/datasets/NifferLi/Cold-Chain-Transportation-Strawberry/resolve/main/article_release/ALL_benchmark_W60.parquet?download=true) |
+| `ALL_benchmark_W60.xlsx` | Microsoft Excel | [View file](https://huggingface.co/datasets/NifferLi/Cold-Chain-Transportation-Strawberry/blob/main/article_release/ALL_benchmark_W60.xlsx) | [Download](https://huggingface.co/datasets/NifferLi/Cold-Chain-Transportation-Strawberry/resolve/main/article_release/ALL_benchmark_W60.xlsx?download=true) |
+
+The Parquet file is recommended for programmatic analysis. The Excel file is provided for convenient inspection and use in spreadsheet software.
+
+## Backup Download — Google Drive
+
+If the Hugging Face preview or download is temporarily unavailable, the same article-release files can be downloaded from the following public Google Drive backup folder:
+
+[Open the Google Drive backup folder](https://drive.google.com/drive/folders/1nGwz-wM6gM68djXA60qpG-73-kPFidKW?usp=sharing)
+
+The backup folder contains:
+
+```text
+ALL_benchmark_W60.parquet
+ALL_benchmark_W60.xlsx
+```
+
+The Google Drive folder is configured as:
+
+> **Anyone with the link → Viewer**
+
+No access request should normally be required.
+
+## Dataset Description
+
+Both files contain the same finalized article-release benchmark dataset in different formats.
+
+The dataset contains:
+
+- **14,398 rows**
+- **107 columns**
+- data from six strawberry cold-chain shipments;
+- resampled multi-sensor temperature measurements;
+- engineered W60 features;
+- current risk-stage labels;
+- future severe-risk prediction targets;
+- explanation-consistency cause flags;
+- data-quality, confidence, and audit-related fields.
+
+## Recommended Format
+
+### Parquet
+
+Use `ALL_benchmark_W60.parquet` for:
+
+- Python or R analysis;
+- machine-learning experiments;
+- preservation of data types;
+- efficient loading and storage.
+
+### Excel
+
+Use `ALL_benchmark_W60.xlsx` for:
+
+- manual inspection;
+- spreadsheet-based review;
+- convenient viewing of columns and values.
+
+## Loading with Python
+
+### Parquet
+
+```python
+from huggingface_hub import hf_hub_download
+import pandas as pd
+
+repo_id = "NifferLi/Cold-Chain-Transportation-Strawberry"
+
+path = hf_hub_download(
+    repo_id=repo_id,
+    filename="article_release/ALL_benchmark_W60.parquet",
+    repo_type="dataset"
+)
+
+df = pd.read_parquet(path)
+
+print(df.shape)
+print(df.head())
+```
+
+### Excel
+
+```python
+from huggingface_hub import hf_hub_download
+import pandas as pd
+
+repo_id = "NifferLi/Cold-Chain-Transportation-Strawberry"
+
+path = hf_hub_download(
+    repo_id=repo_id,
+    filename="article_release/ALL_benchmark_W60.xlsx",
+    repo_type="dataset"
+)
+
+df = pd.read_excel(path)
+
+print(df.shape)
+print(df.head())
+```
+
+## Loading Files Downloaded from Google Drive
+
+If the files were downloaded from the Google Drive backup folder, load them directly from the local directory:
+
+```python
+import pandas as pd
+
+df_parquet = pd.read_parquet("ALL_benchmark_W60.parquet")
+df_excel = pd.read_excel("ALL_benchmark_W60.xlsx")
+
+print(df_parquet.shape)
+print(df_excel.shape)
+```
+
+## Availability Note
+
+Hugging Face is the primary hosting and documentation platform for this dataset.
+
+The public Google Drive folder is maintained as a backup mirror to ensure continuous access if the Hugging Face file preview, content-delivery service, or direct download is temporarily unavailable.
+
+Both locations provide the same finalized article-release files.
+
+## Citation
+
+When using this dataset, please cite the associated article:
+
+```text
+Li, H., Uygun, Ö., Yu, X., Zhou, Y., Chang, X., & Chen, C.-H.
+A Human-Centric Edge-Oriented Decision Support System for Cold Chain Transportation:
+Early Warning, Trigger-Time Explanation, and Prescriptive Action Ranking.
+Advanced Engineering Informatics, forthcoming.
+```
+
+The DOI and final bibliographic details will be added once available.
+
+The dataset repository may also be cited as:
+
+```bibtex
+@dataset{li_coldchain_transportation_strawberry_advei,
+  author    = {Li, Hu},
+  title     = {Cold-Chain Transportation Strawberry Dataset for ADVEI Article Release},
+  publisher = {Hugging Face},
+  year      = {2026},
+  note      = {Processed dataset for the accepted Advanced Engineering Informatics article}
+}
+```
+
+## Contact
+
+For questions about the dataset, file contents, or download access, please open a discussion in the Hugging Face dataset repository.

benchmark_v2/S1_aligned_strict_linear_with_labels.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:849e959ffddf0a524202dee72b3d337a8f06c6956f94585b1da3f2fd04eadc25
+size 145155

benchmark_v2/S2_aligned_strict_linear_with_labels.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e086dde035554a9535104481196ed801850125690a9fc8bea1dcbd23e7af073c
+size 135329

benchmark_v2/S3_aligned_strict_linear_with_labels.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3267578e79ef349a88a939b42d6fec215d8b69c520def1f2b218f4067e3e6008
+size 120816

benchmark_v2/S4_aligned_strict_linear_with_labels.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7971fb61b41c0fadfe915f41beb5c1fe7c828c9b4e1190eb4c698051e1921cb
+size 120750

benchmark_v2/S5_aligned_strict_linear_with_labels.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb6157685e2a53a4a483c7cae438f4f20fa87314aeb7433639abaad69d71faed
+size 105608

benchmark_v2/S6_aligned_strict_linear_with_labels.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d130834bbcf92c976e7c8c466c8797955f2bfb02d46ba887323e6458e1c5c57f
+size 120205

benchmark_v2/all.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd1f2458c7c3d516e3c8bb6589968dd81610b8cd3d218fa10efdd271f4f7a71e
+size 412319

benchmark_v2_pca/S1_aligned_strict_linear_with_labels_pca10.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b93a07ca4312035757a76fe8ff37f62f90819a985a723806bfb75aa1adb91dd
+size 256654

benchmark_v2_pca/S1_aligned_strict_linear_with_labels_pca20.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:770ea2bfbe8785c7e6e8d4e138604737ad2143a8c739af03f54018c3c7c0dc78
+size 368169

benchmark_v2_pca/S1_aligned_strict_linear_with_labels_pca50.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4517612a17bffdb9b238accbfe6f1ba143776bc2622ab311ac0280a9d7ad28b6
+size 702743

benchmark_v2_pca/S2_aligned_strict_linear_with_labels_pca10.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:671cf1474b6923a35ff2bbfa120b17c63f94d09425a769f076a2b4dd2a2a4c4e
+size 276544

benchmark_v2_pca/S2_aligned_strict_linear_with_labels_pca20.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:893d46d46033fcb1fe48845750e24dc37d05f1e21220d3f921157868c69f08df
+size 417767

benchmark_v2_pca/S2_aligned_strict_linear_with_labels_pca50.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61c22f6bcf81997915fc307b99c7ba2c88a65e7811d42716a76f0b0dc3ccf22e
+size 841441

benchmark_v2_pca/S3_aligned_strict_linear_with_labels_pca10.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea90b2e7bf93c0d4c23957d48f2c72e17f919c69eb5baa88afbd718f45f9de73
+size 244743

benchmark_v2_pca/S3_aligned_strict_linear_with_labels_pca20.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f69013599b75abfcfc4b08cd6139d5d9675a324a5f7c5b4787dd86585d0bb598
+size 368682

benchmark_v2_pca/S3_aligned_strict_linear_with_labels_pca50.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bd2c6cf3244e859e51c83878c258213b49f1313fe7d7f056ef8b99fdcac6563
+size 740508

benchmark_v2_pca/S4_aligned_strict_linear_with_labels_pca10.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd7211d818af77d31b3e8d271bd917b63c577e64ff2518cc6a336974207cdc90
+size 244805

benchmark_v2_pca/S4_aligned_strict_linear_with_labels_pca20.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:082bbe522aaeb76ec9bfb1acb8eb93261d88935a132a969152ec1f5ba4fe2642
+size 368864

benchmark_v2_pca/S4_aligned_strict_linear_with_labels_pca50.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3fad969e5eee6b8398b24f6f0a4affdb6a0a3aa3533749fdfd4f0f17a2923a9e
+size 741058

benchmark_v2_pca/S5_aligned_strict_linear_with_labels_pca10.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:298482458593c3696ba2663d63302ca23adfcd4f31661443347be9aeb696c9a0
+size 209853

benchmark_v2_pca/S5_aligned_strict_linear_with_labels_pca20.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6178e8f41a1cb927b229bd21eca494866eff61b946a144f306293439ad453f06
+size 314190

benchmark_v2_pca/S5_aligned_strict_linear_with_labels_pca50.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55b63a8b1cdf07650d0dfad5e703bfbe270bdc1e924755b044f50aeb841a3a13
+size 627254

benchmark_v2_pca/S6_aligned_strict_linear_with_labels_pca10.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc03c7a51eddac88490194e000b7e42857a0a4c44428ca5aaf1ae6d20e056388
+size 249492

benchmark_v2_pca/S6_aligned_strict_linear_with_labels_pca20.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:392ae5cb4ac7fb074924d9ef8b211731075d1fe68d1bebfb25dcff2c4e55f661
+size 378779

benchmark_v2_pca/S6_aligned_strict_linear_with_labels_pca50.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90f31e25a5eed3285a0644433f7fdb811f2b9721ec30b4379eeb8548fcde6a33
+size 766689

benchmark_v2_pca/all_pca10.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:744990c13bdff07f2dad46e86b7b06536063d002165126cd293767e3995ea8e3
+size 1159256

benchmark_v2_pca/all_pca20.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21cdbc1bda2b9acfda2c3f9d04fca597d63646241b57167b0ff857d02dccd9a2
+size 1906313

benchmark_v2_pca/all_pca50.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:172ec805cd326f3cdb0aa0ce8d7a91749172406447e8d51eb964fa52413f3680
+size 4147705

data/w60_S1.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2499dc4fa8abd1f6f0691d523c4d6f8ec9e5356bc9350d85ad07d0aecaacfe2
+size 350830

data/w60_S2.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9c062a86d77db6071d67e789f5d227e55ae7101c87a6cdb9e89bd0c37e75eac
+size 407018

data/w60_S3.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afbc8c11ba23a07d0e2295375f7ed58538b8e9d855464b9ace863bebb78e4ce8
+size 355548

data/w60_S4.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16059747de2e0dd15e929d80bc2f1f656e94c330b74b0b459a321bea639c7028
+size 353309

data/w60_S5.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:854fc7f57f61890d381eb12e4fe71ed47e77b591d30a8914b9d4cc26a0f29640
+size 269179

data/w60_S6.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b93efd209ad572fd6467f824f33db8c648f5cd51e5e513223cafabc492a2c76
+size 345190

data/w60_all.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffcc49fc63e84663638499dff9c13bacc40eef42b821f2d250ca62839b7789a0
+size 1394399

folds/w60_loso_holdout_S1_test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2499dc4fa8abd1f6f0691d523c4d6f8ec9e5356bc9350d85ad07d0aecaacfe2
+size 350830

folds/w60_loso_holdout_S1_train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12567f437eb0b395356eadbfaa36bfe34f78a1b69041c55a405406c752683918
+size 1167456

folds/w60_loso_holdout_S2_test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9c062a86d77db6071d67e789f5d227e55ae7101c87a6cdb9e89bd0c37e75eac
+size 407018

folds/w60_loso_holdout_S2_train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3a9ccd2c5d2d0feced174f9628366b0fe750f3393bf6ddc9aa9b28306be6718
+size 1137274

folds/w60_loso_holdout_S3_test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afbc8c11ba23a07d0e2295375f7ed58538b8e9d855464b9ace863bebb78e4ce8
+size 355548

folds/w60_loso_holdout_S3_train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b76b3e202249fac52ee58dbe056ae3a6c24ab2b94a868a0858ea184c4fcdef11
+size 1203619

folds/w60_loso_holdout_S4_test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16059747de2e0dd15e929d80bc2f1f656e94c330b74b0b459a321bea639c7028
+size 353309

folds/w60_loso_holdout_S4_train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebb9be09e70b7efa1ffef8f9f81c1c6da1fb71d53ab712f00cde3a86e5b970b6
+size 1212195