Lê Đăng Khoa (Liam) commited on
Commit
b3fce1a
·
verified ·
1 Parent(s): 96d6173

RLHF training update v1.1

Browse files
v1.1/README.md ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ tags:
4
+ - driver-behavior
5
+ - obd-data
6
+ - xgboost
7
+ - rlhf
8
+ - reinforcement-learning
9
+ ---
10
+
11
+ # Driver Behavior Classification Model (RLHF v1.1)
12
+
13
+ This model classifies driver behavior based on OBD (On-Board Diagnostics) sensor data using XGBoost.
14
+
15
+ ## Model Information
16
+
17
+ - **Model Type**: xgboost_classifier
18
+ - **Version**: 1.1
19
+ - **Created**: 2025-10-01T06:26:14.198564
20
+ - **Framework**: xgboost
21
+ - **Task**: driver_behavior_classification
22
+
23
+ ## Performance Metrics
24
+
25
+ - **accuracy**: 0.8814
26
+ - **cv_mean**: 0.5067
27
+ - **cv_std**: 0.1100
28
+ - **cv_scores**: [0.35789473684210527, 0.4842105263157895, 0.5425531914893617, 0.4574468085106383, 0.6914893617021277]
29
+
30
+ ## Training Data
31
+
32
+ - **Datasets Used**: 2
33
+ - **Total Samples**: 472
34
+ - **Training Date**: 2025-10-01T06:26:14.198509
35
+
36
+ ## Labels
37
+
38
+ The model predicts one of the following driver behavior categories:
39
+ - aggressive
40
+ - normal
41
+ - conservative
42
+
43
+ ## Usage
44
+
45
+ ```python
46
+ import joblib
47
+ import pandas as pd
48
+
49
+ # Load the model
50
+ model = joblib.load('xgb_drivestyle_ul.pkl')
51
+ label_encoder = joblib.load('label_encoder_ul.pkl')
52
+ scaler = joblib.load('scaler_ul.pkl')
53
+
54
+ # Prepare your OBD data
55
+ # (Ensure features match the training data format)
56
+
57
+ # Make predictions
58
+ predictions = model.predict(scaled_data)
59
+ behavior_labels = label_encoder.inverse_transform(predictions)
60
+ ```
61
+
62
+ ## Files
63
+
64
+ - `xgb_drivestyle_ul.pkl`: Main XGBoost model
65
+ - `label_encoder_ul.pkl`: Label encoder for behavior categories
66
+ - `scaler_ul.pkl`: Feature scaler
67
+ - `metadata.json`: Model metadata and performance metrics
68
+
69
+ ## RLHF Training
70
+
71
+ This model was trained using Reinforcement Learning from Human Feedback (RLHF) to improve performance based on human-labeled data and feedback.
v1.1/label_encoder_ul.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e510340f497abdf3b0da0f87274d35dae07c592c5a5051ed67fd3eac7b8e4f19
3
+ size 534
v1.1/metadata.json ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "xgboost_classifier",
3
+ "version": "1.1",
4
+ "created_at": "2025-10-01T06:26:14.198564",
5
+ "training_data": {
6
+ "datasets": [
7
+ "labeled/001_processed-004_2025-10-01-labelled.csv",
8
+ "labeled/002_processed-002_2025-10-01-labelled.csv"
9
+ ],
10
+ "total_samples": 472,
11
+ "training_date": "2025-10-01T06:26:14.198509",
12
+ "features_count": 95
13
+ },
14
+ "performance_metrics": {
15
+ "accuracy": 0.8813559322033898,
16
+ "cv_mean": 0.5067189249720044,
17
+ "cv_std": 0.110006561067466,
18
+ "cv_scores": [
19
+ 0.35789473684210527,
20
+ 0.4842105263157895,
21
+ 0.5425531914893617,
22
+ 0.4574468085106383,
23
+ 0.6914893617021277
24
+ ]
25
+ },
26
+ "framework": "xgboost",
27
+ "task": "driver_behavior_classification",
28
+ "labels": [
29
+ "aggressive",
30
+ "normal",
31
+ "conservative"
32
+ ],
33
+ "features": "obd_sensor_data",
34
+ "rlhf_metadata": {
35
+ "labeled_samples": 472,
36
+ "original_samples": 0,
37
+ "model_confidence": 0.0,
38
+ "datasets_processed": 2
39
+ }
40
+ }
v1.1/scaler_ul.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a9cf55feb20e9616ded2a1e778a3efc2976a88e1b3d3f2924546cc7f6cca11c
3
+ size 2879
v1.1/xgb_drivestyle_ul.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1959949e16e1a68b761c2da1fa46e7decac5f5ad2c5908bf080923733179b3d
3
+ size 1993173