# ADAlgorithm

**Repository Path**: nie-shiqin/adalgorithm

## Basic Information

- **Project Name**: ADAlgorithm
- **Description**: 异常检测算法库
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-12-23
- **Last Updated**: 2025-03-20

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# ADAlgorithm

#### 介绍
异常检测算法库。


##### DCdetector:
export CUDA_VISIBLE_DEVICES=2

###### SWaT:
export CUDA_VISIBLE_DEVICES=2  # 不然会爆显存

python main.py --model DCdetector --dataset SWaT --lookback 105 --epochs 3 --bs 64  --lr 1e-4  --device cuda:0  --save_path  ../checkpoints/DCdetector/  --output_save_path  ../results/DCdetector/  --anormly_ratio 1  --input_c 51  --output_c 51  --loss_fuc MSE --patch_size 357

###### SMAP:
python main.py --model DCdetector --dataset SMAP --lookback 105 --epochs 3 --bs 128  --lr 1e-4  --device cuda:0  --save_path  ../checkpoints/DCdetector/  --output_save_path  ../results/DCdetector/  --anormly_ratio 0.85  --input_c 25  --output_c 25  --loss_fuc MSE --patch_size 357

##### AnomalyTransformer:
export CUDA_VISIBLE_DEVICES=2

###### SWaT:
python main.py --model AnomalyTransformer --dataset SWaT --lookback 100 --epochs 10 --bs 256  --lr 1e-4  --device cuda:0  --anormly_ratio 1  --input_c 51  --output_c 51

Best Performance:
AUPR: 0.4554
AUCROC: 0.8000761805057357
PR:0.6428571304467736, REC:0.5893805205419377, F1:0.6149534468543073, TP:333.0, TN:3602.0,FP:185.0,FN:232.0,Macro_F1:0.7801156079400238

###### SMAP:
export CUDA_VISIBLE_DEVICES=2  # 不然会爆显存

python main.py --model AnomalyTransformer --dataset SMAP --lookback 100 --epochs 10 --bs 256  --lr 1e-4  --device cuda:0  --anormly_ratio 1  --input_c 25  --output_c 25

Best Performance:
AUPR: 0.2243
AUCROC: 0.6490458757523604
PR:0.2365721168583457, REC:0.6689419681067924, F1:0.3495280138385374, TP:392.0, TN:2245.0,FP:1265.0,FN:194.0,Macro_F1:0.5521359353233308

###### MSL:
python main.py --model AnomalyTransformer --dataset MSL --lookback 100 --epochs 10 --bs 256  --lr 1e-4  --device cuda:0  --anormly_ratio 1  --input_c 55  --output_c 55

Best Performance:
AUPR: 0.2668
AUCROC: 0.6171980178167339
PR:0.2676056149573511, REC:0.46913574455114265, F1:0.3408025185127221, TP:38.0, TN:327.0,FP:104.0,FN:43.0,Macro_F1:0.5786384639516604

###### PSM:
python main.py --model AnomalyTransformer --dataset PSM --lookback 100 --epochs 10 --bs 256  --lr 1e-4  --device cuda:0  --anormly_ratio 1  --input_c 25  --output_c 25

Best Performance:
AUPR: 0.2721
AUCROC: 0.42624521072796934
PR:0.32348804045727086, REC:0.9349593115870198, F1:0.4806649269778663, TP:230.0, TN:41.0,FP:481.0,FN:16.0,Macro_F1:0.3111433179002132

###### SMD machine-1-1:
python main.py --model AnomalyTransformer --dataset machine-1-1 --lookback 100 --epochs 10 --bs 256  --lr 1e-4  --device cuda:0  --anormly_ratio 0.5  --input_c 38  --output_c 38

Best Performance:
AUPR: 0.2873
AUCROC: 0.3712332589285714
PR:0.5333329777780148, REC:0.2499999218750244, F1:0.3404210412500942, TP:8.0, TN:217.0,FP:7.0,FN:24.0,Macro_F1:0.6368746705748135

**COUTA_result:**

dataset: SMD:

```
['machine-1-1_train.pkl','machine-1-1_test.pkl','machine-1-1_test_label.pkl']
```

F1指标 [0.9994385169992943]

aupr指标 [0.9996500199047742]

```
['machine-1-2_train.pkl','machine-1-2_test.pkl','machine-1-2_test_label.pkl']
```

F1指标 [0.9256995536719107]
aupr指标 [0.9190661287117381]

```
['machine-2-1_train.pkl','machine-2-1_test.pkl','machine-2-1_test_label.pkl']
```

F1指标 [0.9793460334408411]
aupr指标 [0.9955544808869823]



**CARLA_result**

dataset: MSL

entities：M-1（55维度） 

F1 0.9230769230769231
precision 0.9970149253731343
recall 0.8593481989708405

entities：M-2（55维度）

F1 0.9192100538599641
precision 0.9642184557438794
recall 0.8782161234991424