No Description

liudawei 8e5729bf39 Merge branch 'dev_david' of anweiguo/algorithm_platform into dev_awg 3 months ago
common c3d0a750f7 awg commit algorithm components 3 months ago
data_processing bcf3bd21ec test 3 months ago
evaluation_processing eaa15a730c awg commit algorithm components 3 months ago
models_processing 369873574c Merge branch 'dev_awg' of anweiguo/algorithm_platform into dev_david 3 months ago
post_processing 1f99c93c25 awg commit algorithm components 3 months ago
test_unit 3aa4c4df29 awg commit 4 months ago
.gitignore 3aa4c4df29 awg commit 4 months ago
Dockerfile 5d810c3670 01031453 4 months ago
README.md 1a582bb983 awg commit algorithm components 4 months ago
requirements.txt f7d2f857f6 01171710 3 months ago
run_all.py 1f99c93c25 awg commit algorithm components 3 months ago

README.md

算法平台

该平台集成了每人经反复验证过的、有效的、通用的数据处理组件与算法组件,提升了组件的可读性、复用性、扩展性与可维护性。

1.代码目录结构

按照组件功能划分目录结构

数据处理 data_processing

  • [异常处理] abnormal_processing
  • [限电处理] processing_limit_power
  • ·agc-avc清洗 ·样板机法清洗 ·限电记录清洗 ·统计法清洗-光 ·统计法清洗-风

特征工程 feature_processing

  • [归一化] feature_normalizer
  • ·‌Min-Max归一化 ·‌Z-Score归一化
  • [PCA降维] feature_pca
  • [特征选择] feature_select
  • [特征变换] feature_transform

模型 model_processing

  • [模型训练] model_train
  • ·‌LSTM ·‌CNN ·DNN ·Lightgbm ·Xgboost ·RandomForest
  • [模型预测] model_predict

模型后处理 post_processing

评估 evaluation_processing

  • 准确率计算
  • 分析报告

2.代码测试

1.Docker镜像为python3.7,开发完后,需要做单元测试,requirements.txt加入所需要的包,run_all.py中services中加入启动程序的路径和端口。

2.启动run_all,可采用postman或代码等其他形式进行接口调用测试。测试通过方可提交。