datanalysis-credit-risk

The data cleaning pipeline consists of the following 11 steps, each executed independently without deleting the original data:

| getdataset() | Load and format data | references.func | | organalysis() | Organization sample analysis | references.func | | missingcheck() | Calculate missing rate | references.func | | dropabnormalym() | Filter abnormal months | references.analysis | | drophighmissfeatures() | Drop high missing rate features | references.analysis |

| droplowivfeatures() | Drop low IV features | references.analysis | | drophighpsifeatures() | Drop high PSI features | references.analysis | | drophighnoisefeatures() | Null Importance denoising | references.analysis | | drophighcorrfeatures() | Drop high correlation features | references.analysis |

用於貸前建模的信用風險資料清理和變數篩選管道。在處理建模前需要品質評估、缺失值分析或變數選擇的原始信用資料時使用。它涵蓋了資料載入和格式化、異常週期過濾、缺失率計算、高缺失變數移除、低IV變數過濾、高PSI變數移除、Null Importance去噪、高相關變數移除和清理報告產生。適用場景有信用風險資料清洗、變數篩選、貸前建模預處理。 來源：github/awesome-copilot。