메뉴
Use when starting on a new dataset, before modeling, or when asked to "explore", "profile", "understand", or "summarize" data. Covers structured EDA, distributions, correlations, target leakage checks, and visualization.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | exploratory-data-analysis |
| description | Use when starting on a new dataset, before modeling, or when asked to "explore", "profile", "understand", or "summarize" data. Covers structured EDA, distributions, correlations, target leakage checks, and visualization. |
EDA is the disciplined first pass over a dataset: understand shape, types, distributions, missingness, relationships, and red flags before writing a single model. Skipping it is the #1 cause of silent modeling failures (leakage, broken splits, garbage features).
Follow this order. Do not jump to modeling until every step is answered.
import pandas as pd
df = pd.read_csv("data.csv")
# 1. Shape & types
print(df.shape)
print(df.dtypes.value_counts())
print(df.memory_usage(deep=True).sum() / 1e6, "MB")
# 2. Missingness
miss = df.isna().mean().sort_values(ascending=False)
print(miss[miss > 0])
# 3. Target (classification example)
print(df["target"].value_counts(normalize=True))
# 4-5. Numeric summary + correlations with target
num = df.select_dtypes("number")
print(num.describe().T)
print(num.corr()["target"].sort_values(ascending=False))
# 6. Leakage red flag: |corr| ~ 1.0 with target
corr_t = num.corr()["target"].drop("target").abs()
print("LEAKAGE SUSPECTS:", corr_t[corr_t > 0.95].index.tolist())
For a fast automated first look, ydata-profiling (ProfileReport(df)) or df.describe(include="all") is acceptable — but never let a tool replace the manual leakage scan.
.corr() for non-linear / categorical relationships. Use mutual information or grouped means for those.Output of EDA should be: a short written summary (shape, target balance, top missing columns, leakage suspects, candidate features) that downstream steps (data-cleaning, feature-engineering) consume.
Use when preparing raw data for modeling — handling missing values, duplicates, inconsistent types, outliers, and bad categorical values. Emphasizes fitting all imputation on train-only to avoid leakage.
Use when running ML experiments that need to be compared, reproduced, or shared. Covers MLflow/Weights & Biases logging, what to track, run organization, and model registry basics.
Use when creating, encoding, scaling, or selecting features for ML models. Covers categorical encoding, numeric transforms, datetime/text/aggregation features, and leakage-safe target encoding.
Use when optimizing model hyperparameters. Covers search strategy (random vs Bayesian/Optuna), leakage-safe tuning inside CV, search-space design, early stopping, and budget management.
Use when the target is rare (fraud, churn, disease, anomalies). Covers correct metrics, resampling (SMOTE/undersampling), class weights, threshold tuning, and avoiding the accuracy trap and resampling leakage.
Use when fine-tuning a large language model. Covers choosing full vs LoRA/QLoRA, dataset formatting, the transformers/PEFT/TRL stack, key hyperparameters, and evaluating fine-tunes without overfitting.