Python ML Cheat Sheet Repo：涵蓋完整機器學習流程的速查表

Every ML beginner hits the same wall: you know what you want to do, but you can't remember the exact syntax.

Is it fit_transform or transform? Which scaler do I use for outliers? How do I plot a confusion matrix again?

I got tired of Googling the same things, so I built a repo that has all of it in one place.

🔗 github.com/Kavi274/python-ml-cheatsheets

What's Inside

9 cheat sheets covering the full ML workflow:

1. NumPy

Arrays, indexing, broadcasting, linear algebra — the foundation of everything.

python
a = np.array([[1, 2, 3], [4, 5, 6]])
a[a > 3]              # [4, 5, 6]
a.reshape(3, 2)
np.dot(A, B)          # matrix multiplication

2. Pandas

DataFrames, groupby, merge, datetime, string ops.

df.groupby("city").agg({"sales": "sum", "profit": "mean"})
df.isnull().sum()
pd.merge(df1, df2, on="id", how="left")

3. Scikit-learn

Every major model, pipelines, cross-validation, and hyperparameter tuning.

pipe = Pipeline([
    ("scaler", StandardScaler()),
    ("model", RandomForestClassifier())
])
pipe.fit(X_train, y_train)

4. Matplotlib & Seaborn

Every plot type with copy-paste code — histograms, heatmaps, ROC curves, feature importance.

sns.heatmap(df.corr(), annot=True, fmt=".2f", cmap="coolwarm")

5. Data Preprocessing

Missing values, scaling, encoding, class imbalance — the full cleaning pipeline.

num_pipeline = Pipeline([
    ("impute", SimpleImputer(strategy="median")),
    ("scale", StandardScaler())
])

6. Feature Engineering

Datetime features, target encoding, cyclical encoding, PCA, SHAP-based selection.

df["month_sin"] = np.sin(2 * np.pi * df["month"] / 12)
df["month_cos"] = np.cos(2 * np.pi * df["month"] / 12)

7. Model Evaluation

ROC, precision-recall curves, confusion matrix, optimal threshold, SHAP explainability, calibration.

fpr, tpr, _ = roc_curve(y_test, y_prob)
roc_auc = auc(fpr, tpr)

8. Deep Learning (Keras)

CNNs, RNNs, transfer learning, callbacks, tf.data pipelines.

base_model = keras.applications.EfficientNetB0(
    include_top=False, weights="imagenet"
)
base_model.trainable = False

9. Statistics for ML

Distributions, hypothesis tests, A/B testing, Bayes' theorem.

stat, p = stats.ttest_ind(group1, group2, equal_var=False)
print(f"p={p:.4f} — {'Significant' if p < 0.05 else 'Not significant'}")

---
How to Use It

No setup. No notebooks to run. Every cheat sheet is plain Markdown — just open it on GitHub and hit Ctrl+F.

Clone it locally if you want it offline:

git clone https://github.com/Kavi274/python-ml-cheatsheets.git

---
What's Next

Planning to add:
- PyTorch
- XGBoost / LightGBM
- Time Series (Prophet, statsmodels)
- HuggingFace Transformers

PRs are welcome — if you know a topic well, add a cheat sheet!

---
If this saves you even one Google search, consider leaving a ⭐  on GitHub — it helps others find it.

🔗 https://github.com/Kavi274/python-ml-cheatsheets