Lesson · Hinglish · ⏱ 6 min padhne mein
🌲 Decision trees, random forests aur boosting
EnglishHinglish
🏠 Asli zindagi ki misaal
Bank officer loan ka faisla sawaalon ki series se karta hai: salary ₹50k se zyaada hai? Pehle koi default? Naukri 2 saal se zyaada? Yeh flowchart decision tree hai. 100 officers se poochna, jinhone thodi alag files dekhi hain, aur phir vote lena random forest hai. Boosting ek coaching class jaisa hai jahan har naya tutor un sawaalon pe focus karta hai jo pichhle students se galat hue the.
💡 Aasaan shabdon mein
Decision tree data ko simple yes/no sawaalon se split karta hai. Isse explain karna aasaan hai, lekin overfitting ka risk rehta hai. Random forest random samples pe bahut saare trees train karta hai aur unke votes ka average leta hai, jo zyaada accurate aur stable hai. Gradient boosting (XGBoost, LightGBM) trees ek ke baad ek banata hai, har naya tree pichhle ki galtiyaan theek karta hai. Tabular business data pe, jaise customers ya transactions ki spreadsheets, boosted trees aksar sabse accha perform karte hain, deep learning ko bhi peeche chhod dete hain.
🪜 Step by step
- Decision tree haan/nahi sawaal poochta hai: income > ₹50k hai? Customer ne payment miss kiya? Har jawaab agle sawaal tak le jaata hai, aakhir mein prediction.
- Ek gehra tree data ratt leta hai (overfit). Depth ya har leaf ke minimum examples limit karo.
- Random forest: random samples pe sainkdon trees train karo aur unke votes ka average lo. Zyaada stable.
- Gradient boosting (XGBoost, LightGBM): chhote trees ek ke baad ek banao, har ek pichhli galtiyaan sudhaarta hai. Tables pe aksar sabse accurate.
- Kuch key settings (trees ki ginti, depth, learning rate) validation data aur early stopping se tune karo.
🔑 Zaroori shabd
- Decision tree
- Yes/no sawaalon ka flowchart jo prediction tak le jaata hai.
- Random forest
- Bahut saare trees milkar vote karte hain, taaki jawaab zyaada stable ho.
- Gradient boosting
- Trees sequence mein bante hain, har ek pichhli galtiyaan sudhaarta hai.
👀 Example dekho
from xgboost import XGBClassifier
model = XGBClassifier(n_estimators=300, max_depth=4, learning_rate=0.05)
model.fit(X_train, y_train)
print(model.score(X_test, y_test))
# feature_importances_ shows which inputs mattered mostSpreadsheet jaise data ke liye ek strong default.
⚠️ Aam galtiyaan jinse bachna hai
- Early stopping ke bina bahut saare boosting rounds, jo overfit karta hai.
- Default feature importance pe aankh band karke bharosa; yeh zyaada unique values wale features ko favour kar sakta hai. SHAP ya permutation importance se cross check karo.
✅ Jaldi se check karo
30 columns wali table se customer churn predict karne ke liye ek strong pehla choice kya hai?
- Ek bahut bada deep neural network
- Chatbot
- Gradient-boosted trees
Jawaab dekho
C. Gradient-boosted trees: Tabular data pe boosted trees fast, accurate aur explainable hain. Deep learning se pehle inhe try karo.
🙋 Seekhne wale yeh poochhte hain
Kya trees ko scaled features chahiye?
Nahi. Trees thresholds pe split karte hain, isliye scaling se farak nahi padta. Isiliye yeh use karne mein aasaan hain.
Tables ke liye deep learning kyun nahi?
Aam business tables pe gradient boosting aksar utna hi ya zyaada accha, train karne mein tez aur samjhaane mein aasaan hota hai.
📌 Yaad rakho
Tables ho to pehle boosted trees try karo.
📖 Is lesson ke shabd (AI dictionary)
Poora course, aapke roadmap ke hisaab se
Progress tracking, quick checks aur step-by-step order ke saath. Yeh lesson in courses ka hissa hai:
Course mein seekho