Naive Bayes Classifier Kya Hai? Shuruaati Guide (Hindi)

Updated · By Viviai

Naive Bayes classifier ek supervised machine learning algorithm hai jo Bayes theorem par based hai aur kisi naye data point ki category ka anuman lagata hai, yeh maante hue ki uske features ek-doosre se conditionally independent hain.

Simple bhasha mein

Naive Bayes probability ka istemaal karta hai. Iske paas jo purane labeled examples hain, unse yeh seekhta hai ki har class ke liye alag-alag features kitne common hain, phir kisi naye point ko us class mein daalta hai jiski probability sabse zyada banti hai.

Bayes theorem ka role

Bayes theorem kisi hypothesis ki probability ko evidence ke aadhar par update karta hai. Basic relation is tarah likha jaata hai:

[ P(y \mid x) = \frac{P(x \mid y)\,P(y)}{P(x)} ]

Yahan (P(y)) prior probability hai, (P(x \mid y)) likelihood hai, aur (P(y \mid x)) posterior probability hai. Multiple features ke liye likelihood ko unke product se joda jaata hai.

'Naive' shabd kyun

Model yeh maan leta hai ki class diye jaane par saare features ek-doosre se independent hain. Real data mein yeh assumption aksar poori nahi hoti, isliye ise "naive" (भोला) kaha jaata hai. Is simple assumption ke bawajood algorithm text jaise high-dimensional tasks par kaafi achha kaam karta hai.

Main variants

Variant Kis data ke liye Example
Gaussian Naive Bayes Continuous numeric values Umar, aay jaisi sankhyaayein
Multinomial Naive Bayes Word counts Document mein shabd kitni baar aaye
Bernoulli Naive Bayes Binary (haan/na) features Shabd maujood hai ya nahi

Kahan istemaal hota hai

Istemaal Example
Spam filtering Email ko spam ya normal classify karna
Sentiment analysis Review positive ya negative hai
Document classification Article ko topic ke hisaab se baantna
Medical screening Lakshanon ke aadhar par category ka ishara

Zero probability ka dhyaan

Agar koi feature training data mein kisi class ke saath kabhi na dikhe, to uski probability zero ho sakti hai, jisse poora product zero ho jaata hai. Isse bachne ke liye smoothing (jaise Laplace smoothing) ka istemaal kiya jaata hai, jahan default alpha value 1 rakhi jaati hai.

Seemaaein

Independence assumption galat hone par probabilities overconfident ho sakti hain, isliye predicted probability ko actual result se compare karna zaroori hai. Naive Bayes aksar ek halka baseline maana jaata hai jise zyada complex models se compare kiya jaata hai.

FAQ

Naive Bayes supervised hai ya unsupervised?

Yeh supervised classification algorithm hai, kyunki yeh labeled data se seekhta hai.

Isse 'naive' kyun kaha jaata hai?

Kyunki yeh maanta hai ki saare features conditionally independent hain, jabki real data mein aisa aksar nahi hota.

Spam filtering mein yeh kaise kaam karta hai?

Yeh shabdon ki frequency se spam aur normal email ki probability gina kar naye email ko zyada probability wali class mein daalta hai.

Smoothing kyun zaroori hai?

Smoothing zero probability se bachati hai, warna ek hi absent feature poore prediction ko zero bana sakta hai.

Mukhya baatein

Aage padhiye machine learning kya hai aur logistic regression kya hai.

Sources

  1. https://www.datacamp.com/hi/tutorial/naive-bayes-scikit-learn
  2. https://www.primelearnpoint.com/machine-learning/naive-bayes-classifier/

Learn by doing

Read a guide, then practice with bite-sized AI lessons you can take offline.

Start learning free