Laplace smoothing | Laplace Correction | Zero Probability in Naive Bayes Classifier by Mahesh Huddar

Поділитися
Вставка
  • Опубліковано 29 лис 2024

КОМЕНТАРІ • 16

  • @bathalamallikarjuna2316
    @bathalamallikarjuna2316 11 місяців тому +2

    Great work sir...💖

    • @MaheshHuddar
      @MaheshHuddar  11 місяців тому

      Thank You
      Please do like share and subscribe

  • @sagaradoshi
    @sagaradoshi Рік тому +4

    Hi, Thanks for this excellent video. I have a question that: Can we apply Laplace transform only to p(outlook = Overcast| no)? or should we also apply to other terms of product like p(temp = cool | no), p(hum = hight | no) and p(wind = strong | no) although the other terms are not 0

    • @dastan-e-lahore2487
      @dastan-e-lahore2487 Рік тому +1

      i have same question

    • @adarshshukla2055
      @adarshshukla2055 Рік тому +2

      @@dastan-e-lahore2487 that term only in case of both P(overcast/yes) and P(overcast/No)

    • @akshitajha6382
      @akshitajha6382 Рік тому +1

      It's needless to apply there
      It's only probability 0 problem ig

    • @tuananhtran5071
      @tuananhtran5071 Рік тому

      @@akshitajha6382 so what is the answer, i don't quite get it. If we have p(outlook = Overcast/no), we should use Laplace for p(outook = sunny/ no) and P(rainy/no) too.

    • @tuananhtran5071
      @tuananhtran5071 Рік тому

      But if i have (sunny, hot, high, strong), there is no zero probability here, so when i calculate, do i need to use the laplace for sunny or just take the initial value (2/9, 3/5)???. Please reply me as soon as possible

  • @xyz_12381
    @xyz_12381 Місяць тому

    Sir , there is a confusion, when you changed the prob for no class, is there an impact of yes class , as the counts also incresed that class also , or that will remain same for yes class, please clear 🙏🏻

    • @TabFamily-c1g
      @TabFamily-c1g 12 днів тому

      Yes, you also have to apply Laplace smoothing in overcast = yes so they will be equal

  • @faisalhawlader6822
    @faisalhawlader6822 Рік тому

    Sir, will the other probabilities remain same? or it will change due to inserting new record.

    • @Samurai_Jack__
      @Samurai_Jack__ Рік тому +2

      all the probabilities for outlook feature should change so that to keep their summation 1

  • @soapmactavish6001
    @soapmactavish6001 Рік тому

    God 🙌

    • @MaheshHuddar
      @MaheshHuddar  Рік тому

      Thank You
      Do like share and subscribe

    • @tuananhtran5071
      @tuananhtran5071 Рік тому

      @@MaheshHuddar If we have p(outlook = Overcast/no), we should use Laplace for p(outook = sunny/ no) and P(rainy/no) too. But if i have (sunny, hot, high, strong), there is no zero probability here, so when i calculate, do i need to use the laplace for sunny or just take the initial value (2/9, 3/5)???. Please reply me as soon as possible

    • @elurijahnavi
      @elurijahnavi Рік тому

      @@tuananhtran5071 actually p(outlook= overcast/no) + p(outlook= sunny/ no) + p(Outlook=rainy/ no) = 1
      When we are smoothing the probability mass from large probabilities are given to the smaller ones. So, we have to smooth posterior probabilities for the Outlook feature conditioned on play = no