Scribd
Upload
44 views

Naïve Bayes Classifier: April 25, 2006

The document discusses Naive Bayes classifiers and their use in document classification. It begins by introducing classification methods like manual classification, rule-based classification, and supervised machine learning classification. It then focuses on Naive Bayes classifiers, explaining that they are a simple and commonly used supervised learning method that uses Bayes' theorem. The document outlines the parameters and assumptions of Naive Bayes classifiers, including conditional independence of features. It discusses properties like incremental learning, combining prior and observed data, and probabilistic outputs. Finally, it covers techniques like log probability calculations to prevent underflow and Laplace smoothing to avoid overfitting.

Uploaded by

aaminj
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
44 views

Naïve Bayes Classifier: April 25, 2006

The document discusses Naive Bayes classifiers and their use in document classification. It begins by introducing classification methods like manual classification, rule-based classification, and supervised machine learning classification. It then focuses on Naive Bayes classifiers, explaining that they are a simple and commonly used supervised learning method that uses Bayes' theorem. The document outlines the parameters and assumptions of Naive Bayes classifiers, including conditional independence of features. It discusses properties like incremental learning, combining prior and observed data, and probabilistic outputs. Finally, it covers techniques like log probability calculations to prevent underflow and Laplace smoothing to avoid overfitting.

Uploaded by

aaminj
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 19

Naïve Bayes Classifier

April 25th, 2006


Classification Methods (1)
 Manual classification
 Used by Yahoo!, Looksmart, about.com, ODP
 Very accurate when job is done by experts
 Consistent when the problem size and team is
small
 Difficult and expensive to scale
Classification Methods (2)
 Automatic classification
 Hand-coded rule-based systems
 One technique used by CS dept’s spam filter, Reuters,
Snort IDS …
 E.g., assign category if the instance matches the rules
 Accuracy is often very high if a rule has been carefully
refined over time by a subject expert
 Building and maintaining these rules is expensive
Classification Methods (3)
 Supervised learning of a document-label assignment function
 Many systems partly rely on machine learning (Google, MSN,
Yahoo!, …)
 Naive Bayes (simple, common method)
 k-Nearest Neighbors (simple, powerful)
 Support-vector machines (new, more powerful)
 … plus many other methods
 No free lunch: requires hand-classified training data
 But data can be built up (and refined) by amateurs

 Note that many commercial systems use a mixture of methods


Decision Tree
 Strength
 Decision trees are able to generate understandable rules.
 Decision trees perform classification without requiring
much computation.
 Decision trees are able to handle both continuous and
categorical variables.
 Decision trees provide a clear indication of which fields are
most important for prediction or classification
 Weakness
 Error-prone with many classes
 Computationally expensive to train, hard to update
 Simple true/false decision, nothing in between
Does patient have cancer or not?
 A patient takes a lab test and the result comes back positive.
It is known that the test returns a correct positive result in
only 99% of the cases and a correct negative result in only
95% of the cases. Furthermore, only 0.03 of the entire
population has this disease.

How likely that this patient has cancer?


Bayesian Methods
 Our focus this lecture
 Learning and classification methods based on
probability theory.
 Bayes theorem plays a critical role in probabilistic
learning and classification.
 Uses prior probability of each category given no
information about an item.
 Categorization produces a posterior probability
distribution over the possible categories given a
description of an item.
Basic Probability Formulas
 Product rule
P ( A  B)  P( A | B) P( B )  P( B | A) P( A)
 Sum rule
P ( A  B )  P ( A)  P ( B )  P ( A  B )
 Bayes theorem
P ( D | h) P ( h)
P(h | D) 
P( D)
 Theorem of total probability, if event Ai is
mutually exclusive and probability sum to 1
n
P( B )   P ( B | Ai ) P ( Ai )
i 1
Bayes Theorem
 Given a hypothesis h and data D which bears on the
hypothesis: P ( D | h) P ( h)
P (h | D) 
P( D)
 P(h): independent probability of h: prior probability
 P(D): independent probability of D
 P(D|h): conditional probability of D given h:
likelihood
 P(h|D): conditional probability of h given D: posterior
probability
Does patient have cancer or not?
 A patient takes a lab test and the result comes back positive. It is
known that the test returns a correct positive result in only 99% of
the cases and a correct negative result in only 95% of the cases.
Furthermore, only 0.03 of the entire population has this disease.

1. What is the probability that this patient has cancer?


2. What is the probability that he does not have cancer?
3. What is the diagnosis?
Maximum A Posterior
 Based on Bayes Theorem, we can compute the
Maximum A Posterior (MAP) hypothesis for the data
 We are interested in the best hypothesis for some space
H given observed training data D.

hMAP  argmax P ( h | D )
hH

P ( D | h ) P ( h)
 argmax
hH P ( D)
 argmax P ( D | h) P (h)
hH
H: set of all hypothesis.
Note that we can drop P(D) as the probability of the data is constant
(and independent of the hypothesis).
Maximum Likelihood
 Now assume that all hypotheses are equally
probable a priori, i.e., P(hi ) = P(hj ) for all hi,
hj belong to H.
 This is called assuming a uniform prior. It
simplifies computing the posterior:
hML  arg max P ( D | h)
hH

 This hypothesis is called the maximum


likelihood hypothesis.
Desirable Properties of Bayes Classifier
 Incrementality: with each training example,
the prior and the likelihood can be updated
dynamically: flexible and robust to errors.
 Combines prior knowledge and observed
data: prior probability of a hypothesis
multiplied with probability of the hypothesis
given the training data
 Probabilistic hypothesis: outputs not only a
classification, but a probability distribution
over all classes
Bayes Classifiers
Assumption: training set consists of instances of different classes
described cj as conjunctions of attributes values
Task: Classify a new instance d based on a tuple of attribute values
into one of the classes cj  C
Key idea: assign the most probable class c MAP using Bayes
Theorem.

cMAP  argmax P (c j | x1 , x2 , , xn )
c j C

P ( x1 , x2 , , xn | c j ) P(c j )
 argmax
c j C P( x1 , x2 ,  , xn )
 argmax P( x1 , x2 ,  , xn | c j ) P(c j )
c j C
Parameters estimation
 P(cj)
 Can be estimated from the frequency of classes in the
training examples.
 P(x1,x2,…,xn|cj)
 O(|X|n•|C|) parameters
 Could only be estimated if a very, very large number of
training examples was available.
 Independence Assumption: attribute values are
conditionally independent given the target value: naïve
Bayes. P( x , x , , x | c )  P( x | c )
1 2 n j 
i
i j

c NB  arg max P (c j ) P ( xi | c j )
c j C i
Properties
 Estimating P( xi | c j ) instead of P( x1 , x2 , , xn | c j ) greatly
reduces the number of parameters (and the data
sparseness).
 The learning step in Naïve Bayes consists of
estimating P( xi | c j ) and P(c j ) based on the
frequencies in the training data
 An unseen instance is classified by computing the
class that maximizes the posterior
 When conditioned independence is satisfied, Naïve
Bayes corresponds to MAP classification.
Question: For the day <sunny, cool, high, strong>, what’s
the play prediction?
Underflow Prevention
 Multiplying lots of probabilities, which are
between 0 and 1 by definition, can result in
floating-point underflow.
 Since log(xy) = log(x) + log(y), it is better to
perform all computations by summing logs of
probabilities rather than multiplying
probabilities.
 Class with highest final un-normalized log
probability score is still the most probable.
c NB  argmax log P(c j ) 
c jC
 log P( x | c )
i positions
i j
Smoothing to Avoid Overfitting
N ( X i  xi , C  c j )  1
Pˆ ( xi | c j ) 
N (C  c j )  k
# of values of Xi
 Somewhat more subtle version overall fraction in
data where Xi=xi,k

N ( X i  xi ,k , C  c j )  mpi ,k
Pˆ ( xi ,k | c j ) 
N (C  c j )  m
extent of
“smoothing”

You might also like