ClassPriors

✖
`ClassPriors`

✖

is an option for Classify and related functions that specifies explicit prior probabilities to assume for output classes, independent of anything deduced from the training set.

Details

ClassPriors<|class₁p₁,class₂p₂,…|> specifies that the prior probability to classify a particular input as being in class class_i should be p_i.
By default, the class priors are computed from the class frequencies of the training set.
In ClassPriors<|class₁p₁,class₂p₂,…|>, unspecified class priors will be set to their rescaled default value to ensure normalization of the distribution.

Examples

open allclose all

Basic Examples (1)Summary of the most common use cases

Train a classifier on an unbalanced training set:

In[1]:=1

✖

https://wolfram.com/xid/0btnc8ay4rdg-vvjkap

In[2]:=2

✖

https://wolfram.com/xid/0btnc8ay4rdg-jgw9q2

Out[2]=2

Obtain class probabilities for a new example:

In[3]:=3

✖

https://wolfram.com/xid/0btnc8ay4rdg-pzmtax

Out[3]=3

Set a uniform prior distribution for the classes:

In[4]:=4

✖

https://wolfram.com/xid/0btnc8ay4rdg-3t4cal

Out[4]=4

Set the prior distribution of classes inside the classifier:

In[5]:=5

✖

https://wolfram.com/xid/0btnc8ay4rdg-252zkb

Out[5]=5

In[6]:=6

✖

https://wolfram.com/xid/0btnc8ay4rdg-no60bq

Out[6]=6

Priors of unspecified classes are functions of the remaining probability mass:

In[7]:=7

✖

https://wolfram.com/xid/0btnc8ay4rdg-i5vc31

Out[7]=7

In[8]:=8

✖

https://wolfram.com/xid/0btnc8ay4rdg-svuflv

Out[8]=8

Applications (3)Sample problems that can be solved with this function

Restrict the built-in language classifier to only two languages:

In[1]:=1

✖

https://wolfram.com/xid/0btnc8ay4rdg-1v24g1

Out[1]=1

Train a classifier to recognize if an organism is diseased on a balanced training set:

In[1]:=1

✖

https://wolfram.com/xid/0btnc8ay4rdg-3ot9yq

Out[1]=1

Find the probability that a new organism is diseased:

In[2]:=2

✖

https://wolfram.com/xid/0btnc8ay4rdg-xfsyy0

Out[2]=2

Find the probability that the same organism is diseased, incorporating prior knowledge that diseased organisms represent only 1% of the population:

In[3]:=3

✖

https://wolfram.com/xid/0btnc8ay4rdg-e93hbc

Out[3]=3

Train a classifier able to predict if a website user will buy a given item based on his behavior:

In[1]:=1

✖

https://wolfram.com/xid/0btnc8ay4rdg-w7fes8

Out[1]=1

Define and visualize a model for the probability of buying the item as a function of the outside temperature:

In[2]:=2

✖

https://wolfram.com/xid/0btnc8ay4rdg-lrelis

In[3]:=3

✖

https://wolfram.com/xid/0btnc8ay4rdg-bvhlwv

Out[3]=3

Use this model and WeatherData to increase the prediction performance for a Parisian user, without additional training:

In[4]:=4

✖

https://wolfram.com/xid/0btnc8ay4rdg-tb94fq

Out[4]=4

Top