Machine Learning For Beginners: A Complete Overview for Understand How to Build Artificial Intelligence Through Data Science. by Mendez Richard

Author:Mendez, Richard
Language: eng
Format: epub, pdf
Published: 2021-10-23T00:00:00+00:00

The same may be said for the Split on Class:

The Gini coefficient for sub node Class IX is (0.43)*(0.43) + (0.57)*(0.57) =0.51. (0.56)*(0.56) + (0.44)*(0.44) = 0.51 Gini coefficient for sub node Class X

Calculate the weighted Gini coefficient for Split Class = (14) of 30. *0.51 plus (16/30)*0.51 equals 0.51. As you can see in the chart above, the Gini score for Split on Gender is higher than the Gini score for Split on Age.

As a result, the node split will occur on Gender, rather than on Class.

Chi-Square

This algorithm is used to determine whether or not the differences between the parent node and the sub-node are statistically significant. In order to calculate it, we need to add up each and every square of each and every difference between expected frequencies of a target variable and observed frequencies of that same target variable.

The target variable "Success" or "Failure" is a categorical variable with two possible outcomes. It has the capability of performing two or more splits.

The greater the value of the Chi-Square, the greater the value of the statistical significance of differences between the parent node and the sub-node between the parent node and the sub-node.

Calculating the Chi-Square of each node is accomplished using the following formula: Chi-square = ((Actual – Expected)2 Expected)12 = ((Actual – Expected)2 Expected)12

The formula results in the formation of a tree known as CHAID (Chi-square Automatic Interaction Detector) The following are the procedures that must be completed in order to compute the Chi-square for a split:

The Chi-square for a single node is derived by computing the deviations for both 'Success' and 'Failure' outcomes for that node.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.