Okay then we have this decision tree algorithm for decision tree, we will only be talking about this ID tree algorithm. So, Id tree algorithm uses our entropy and information gain to select a variable to build a decision tree. So, entropy can be the formula is something like this entropy of RSA is equal to some n minus probability law a lot to the to base to probability. So, entropy of prey guava attribute or variable equal to entropy phi equal to entropy 0.36 comma 0.64 equal to minus 0.36 law to 0.36 minus 0.64 a lot to 2.64 equals 0.94 So, in prego via yes is nigh noise phi. So 0.36 is actually our pi divided by nine plus pi equal to five divided by 14. So is 0.362 point Paul says why is nine divided by nine plus by is equal to nine divided by 40 is equal to 0.64.

So, I will say, essentially a probability then we look into the entropy aka two variables that the entropy of sn x equal to some of our probability C and entropy c. So, we have a frequency tables. So, let's say we have two variables Although and the protocol so entropy of play golf are low equal to probability of sunny time is the entropy of tree to brass probability of overcast times the entropy for zero brass probably 80 of rainy times the entropy of two, three. So, probability of Sunny, so sunny, you have five Sunny, divided by overall pi over 14. So four, five plus four plus five is 14. So five divided by 14 is the probability of sunny times the entropy of a tree to grass the probability of overcast will be four over 14 times the entropy of four zero brass the probability of rainy should be fine over 14 times the entropy of our two tree So, we will get around 0.6934 entropy of two variables.

So, for information gained information gain is a gain of s s equal to entropy of s minus entropy s s. So, are we have RSA information gain or precog and outlook is equal to entropy or prego minus entropy or pre growth and outlook. So, we have 0.940 minus 0.693 equal to 0.247. So, we calculate all the information gain for all the variables. So, for our Do we have information gain of 0.247 temperature our variable information gain is 0.029 Humidity variable information gain is 0.15 to windy variable information gain is 0.04. So, for how low we select our variable because our variable has the highest information gain, so 0.247 is the highest. So we select these are all available as the root note and then we split the data into three data set.

So are low then we have a tree data tree data set here. So far these are the data here. Although variable is equal to sunny bodies, our data set here or the hollow variable is equal to overcast for all the data here, although is equal to rainy So, after we have our loop and we had our tree data set, so we try to print the data again. So let's say for how low for overcast we do not need to spray because overcast, entropy is zero, entropy is zero because all the pre golf all the values here is equal to Yes. Then, for sunny and rainy we need to spray the low or we need to spray the data. So, for sunny we will calculate all the information gain of all the attributes or variables again, very select attributes or variables with the highest are information gain.

And then from there we will select windy, windy has the highest information gain. So, from this sunny data set after we calculate all the information gain for all the variables and attributes, we select when the variables because when the variables has a highest information gain, then from this windy variable we split the data set into false and true again then bodies are any data set we will calculate all the information gain of all the variables and then we will select the variables into highest information gain. So, we select humidity because our humidity variable has the highest information gain there is greater data into higher normal So, we will Continue to spray and we will continue to calculate all the information gain and select a variable with the highest information gain. Click continue and repeating until you are the whole decision trees have been being developed or been built.

And from this decision tree, we can actually come out or the rival bodies are loose. So in this decision tree we can derive out all these rules. So let's say our a ALU equal to sunny, windy equal to false praise equal to Yes, if our equal to sunny windy equal to true then praise equal to know if other equal to overcast spray equal to Yes, it all equal to rainy he Immediately equal to high grade equal to no it all equal to rainy humidity equal to normal pre equal to Yes. So, based on all these rules, we can actually use all these rules to let's say our classify our predictor variable. So for decision tree, so let's see, how our decision tree can classify a variable or predict a variable ah the decision tree classifier variable pretty variable essentially, based on the rules that can be derived from the decision tree.

So he is actually based on all these rules here. So for decision tree for AI D tree algorithm, we use information gain to select variables or attributes in some other algorithms or decision tree like a chat The algorithm you'll use Ty square test to select the variables to build this our decision tree