How to Plot an ROC Curve in Python | Machine Learning in Python

Key Takeaways

welcome back to the beta professor YouTube channel in this video I'm going to show you how to create the RLC curve which you can use to compare the performance of different machine learning models so without further ado let's get started so the first thing that you want to do is head over to the github of the data professor click on the code repository go to Python and then find the RLC curve click on that right click on the raw link and save a local copy to your computer or you could also follow along by looking at the github right save link s and save it to your computer so if you would like to use the Google collab you can also free free to do so as well so first thing that you want to do is click on the github and then type in data professor enter and then find our LLC curve click on that and so today is going to be machine learning and python series and we're gonna create receiver operating characteristic curve or shortly known as roc okay and today we're not going to use the irish dataset but we're going to use a synthetic data set so in a nutshell the ROC curve will summarize the prediction performance of a classification model at all classification thresholds as a function of the true positive rate and false positive rates okay and the true positive rate will be on the y axis and the false positive rate will be on the x axis and for the true positive rate it is also known as sensitivity and for the false positive rate it is also known as 1 minus specificity okay and the equation is provided right here okay so let's stop on to the next step is to generate the synthetic data set and for this one we're going to use the make classification function of scikit-learn and also we're going to use the numpy and actually number I will be used here so let me move this over to here okay and here we're gonna create two thousand samples in the data set and it's going to create two classes and the features will be ten can import numpy SNP and then we're going to make noisy features in order to make the data set look more real in otherwise it will make perfect prediction and it would look just too good to be true so let's make it a little bit more difficult for the machine learning model to perform okay and so let's now perform the data splitting and now we're going to perform the actual trains have split using import argument of XY data matrices and the test size we're going to set it to be twenty percent okay and now we're going to build two classification models which we will compare the first one will be the random forest and the second one will be the Gaussian naive Bayes okay and then we're going to assign the random forest to the RF variable and that are F dot fit to create the model and the import argument will be the X screen and the y train which will be the 80% training set and finally it will be evaluated on the test set okay and now we're going to create the naive Bayes model so here we're going to create the prediction probability data matrices and ask the baseline we're going to have a variable called our probs and so this one will contain zero or the worst case scenario and here will contain the probability of the predicted values by the random forest model and by the naive Bayes model and here we are using the predict proper function to get the probability from the prediction and so in the following cell code we're going to keep the positive outcome and here we're going to import the library as fill n dot metrics and we're gonna use the function RLC curve roc AUC score okay now we're going to compute the a URL see a URL see is the area under the ROC curve so under the curve what is the area and then we're gonna print the scores and then here you see so by random chance prediction it is 0.5 so random chance prediction will mean that all predictions are wrong in here okay so when we have a sign a probability of the prediction to be all 0 it means that all predictions are wrong and for the worst-case scenario the a URL C will equal to 0.5 okay and then the performance of the random forest will be about 0.894 and now you based performed better by having a URL C of 99.3 okay and now we're gonna compute the fpr and TPR which will be used to create the RLC curve all right so now I think we're ready to finally plot the curve okay and here we go we have the RLC curve so in a nutshell the ROC curve is commonly used in the machine learning community to compare the performance of different learning models so here we can see that the naive Bayes provided the best performance as it occupies the curve at the far left and top and so the area under the curve will essentially be one almost one 0.99 3 we're asked for the random forest the area under the curve is about 0.894 right so it's the area under this curve okay and so for the labels here in the legend we put in the a URL see value in the parentheses so you can see that the area under the curve values are provided by these three variables and be a UC RF a USC and are a USC okay so if you run it individually you will see the values right 7.5 are FA you see 0.89 and NBA C 0.99 okay and then we're gonna put the modulo operator in here and then we're gonna print something else we're gonna say and B : and then it's going to be encapsulated in the print function and we're going to say % point 3f shift enter okay there you go actually okay there you go so you see that it round the number off to only three digit decimal 0.99 three so if you make it 2 it will be only two digits all right you can make it for as small 993 two okay so that's the label here which will be present in the legend and so the RLC plot title is in the PLT title function okay and you could also modify the X label Y label and smell in here alright so congratulation you have now created the RLC plot so you can improvise and try this on a different data set and then upload it to github grow your data science portfolio because the best way to learn data science is by doing data science not only by learning not by watching the science but by doing data science because when you do data science you will encounter errors you will encounter problems and the journey that you use to solve the problems will allow you to mature and learn and grow as a data scientist okay so embark on this journey to become a data scientist by doing the science so enjoy thank you for watching please like subscribe and share and I'll see you in the next one but in the meantime please check out these videos