Now in this video we will be discussing about confusion matrix and different measures of confusion matrix and what is our OC curve that is receiver operating characteristic confusion matrix is a tabular presentation of actual versus predicted values actual means observed that is observed and predicted values this helps us to find the accuracy of the model and avoids over a confusion matrix is the most crucial metric commonly used to evaluate the classification models It is used to measure the accuracy of a model this is how a confusion matrix looks like observe this these are the observed cases these are the predicted cases zero means it's a non event one means it's an event 00 means it's observed that the observation where it is observed as zero means observed as non event as predicted as non event there are 50 cases 01 means observed as non event and predicted as event there are 30 cases one zero is obvious Death event but credited as non event there are 40 cases one one is observed that event as illustrated as even there are 80 cases.

So, this is the total the total number of observations the total number of cases that is 200. Now, using this table I will be explaining you all the different measures of confusion matrix the first measure of confusion matrix is water known events known events as I told the probability y equals to zero next is correctly classified event correctly classified event is for probability level prediction is an event and observed outcome is also an event. So, according to the last table correctly classified event is this is this one one. So, there are a number of observations which is observed as one as respected as back next is correctly classified known evil that is for a probability level production is a non event and observed outcome is also a non event. So, according to our table that is this this case that is there are 50 cases next incorrectly classified event this is for a probability level prediction is an event but observed outcome is a non event.

So, that is this case that is 13 01 there are 30 cases which are incorrectly classified events then incorrectly classified known events for a priority level prediction is a non event, but observed outcome is an event. So, that is this case that is there are 40 cases where it is observed as one and predicted as zero that is one zero 40 cases Next comes the percentage correct percentage correct is percentage of correct predictions out of the total production that is correctly classified percentage total amount of correctly classified events and non events by total number of observations in 200. So, that is 00 plus one one by total into hundreds that is 50 plus 80 by 202 hundred which is around 150 by 202 hundred so, that is 65% Next is sensitivity sensitivity it measures the ability to predict an event correctly which is calculated as correctly predicted as events By total number of observed events in 200, so, that is 80 by 120 200.

Next is specificity it measures the ability to predict a known event correctly, which is calculated as correctly predicted as non events by total number of observed non events in 200 that is equal to 50 by 80. In 200, next is false positive false positive is incorrectly predicted as event by total production as event in 200. So, that is 30 by 110 and 200 and false negative is the ratio of incorrectly classified known events by total production as known evil in 200. So, that is 14 by 19 200. Now, this correctly classified event correctly classified non event incorrectly classified events incorrectly classified known events then false positives and false negatives then sensitivity and specificity these also have some other names when we call 00 that is called correctly classified known events. So, this these 50 cases that is correctly classified known events are also called True negative than 01.

That is incorrect classified even this is also called false positives these 30 cases these are also called false positives or another name of incorrect classified even is called false positive next 40 cases that is one zero that is incorrectly classified known events are also called false negative and one one that is observed as one and predicted as one that is correctly classified even these are also called true positives. So, this is true negative This is false positive this is false negative and this is true positive and the percentage correct that we are calculating that is also used to check the accuracy of the model my sensitivity is also known as true positive rate and one minus sensitivity is called false negative rate true negative rate is specificity and false positive rate is one minus specificity. Now let's move to the concept of our OC curve rasika stands for receiver operating characteristic of our receiver operator characteristic curve.

So, receiver operating characteristic curve summarizes the models performance by evaluating the trade offs between true costs rate that is sensitivity and false positive rate that is one minus specificity. So this is our OC curve is project between one minus specificity and sensitivity our OC curve summarizes the predictive power of the model the area under our OSI is referred to as the index of accuracy or concordance index is a perfect performance metric for our speaker higher is the area under our OSI better is a predictive power of the more so our OSI determines the accuracy of a classification model after user defined threshold value it determines the models accuracy using the area under curve that is AUC the area under the curve that is AUC also referred to as index of accuracy or convert an index represents the performance of the rasika higher is the idea better is the model our OC curve is plotted between true positive rate that is sensitivity and false positive rate it is one minus specificity.

So in this video, we will be learning till here so for now, let's end this video over here. Thank you Goodbye and see you all for the next video.