Performance Metrics Interview Questions- Data Science

Key Takeaways

hello on my name is krishna and welcome to my youtube channel so guys today in this particular video we'll be discussing some of the interview questions with respect to classification matrix which is pretty much important i have uploaded a lot of videos regarding classification matrix but i've covered everything almost like confusion matrix true positive rate false positive rate true negative rate false negative rate recall precision f1 score FB TAS called ROC curve AUC come everything i've actually covered it in two videos so why not actually covered it down because most of the interview questions they ask you some of the questions regarding classification matrix so it is pretty much important so I've noted down some of the questions which you know recruiter has actually asked me so you also try to find out these particular answers and if you have any queries you can actually watch my videos again in my complete machine learning playlist so the first question is one of my interviewer asked this particular question to tell me some examples where false positive is more important that false negative so this basically says that you have to try to find out some use case in which your false positive is actually more important than false negatives know you can take some examples from disease classification you can take some examples from stock market predictions different kind of examples you can actually prepare for this kind of question now when you are preparing for this you should actually know what exactly is false positive what exactly is false negative and when is false positive more important than the false negative in which kind of use cases in which kind of scenario okay if I give you an example of stock market crashing of the stock market and the stock market is not crashing okay in this particular case can you tell me like whether false positive will be important or false negative will be important okay just comment down in this particular video in the comment section I will have a look and again I will give you a reply if you actually right apart from that guys if you are preparing for false positive and false negative make sure that you also try to find out some more examples where false negative is more important than false positive okay so this examples will also help apart from that you can also you know prepared for such examples where false negative is equal to the false positive and also both the scenarios are actually equal you know in this kind of scenario Z and this will be pretty much interesting you will get confused initially but you'll love it when you learn different different kind of examples the next interview question which I have actually returned on and this was again asked in one of my interview questions okay if they have given you option whether do you want to go with model accuracy or with modelled performance which one will you choose okay suppose if you want to go with model accuracy all you want to go with model performance which one will you choose and you should be able to explain them in a proper way because many people will say that okay I'll go with model performance some people may say that I'll go with model accuracy but just understand you have to make them explain why exactly you are going with one of this particular option which is so which is the best way to actually find out through which metrics whether through model accuracy or model performance should I go in order to understand whether the model is performing well or not so you should be able to explain pretty much well so these are the two common interview questions that are asked to me the next interview question was also regarding ROC and AUC curve and trust me guys ROC in use occur if I consider that particular example you know you'll be able to find out the area under the curve and based on that you will also be able to select your threshold value and what I am saying about the threshold value by default in logistic regression if you are selecting Thresher the default threshold values that is selected is 0.5 and anything greater than 0.5 if the probability comes greater than 0.5 you are taking that as a positive value otherwise you are taking the output as 0 okay so in this particular scenario sometimes based on the different use cases you need to select different different threshold values so that you get a good accuracy good performance rate okay and when you are selecting this you basically try to use ROC mu C curve and when you are using those kind of curve you also have to make sure that on what parameter you are actually focusing do you want to reduce the false positive very very less or do you want to increase the true positive rate in a higher manner so all this kind of scenarios has to be taken so this was the some of the interview questions please do comment down on this all interview questions in just two to three lines it will be I'll have a look and I'll give you a reply if you are absolutely right and yes keep on learning asean in the next video please do subscribe the channel till then bye bye