Question

Recall and Precision are often discussed together as their focus is on complementary information. If precision is important, the we don’t not want to see any non-relevant documents. That is, whatever is retrieved, should be relevant. If recall is important, we want to see all the relevant documents, even if it requires sifting through some non-relevant ones. Provide and Justify two information-seeking tasks where precision may be considerably more important than recall. Similarly, Provide and Justify two information-seeking tasks where recall may be more important than precision. [Don’t forget to justify your choices: Justification will be graded, not the particular choices

Answer 1

Precision and recall are the two terms used in machine learning to measure the accuracy of our machine learning model. Generally, in many cases we want to to protect among two two cases accurately. Let's say we want to predict whether a given fruit is Apple or bananas. If the model make very less mistake in predicting bananas as bananas then it is high precision. If the model makes less mistake in predicting apples apples then it is high recall. In this case we want a model to have a high Precision and high record so that both apples and bananas a predicted accurately through our model. In some cases which will discuss now onwards are model is required to have either a high precision or a high recall.

In the case of accurately predicting spam in email, the model needs to have the high precision which is we need to to predict all those emails which are actually a spam. And the precision formula leads to the number of positives we have predicted as opposed to number of positive value actually present.

Precision = True positive/ True positive+False positive

In high precision, the model needs to minimise the false positive which is in this case is flasely predicting an email as spam when it is actually not a spam and then precision will increase. So in this case precision should be high.

In the second case where you want to predict whether a person has cancer or not, they will lead to give more importance to recall where we need to maximize the chance of model creditor accurately the people who actually have cancer. It's very important to not leave out any patient who actually happened sir but has been predicted as not having cancer.

Recall = True positive/ True positive+False negative

Here recall should be higher which measure the accuracy of predicting positive among all the actual cases having cancer. False negative here is that the model fails to recognise the presence of cancer in the patient who actually has cancer and it needs to be minimised to increase recall.