The goal of this assignment is to create a k$-$nearest neighbor model to classify births as low$-$risk. This is a two$-$step process requiring you to first run a model, then answer questions based on the model.Step $1$Run the model:Before you run the model, get some insights into the data. Of the variables you have, what factors impact low birth weight? Show evidence using charts. Remember, these are hunchesso you are not looking for a specific answer, just your take on how you read the data.Run a k$-$nearest neighbor model and search for a k between $1$ and $10.$ Partition your dataset into $60\%$ training and $40\%$ validation $($these are defaults$)$ to run your model. You will be using a cutoff probability of $0\mathrm{.}5($that is$,$ if half or more of the nearest neighbors have a low$-$birth$-$weight baby, the mother in question will be also classified as having a low$-$birth$-$weight baby$).$Step $2$Answer the following questions after running the model:What is the best k that minimizes the error of classification?For this k$,$ what are the Precision, Specificity, and Sensitivity of the model? $($Remember to use the confusion matrix from the validation dataset$)$What is the Precision, Specificity, and Sensitivity if the cutoff probability is decreased to $0\mathrm{.}3?$ If it is increased to $0\mathrm{.}7?($Hint: Simply rerun the model with the appropriate cutoff to get the confusion matrix. Analytic Solver will put new tabs on the spreadsheet with the updated cutoff$).$Mark the three cutoffs on the ROC curve. $($Hint: See Professor Ganeshan's slides just cut and paste the ROC curve from the output and mark it$)$What is the AUC of your model? What does it tell you about the model?Given a choice, would you use a higher or lower cutoff $($when compared to $0\mathrm{.}5)$ for the classification? Why? Why not?