Challenge - Diabetes Classification

#1

i am not able to understand problem statement

#2

Hi @rishigarg,
In this challenge, first download the training and testing dataset…
Make a knn algorithm from scratch, then use X_test values from test dataset iterate over them and one by one predict the values of these X_test… Store the predictions in a list, and convert the output to a csv file according to the submission file.

Thanks :slight_smile:

#3

hi !! i tried it but was able to score 78% only
i tried almost all values of k between 3 and 21
how to improve accuracy
please help

#4

Hello ,
78 is a good enough score when talking about KNNs. Expecting a higher score would be unrealistic. KNN is mostly used as a threshold for ML models, as one can decipher from the way it works, it’s a very primitive and a simple algorithm mostly based on intuition. Higher level models like SVMs and Random Forests can be used to obtain a much better accuracy.

I hope this resolves your doubt.