how to increase accuracy of classification

[ionic breach]

hello i need help with a python based classifier where we are being assessed on our accuracy but i keep getting really low scores. i'm willing to show the code but i might have to dm it as it's a lot

[sullen creek]

what classification algorithm are you using?

[ionic breach]

knn with pca for dimension reductionality

[sullen creek]

ok I'd say mess around w a bunch of different classification algorithms

logistic regression (maybe multiple if u have multiple classes)

gradient boosting

random forest

unless u have to use this specifically

[ionic breach]

we can't use scikitlearn so i would have to build all that from scratch 🫠

i'm using knn because it's what we were taught in our classes so i'm not really sure on any other stuff

[sullen creek]

oh dang

ok can u use like TensorFlow or pytorch? or just like numpy

[ionic breach]

i can use numpy, scipy or anything else in python standard module library

[sullen creek]

ok, i might suggest looking into some of the algorithms I mentioned and see if any of them you think you can implement easily

also try doing some neural network stuff maybe

there are a number of tutorials on how to do that with just numpy

[ionic breach]

hm i'll see thank you for the suggestions

[ionic breach]

ok i've been implementing the classifier and that's all good but when i use pca the results are worse than just using the classifier on it's own can someone please help with this it's urgent

[ionic breach]

<@&1170904818810880060> sorry for the ping i just want to make sure someone sees this 😭

[sullen creek]

Maybe you could try doing knn on the output of the pca?

if that doesn't work id suggest just to try some other algorithm

[ionic breach]

i've discovered the problem which is that i'm using the principal components of the test set but the classifier is trained on the training set, using the incorrect parameters, i'm trying to fix it by saving pca model but i'm new to python so i really don't get how