# Proportion of positive/negative label in Supervised Learning

I'm working on a Supervised Machine Learning problem and I have a question about the proportion of positive/negative label.

I would like to categorize some batch as OK or NOK. But actually my batchs are most often OK. To give you an idea, at the moment I have approximately 400 batchs OK and 10 NOK.

My questions are : What will be the impact on the results of this proportion? Is it still possible to have conclusive results? If so, with what type of algorithm?