I am new to data science field and I have started to participate in various data science competition like the competition which is conducted by analytics Vidhya.I want to know whenever we create a model on training data set and then test on test data set usually training error is less than test error .Is there any condition when we have test error less than training error.
If your test error is less than the training error, this means that there is a sampling bias in your test.
This can be explained by a simple example. If you are a student studying for an exam, and you understood only 40% of your syllabus. Fortunately for you the examiner asks you question only on the things you learnt and you get a 100% result. This does not mean that you know the whole subject, just that the test was ‘biased’ for you.
Awesome reply jalFaizy. Crystal Clear explanation.