Week 5 — Make World Green Again!
This week we worked on classification for our project. To do this,we divided the difference between minimum and maximum no2 in the data to 12,6 and 4 equal parts.This way, we were able to try many new methods. In the reference we take for our project, we have seen the data was split in 4 different labels.
However, we tried to find the most suitable method by dividing the data to more parts to have more sensitive measurement.
Neural network has been used in the related work we studied for this process. Therefore, we picked it as our first choice.After we tested various label numbers and sizes, we found the ideal one. You can see the results of different experiments.
Also, we saw which feature is more guiding regarding guessing the label of peak values of no2 when we were working on decision tree and plotting the tree.Besides you can see the test results which is done to find the ideal height.
You can examine the data overfit with unnecessary increase on height and also see that the decrease in prediction rates when there is a data from outside that specific dataset.