Data Analytics Made Easy by Andrea De Mauro

Author:Andrea De Mauro
Language: eng
Format: epub
Publisher: Packt
Published: 2021-08-26T06:56:08+00:00

We obtained an R2 of 0.92, which means that our current model accounts for around 92% of the full variability of rental prices in Rome. Considering the limited sample and the few features available, this looks quite good already. Also, the RMSE is €110, which means that 68% of the time (one standard deviation) we will make a prediction error that is, in absolute terms, below €110, and 95% of the time our error will be below €220 (two times the RMSE). The last performance metric, Mean Absolute Percentage Error (MAPE) tells us that, on average, our predicted rent will differ from the actual rent by around 10%: again, not bad at all.

Still, we strive for the best and question ourselves if we can do anything to improve the model. The simplest thing to do will be to consider whether we can improve the selection of features. Let's go back and have a look at the parameters obtained by the regression (Figure 5.9) and if we can remove some unneeded (or damaging) features. When we remove excess features from a model, we obtain at least two advantages: first, we make the model simpler and more explanatory to other human beings, as we have fewer parameters to explain. Secondly, we reduce the possibility for the model to overfit on the training set and, so, we increase its general robustness.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.