Predicting the number of bike rentals using 4 different algorithms

Hi everyone!

I would like to share with you my guided project regarding bike rentals. In this project I applied 4 algorithms (3 requested by the task and SVM regression) and performed extensive hyperparameter tuning for a decision tree algorithm. I would like to hear any feedback about algorithm implementation, my conclusions and project in general.

Las scree of the guided project’s page: Learn data science with Python and R projects

Predicting bike rentals.ipynb (180.7 KB)

Click here to view the jupyter notebook file in a new tab

4 Likes

nice one,

  • I’d try Gboost - improves the score a touch
  • fiddle more with hr labels , the lesson tells us to divide it in 3 or 6 (don’ remember) hr intervals, look at the plot below: I’d merge only the hrs labels that are very similar to each other(eg 17 and 18)
  • to reduce code and work faster I wrote 1 function to tweak most of my plots and I use it all the time:
def spines(ax,yl='Rental counts',xl='',title=''):
    x1 = ax.spines['right'].set_visible(False)
    x2 = ax.spines['top'].set_visible(False)
    x3 = ax.spines['left'].set_linewidth(2)
    x4 = ax.spines['bottom'].set_linewidth(2)
    x5 = ax.set_ylabel(yl)
    x6 = ax.set_xlabel(xl)
    x7 = ax.set_title(title)
    return x1, x2, x3, x4, x5, x6

so now with 1 line of code I can deal with x, y labels, title and spines on my plots

3 Likes

Hi @adam.kubalica. Thank you for your feedback. Quite an interesting approach to combine only similar hr labels. I will try it out!