What did you realize when you tried to submit your predictions? What changes were needed to the output of the predictor to submit your results?
I described the final predictions series to see if there are any negative values, because Kaggle will reject the submission if we don't set everything to be > 0.
A LightGBMLarge_BAG_L1 model with the kaggle score of 0.45586
I found out that there are some features that don't have a normal or pert distributions and they are in category format. So I changed their type and also I seperated the 'datetime' column to two distinct columns and then ignored the 'datetime' column itself to prevent dependency in features.
How much better did your model preform after adding additional features and why do you think that is?
It performed in a significant better way, and I think it's because before this change, models were predicting based on a mixed feature of two independent features('datetime' column) but after separating it to two column and also categorizing some other features our models predicted in a much better way.
It became very better, because I neglected some time-consuming models like KNN and NN and also fine tuned the hyperparameters using some argument in fit function.
I should spend more time in discovering the depencency between features and also the threshold weight of the features based on their importance. For example I think that 'hour' or 'time' feature is very important because in most of the stations we don't have any demand on bake sharing from 10 p.m. to 5 a.m. and our models should predict a count of near zero in these times.
| model | hpo1 | hpo2 | hpo3 | score |
|---|---|---|---|---|
| initial | problem_type='regression' | learner_kwargs={'ignored_columns':['casual' , 'registered']} | time_limit=600 | 1.39407 |
| add_features | Added two distinct features 'date' and 'time' to train data | Changing the format of 'season' and 'weather' to category | learner_kwargs={'ignored_columns':['casual', 'datetime' , 'registered']} | 0.47959 |
| hpo | hyperparameter_tune_kwargs={'num_trials':5, 'scheduler' : 'local', 'searcher':'auto} | Changing the format of 'workingday' and 'holiday' to category | excluded_model_types = ['KNN', 'NN'] | 0.45586 |
Create a line plot showing the top model score for the three (or more) training runs during the project.
Create a line plot showing the top kaggle score for the three (or more) prediction submissions during the project.
TODO: I used a TabularPredictor to predict the 'count' column in the test file. I chose the problem type to be a Regression problem (because that's it!) and I used ‘root_mean_squared_error’ for regression. I ignored the 'casual' and 'registered' columns because they are not existed in the test file. During reading the csv files, I remembered to parse the datetime, so that I could create two new features 'date' and 'time. I changed the types of features l'season' and 'weather' to the Category format, because AutoGluon originally sees them as ints, but in reality they are int representations of a category. Setting the dtype to category will classify these as categories in AutoGluon. I gave a time limit of 10 minutes to the model fitting and then I predicted the test file's 'count' column using my predictor. At the end I described the final predictions series to see if there are any negative values, because Kaggle will reject the submission if we don't set everything to be > 0.