First, I downloaded the data file, titled car.data, from [https://code.datasciencedojo.com/datasciencedojo/datasets/tree/master/Car%20Evaluation].
It should be noted that, given the fact that the desired output is a label, not a quantity, multiclass classification employing approaches such as Random Forest Classifier would be more appropriate for this particular dataset. I wrote a python script titled multiclassification.py to process the data and apply a predictive model on the buying price, employing Random Forests. The car.data file came without headers, so I wrote function save_file_with_headers to add the headers described in the above url and convert the file to csv, having the car.data.csv file as output. I wrote function ordinal_data_classification to process the data and perform the prediction task. Given the fact that the attribute data were qualitative, but ordinal, I proceeded to encode each attribute column, from lowest to highest quality, starting from 0. Subsequently, I checked for attribute correlation, which could be useful in using stratified sampling based on the attribute most correlated to the buying price, to appropriately split the dataset to a training and testing set. I noticed that there is a slight negative correlation between buying price and car acceptability. Normally, I would choose to do stratified sampling based on correlated attributes, however, this correlation result seems counter-intuitive, so I just performed random sampling instead. I then trained the Random Forest model and tested on the test set (20% of the entire dataset). For Random Forest Classifier, RMSE_train_set < RMSE_test_set, we can conclude that there is some overfitting of the data. Finally, I test each predictive model on a triple row dataframe (since we do not know the capacity, we test for all three available categories). Results show that a low buying price is predicted for all capacities.