decisiontree | ID3-based implementation of the ML Decision Tree algorithm | Machine Learning library

One way to convert an image dataset into X and Y NumPy arrays are as follows:

NOTE: This code is borrowed from here. This code is written by "PARASTOOP" on Github.

The error message is not very helpful, but I think the correct way to load the model back is to call the load method of the model, not of the estimator. The model is fitted to the data already, which is different from the estimator, which only contains the settings/parameters, but is not fitted.

So you can try this:

The plot is correct.

The two values in value are not the number of samples to go to the children nodes; instead, they are the negative and positive class counts in the node. For example, 748=101+647; there are 748 samples in that node, 647 of which are positive class. The child nodes have 685 and 63 samples, and 685+63=647. The left child has 47 of the negative samples, and the right node 54, and 47+54=101, the total number of negative samples.

sklearn.model_selection.cross_val_score gives you the score evaluated by cross validation, which means that it uses K-fold cross validation to fit and predict using the input data. The result is hence an array of k scores, resulting from each of the folds. You have an array of 5 values because cv defaults to that value, but you can modify it to others.

Here's an example using the iris dataset:

You have stored your models in a list of tuples (note that in your example the closing bracket is actually missing):

Your implicit assumption that the best hyperparameters found during CV should definitely produce the best results on an unseen test set is wrong. There is absolutely no guarantee whatsoever that something like that will happen.

The logic behind selecting hyperparameters this way is that it is the best we can do given the (limited) information we have at hand at the time of model fitting, i.e. it is the most rational choice. But the general context of the problem here is that of decision-making under uncertainty (the decision being indeed the choice of hyperparameters), and in such a context, there are no performance guarantees of any kind on unseen data.

Keep in mind that, by definition (and according to the underlying statistical theory), the CV results are not only biased on the specific dataset used, but even on the specific partitioning to training & validation folds; in other words, there is always the possibility that, using a different CV partitioning of the same data, you will end up with different "best values" for the hyperparameters involved - perhaps even more so when using an unstable classifier, such as a decision tree.

All this does not of course mean either that such a use of CV is useless or that we should spend the rest of our lives trying different CV partitions of our data, in order to be sure that we have the "best" hyperparameters; it simply means that CV is indeed a useful and rational heuristic approach here, but expecting any kind of mathematical assurance that its results will be optimal on unseen data is unfounded.

1. You have missed something elsewhere cause the object is indeed fitted. To check that use check_is_fitted().

2.You need to pass the best estimator to the export_graphviz()and not the Gridsearch i.e. export_graphviz(dt_clf.best_estimator_)

Example:

Per the docs, splitter must be either "best" or "random".

The cause is mentioned in the error stack trace.

ModuleNotFoundError: No module named 'numpy'

You just need to install numpy