Fit, Then Predict

A decision tree guesses an answer by asking yes/no questions about a row, one after another: fewer than 2 orders? More than 100 days since the last visit? Each answer leads to the next question, until the tree reaches a guess. You don't write those questions. The tree finds them itself, in rows whose answers you know.

DecisionTreeClassifier(max_depth=…, random_state=…), from sklearn.tree, gives back a new, empty model: your settings, no questions yet. max_depth is how many questions it may ask in a row, and random_state makes it build the same tree on every run. model.fit(X, y) is the learning: it finds the questions that best link the rows in X to their answers in y. model.predict(rows) gives back an array of guesses, one 0 or 1 for each row you pass it. For example, a tree that learns from 4 students how many hours of study it takes to pass:

hours = pd.DataFrame({"hours": [1, 2, 8, 9]})
passed = pd.Series([0, 0, 1, 1])
tree = DecisionTreeClassifier(max_depth=1, random_state=0)
tree.fit(hours, passed)
print(tree.predict(pd.DataFrame({"hours": [3, 7]})))   # [0 1]

From the 4 students the tree learned one question, "5 hours or less?", so it guesses 0 (fails) for 3 hours and 1 (passes) for 7. X_test[:5] gives back the first 5 test rows, and rows["predicted"] = … adds a column to a table, like a new key in a dictionary. Your script must print the 5 test rows with the tree's guess and the real answer side by side. The last two lines come from the lesson's check:

     visits  orders  spent  days_since_last  returned_item  predicted  real
208      18       6    357                8              0          1     1
188       5       2    310               20              0          1     1
12        8       1     22              140              1          1     0
221       2       3    104              140              0          0     1
239       9       1      0               87              0          1     0
Rows the tree learned from: 225
Tree depth: 3

Your task: the starting code loads the customers, splits them, and prints the first 5 test rows. Make a DecisionTreeClassifier with max_depth=3 and random_state=0 in model, fit it on X_train and y_train, keep its guesses for X_test[:5] in predictions, and add the columns "predicted" and "real" to rows.

Assignment
Follow the Coding Tutorial to fit a decision tree on the training rows and set its guesses for 5 test rows next to the real answers.

Hint
fit gets rows and their answers; predict gets rows only and gives back guesses. Fit first: an empty tree has no questions to guess with.