Question

我正在尝试创建自己的分类器，以下代码从理论上讲应该达到〜90％+的准确性，但是，它只能达到〜30％。我意识到我可以只使用kneighborsclassifier，但是，我正在尝试编写自己的。请帮助我确定这里出了什么问题。

以前是从库中导入分类器的。 ...这次，我们将编写自己的内容。

from scipy.spatial import distance

# a=trainingPoint & b=testingPoint
def euc(a,b):
    return distance.euclidean(a,b)

# Implement a class for our classifier:
class ScrappyKNN():
    def fit(self, X_train, y_train):
        # Store the training data in this class
        self.X_train = X_train
        self.y_train = y_train

    def predict(self, X_test):
        predicitons = []
        for row in X_test:
            # Method to find the closest point
            label = self.closest(row)
            predictions.append(label)

        return predictions

    # Loop over all training points and keep track of nearest ones, so far
    def closest(self, row):
        #calc dist from train point to nearest test point
        best_dist = euc(row, self.X_train[0])
        # Keep track of index of closest point
        best_index = 0
        # Iterate over all other training points and update when you find a closer one
        for i in range(1, len(self.X_train)):
            dist = euc(row, self.X_train[i])
            if dist < best.dist:
                best_dist = dist
                # use index to return the closest point
                best_index = i

            return self.y_train[best_index]

# Import a dataset
from sklearn import datasets
iris = datasets.load_iris()

# Feature--> f(x)=y <--Label
X = iris.data
y = iris.target

# Import the data, then partition into 1. _train and 2. _test
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = .5)

# Comment out our classifier and write our own...
# from sklearn.neighbors import KNeighborsClassifier

# Change our pipeline to use our new classifier
# my_classifier = KNeighborsClassifier()
my_classifier = ScrappyKNN()

# Use training data on classifier
my_classifier.fit(X_train, y_train)

# Use predict method to classify test data
# predictions = my_classifier.predict(X_test)
# print(predictions)

# To find accuracy, compare predicted labels to true values
from sklearn.metrics import accuracy_score
# In this case, accuracy should be ~%90
print(accuracy_score(y_test, predictions))

编写自己的kneighborsclassifier。准确性应该更高，这是什么问题？

0 个答案: