在python中实现一个简单的高斯朴素贝叶斯算法

Question

所以我是一个真正的业余爱好者，试图在python中实现一种你可能称之为朴素贝叶斯算法的“简化”版本，似乎有很多麻烦，原因可能是我不太确定我不完全理解算法的工作方式……不过，我还是非常感谢大家的帮助/建议。这是我的代码：

class GaussianNB(object):

    def __init__(self):
        '''
        Constructor
        '''
        # This variable will hold the gaussian distribution over your data
        # In fact, you need a distribution per class for each feature variable.
        # This can be done as a list of lists.

        self.classmodels_count = {}
        self.classmodels = {}
        self.classmodelsMeanAndVariance = {}
        self.featureTokenCount= 0;
        self.featureTypeCount = 0;

    def train(self, trainingdata):
        for i in trainingdata:
            current_class = i[0]
            features = i[1]

            if self.classmodels.has_key(current_class):
                current_class_model = self.classmodels[current_class]
                self.classmodels_count[current_class] =  self.classmodels_count[current_class] + 1
            else:
                current_class_model = {}
                self.classmodels_count[current_class]  = 1

            for f in  features:
                feature = f[0]
                value = f[1]
                if current_class_model.has_key(feature):
                    list_of_values = current_class_model[feature]
                    list_of_values.append(value)
                    current_class_model[feature] = list_of_values
                else:
                    list_of_values = []
                    list_of_values.append(value)
                    current_class_model[feature] = list_of_values

            self.classmodels[current_class] = current_class_model
        for a_class in self.classmodels.keys():
            a_class_model = self.classmodels[a_class]
            a_class_model_mean_and_variance = {}
            for feature in a_class_model.keys():
                a_class_model_mean_and_variance[feature] = findMeanSD(np.array(a_class_model[feature]))
            self.classmodelsMeanAndVariance[a_class] = a_class_model_mean_and_variance

    def classify(self, testing_vecs):
        outputs = []
        for vec in testing_vecs:
            features = vec[1]
            class_model_output_prob = {}
            for a_class in self.classmodelsMeanAndVariance.keys():
                a_class_output_prob = 0.0
                a_class_model_mean_and_variance = self.classmodelsMeanAndVariance[a_class]
                for feature_value in features:
                    feature = feature_value[0]
                    value = feature_value[1]
                    #simply ignore a feature if its not seen in training
                    if(a_class_model_mean_and_variance.has_key(feature)):
                        feature_mean = a_class_model_mean_and_variance[feature][0]
                        feature_std = a_class_model_mean_and_variance[feature][1]
                        a_class_output_prob = a_class_output_prob +  math.log10(norm(value,feature_mean,feature_std))

                #ignoring P(class) prior.. assuming equal priors
                class_model_output_prob[a_class_output_prob] = a_class
            probs = class_model_output_prob.keys()
            print probs
            probs.sort()
            max_prob = probs[len(probs)-1]
            max_class =class_model_output_prob[max_prob]
            outputs.append(max_class)
        return outputs 

在一些数据上运行时，我得到的错误是

回溯(最近一次调用)：文件"C:\Users\Toshiba\workspace\Assignment6\src\gnb_test.py"，第34行，在高斯= Model.train(testData)文件特征行91中，正在训练f in "C:\Users\Toshiba\workspace\Assignment6\src\gnb.py"，：TypeError：'numpy.float64‘对象不可迭代

我一点也不明白这是什么意思

Answer 1

回溯表明问题在于您试图遍历features，但features是一个浮点数，而不是列表或元组--基本上，它不能分解为单独的元素。我认为它是一个浮点，因为线条

for i in trainingdata:
   current_class = i[0]
   features = i[1]

建议将features重写为一系列连续的数字，而您似乎想要的是将这些数字保存为可迭代类型。试一试

features = []
for i in trainingdata:
   current_class = i[0]
   features.append(i[1])