Python -在python中计算不同数据类型的距离

Question

我有一个包含11个属性的数据。我想计算每个属性的距离。例如，它的属性(x1, x2, ..., x11)和for x1 & x2具有Nomal型，x3, x4, ... x10具有ordinal类型，那么x11具有binary类型。如何使用python读取属性？如何在python中区分这些属性，以及如何在python中区分这些属性，以便我可以计算距离？有人能告诉我我该怎么做吗？谢谢

样本数据: x1 (林业，人工林，其他，林业) x2 (人工林，人工林，灌木，森林) x3 (高，高，中，低) x4 (低，中，高，高) x5 (高，低，中，高) x6 (中，低，高，中) x7 (3，1，0，4) x8 (低，低，高，中) x9 (297,298,299,297) x10 (1，2，0，4) x11 (t，f)

Answer 1

你可以这样做：

def distance(x,y):
    p = len(x)
    m = sum(map(lambda (a,b): 1 if a == b else 0, zip(x,y)))
    return float(p-m)/p

示例：

x1 = ("forestry", "plantation", "high", "low", "high", "medium", 3, "low", 297, 1, True)
x2 = ("plantation", "plantation", "high", "medium", "low", "low", 1, "low", 298, 2, True)

print distance(x1,x2) # result: 0.636363636364 = (11-4)/7

Answer 2

我将其重写如下：

首先，我创建一个Nominal类型工厂：

class BaseNominalType:
    name_values = {}   # <= subclass must override this

    def __init__(self, name):
        self.name = name
        self.value = self.name_values[name]

    def __str__(self):
        return self.name

    def __sub__(self, other):
        assert type(self) == type(other), "Incompatible types, subtraction is undefined"
        return self.value - other.value

# class factory function
def make_nominal_type(name_values):
    try:
        nv = dict(name_values)
    except ValueError:
        nv = {item:i for i,item in enumerate(name_values)}

    # make custom type
    class MyNominalType(BaseNominalType):
        name_values = nv
    return MyNominalType

现在我可以定义你的名义类型了，

Forest = make_nominal_type(["shrubs", "plantation", "forestry", "other"])
Level  = make_nominal_type(["low", "medium", "high"])
Bool   = make_nominal_type({"f":False, "t":True})

然后我创建了一个MixedVector类型工厂：

# base class
class BaseMixedVectorType:
    types = []          # <= subclass must
    distance_fn = None  # <=   override these

    def __init__(self, values):
        self.values = [type_(value) for type_,value in zip(self.types, values)]

    def dist(self, other):
        return self.distance_fn([abs(s - o) for s,o in zip(self.values, other.values)])

# class factory function
def make_mixed_vector_type(types, distance_fn):
    tl = list(types)
    df = distance_fn

    class MyVectorType(BaseMixedVectorType):
        types = tl
        distance_fn = df
    return MyVectorType

然后创建您的数据类型，

# your mixed-vector type
DataItem = make_mixed_vector_type(
    [Forest, Forest, Level, Level, Level, Level, int, Level, int, int, Bool],
    ??? # have to define an appropriate distance function!
)

..。但是等等，我们还没有定义一个距离函数！我编写了这个类，允许您插入任何您喜欢的距离函数，格式如下：

def manhattan_dist(_, vector):
    return sum(vector)

def euclidean_dist(_, vector):
    return sum(v*v for v in vector) ** 0.5

# the distance function per your description:
def fractional_match_distance(_, vector):
    return float(sum(not v for v in vector)) / len(vector)

所以我们完成了创建

# your mixed-vector type
DataItem = make_mixed_vector_type(
    [Forest, Forest, Level, Level, Level, Level, int, Level, int, int, Bool],
    fractional_match_distance
)

并将其测试为

def main():
    raw_data = [
        ('forestry', 'plantation', 'high', 'low', 'high', 'medium', 3, 'low', 297, 1, 't'),
        ('plantation', 'plantation', 'high', 'medium', 'low', 'low', 1, 'low', 298, 2, 't'),
        ('other', 'shrubs', 'medium', 'high', 'medium', 'high', 0, 'high', 299, 0, 't'),
        ('forestry', 'forestry', 'low', 'high', 'high', 'medium', 4, 'medium', 297, 4, 'f')
    ]

    a, b, c, d = [DataItem(d) for d in raw_data]

    print("a to b, dist = {}".format(a.dist(b)))
    print("b to c, dist = {}".format(b.dist(c)))
    print("c to d, dist = {}".format(c.dist(d)))

if __name__=="__main__":
    main()

这给了我们

a to b, dist = 0.363636363636
b to c, dist = 0.0909090909091
c to d, dist = 0.0909090909091