相关性算法和api

1.相关性

    先写几个相关性的基础算法和相应的api吧

    欧式距离

from scipy import spatial
import math
# 欧式距离
def d_euclidean(*args):  #args为所有向量组成数组,其实就是个二维数组
    ''' 第一个值是其它值的对比对象,其它值都是以第一个值为对比标准'''
    """ 返回也是个数组,为和第一个值对比出来的欧氏距离"""
    """ 欧式距离公式 d(x,y) = ((y1 - x1)^2 + (y2 - x2)^2 +(y3 - x3)^2 .....)^(0.5)"""
    base_vector = args[0]
    d_values = []
    for o_vector in args:
        if base_vector is o_vector:
            continue
        d_value = 0
        for index in range(len(base_vector)):
            d_value += pow((o_vector[index] - base_vector[index]),2)
        d_values.append(math.sqrt(d_value))
    if len(d_values) > 0:
        return d_values
    else:
        return -1

vector1 = [0,1,2,3]
vector2 = [1,2,3,4]
vector3 = [4,3,2,1]
print(d_euclidean(vector1,vector2,vector3))
#结果为 [2.0, 4.898979485566356]
print(spatial.distance.euclidean(vector1,vector2))
print(spatial.distance.euclidean(vector1,vector3))
# 2.0
# 4.898979485566356
#结果是一样的

曼哈顿距离

#曼哈顿距离

import numpy as np
from scipy.spatial.distance import pdist

def d_manhattan(*args):
    """这个就超级简单了,就是求相差的绝对值"""
    """然后把这些绝对值加起来就妥了"""
    base_vector = args[0]
    d_values = []
    for o_vector in args:
        if base_vector is o_vector:
            continue
        d_value = 0
        for index in range(len(base_vector)):
            d_value += abs((o_vector[index] - base_vector[index]))
        d_values.append(d_value)
    if len(d_values) > 0:
        return d_values
    else:
        return -1

vector1 = [0,1,2,3]
vector2 = [1,2,3,4]
vector3 = [4,3,2,1]
print(d_manhattan(vector1,vector2,vector3))
# [4,8]
print(d_manhattan(vector2,vector3))
#[8]
print(pdist(np.vstack([vector1,vector2,vector3]),'cityblock'))  #这个api会返回所有的相似关系,是ndarray形式
# 结果 [4 8 8]

余弦相似度

#余弦相似度
import numpy as np
from scipy.spatial.distance import pdist
def d_cos(*args):
    base_vector = args[0]
    a_vals = [np.sum(np.square(x)) for x in args]
    b_val = a_vals[0]
    d_values = []
    for i in range(len(args)):
        o_vector = args[i]
        if i == 0:
            continue
        d_value = 0
        for index in range(len(base_vector)):
            d_value += (o_vector[index] * base_vector[index])
        d_values.append(1 - d_value/np.sqrt(b_val*a_vals[i]))
    if len(d_values) > 0:
        return d_values
    else:
        return -1

vector1 = [0,1,2,3]
vector2 = [1,2,3,4]
vector3 = [4,3,2,1]
print(d_cos(vector1,vector2,vector3))
#[0.024099927051466796, 0.5120499635257334]
print(d_cos(vector2,vector3))
#[0.33333333333333337]
print(pdist(np.vstack([vector1,vector2,vector3]),'cosine'))
#[0.02409993 0.51204996 0.33333333]

皮尔森相关系数

import numpy as np
from scipy.stats import pearsonr
def get_pearsonr(b,o):
    b_s = np.sum(b)
    o_s = np.sum(o)
    b_ss = np.sum(np.square(b))
    o_ss = np.sum(np.square(o))
    t_bo = 0
    n = len(b)
    for i in range(len(b)):
        t_bo += b[i] * o[i]
    denominator = np.sqrt(b_ss - np.square(b_s)/n)*np.sqrt(o_ss - np.square(o_s)/n)
    if denominator:
        return (t_bo - (o_s * b_s)/n) / denominator
    else:
        return 0


def d_pearsonr(*args):
    pass

vector1 = np.random.normal(1,100,50)
vector2 = np.random.normal(2,10,50)
vector3 = [4,3,2,1]
print(get_pearsonr(vector1,vector2))
#随机数,得到的结果每次都不一样比如这次的结果是-0.10288209400426439
print(pearsonr(vector1,vector2))
#(-0.10288209400426439, 0.4770990724691394)    这次是一样的,但有时候会不一样,有空研究下怎么还会不一样

scipy的皮尔森函数返回两个值,第一个是皮尔森系数,另一个是p-value也是表示相关性的