03 ML KNN 實現的婚戀網站分類匹配

weixin_34320159發表於2016-05-10
KNN網站

marry_data 資料

from numpy import *
import operator
from os import listdir

def knn_class(inx, dataset, labels, k):
    dataset_size = dataset.shape[0]            # shape return size
    diff_mat     = tile(inx, (dataset_size, 1)) - dataset # tile() 計算距離
    sq_diff_mat  = diff_mat**2                 # python ** == ^ 這裡平方算距離
    sq_distances = sq_diff_mat.sum(axis = 1)   # axis = 0 -> 列 asix = 1 -> 行 按列累和 (x^2 + y^2)
    distances    = sq_distances**0.5           # (x^2 + y^2)開方算距離 
    sorted_dist_indicies = distances.argsort() # 距離計算 argsort() 函式返回從小到大的索引值

    # 選取 K 個距離最小的點 進行分類  並且統計各個分類的數量
    class_count = {}
    for i in range(k): # [0, k-1]
        vote_label = labels[sorted_dist_indicies[i]]
        class_count[vote_label] = class_count.get(vote_label, 0) + 1 

    sorted_class_count = sorted(class_count.iteritems(), key=operator.itemgetter(1), reverse=True)
    return sorted_class_count[0][0]

def file_to_matrix(filename):
    love_dict = {'largeDoses' : 3, 'smallDoses' : 2, 'didntLike': 1}
    file = open(filename)

    lines = file.readlines()
    lines_num = len(lines)
    data_matrix = zeros((lines_num, 3)) # -> lines_num * 3 matrix
    
    class_label = []
    idx = 0
    for line in lines:
        line = line.strip()           # 刪除空白字元
        msgs = line.split('\t')
        
        data_matrix[idx, :] = msgs[0:3]   # 放入 對應的行中 40920  8.326976    0.953952    largeDoses
        if (msgs[-1].isdigit()):
            class_label.append(int(msgs[-1]))
        else:
            class_label.append(love_dict.get(msgs[-1])) # 獲取得到該資料的 lable 對應的編號 3 2 1
        idx += 1
    return data_matrix, class_label

# 把資料歸一化到 [0, 1]
def auto_norm(data_set):
    min_vals = data_set.min(0)
    max_vals = data_set.max(0)
    ranges   = max_vals - min_vals

    norm_data = zeros(shape(data_set))
    row_size = data_set.shape[0]
    print('row_size', row_size)

    # 這裡歸一化的演算法思路： [x, y] z 在 x,y 之間 
    # 結果 = (z - x) / (y - x) 比如： [1, 9] z = 4 -> = (4 - 1) / (9 - 1)
    norm_data = data_set - tile(min_vals, (row_size, 1))
    norm_data = norm_data / tile (ranges, (row_size, 1))
    return norm_data, ranges, min_vals    
    

# 檔案中的資料格式
# 40920 8.326976    0.953952    largeDoses
# 14488 7.153469    1.673904    smallDoses
# 26052 1.441871    0.805124    didntLike
# 75136 13.147394   0.428964    didntLike
# 38344 1.669788    0.134296    didntLike
# 72993 10.141740   1.032955    didntLike

def date_class_test():
    ratio = 0.1 # 這裡用 90% 的資料來訓練 10% 資料留作驗證
    data_matrix, class_label = file_to_matrix('./marry_data')
    norm_matrix, ranges, min_vals = auto_norm(data_matrix) # 資料歸一化 使得資料都在 [0,1] 之間 影響因子相同

    norm_size = norm_matrix.shape[0]
    test_num  = int(norm_size * ratio)
    error_count = 0.0

    for i in range(test_num):
        result = knn_class(norm_matrix[i, :], norm_matrix[test_num:norm_size, :],
             class_label[test_num:norm_size], 3)

        if (result != class_label[i]):
            error_count += 1.0
    return (error_count / float(test_num)) * 100

print('error_count: %d') % (date_class_test()) + '%' # 5.0%

相關文章