交叉验证: 为了让评估的模型更加准确可信, 交叉验证通常是和网格搜索搭配使用的.

在给定的建模样本中，拿出大部分样本进行建模型，留小部分样本用刚建立的模型进行预报，并求这小部分样本的预报误差，记录它们的平方加和。
交叉验证详情请看 交叉验证_百度百科

超参数搜索 - 网格搜索

超参数搜索和网格搜索是同一个概念, 通常情况下, 有很多参数需要手动指定的 (例如 K - 近邻算法中的 K 值), 这种叫做超参数. 但是手动过程繁杂, 所以需要对模型预设几种超参数组合. 每组超参数对采用交叉验证来进行评估 . 最后选出最优的参数组合建立模型.

超参数搜索 - 网格搜索 API

• sklearn.model_selection.GridSearchCV

代码

import pandas as pd
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.neighbors import KNeighborsClassifier
from sklearn.preprocessing import StandardScaler


def knncls():
    """
    K-近邻预测用户签到位置
    :return: None
    """

    # 读取数据
    data = pd.read_csv("../data/train/train.csv")

    # print(data)

    # 处理数据
    # 1.缩小数据,查询数据筛选
    # data.query("x > 1.0 & x < 1.15")
    data = data.query("x > 1.0 &  x < 1.25 & y > 2.5 & y < 2.75")

    print("data.size", data.size)

    # 处理时间的数据
    time_value = pd.to_datetime(data['time'], unit='s')

    # print(time_value)

    # 把日期格式转换成字典格式

    time_value = pd.DatetimeIndex(time_value)

    # 构造一些特征
    data['day'] = time_value.day
    # data['hour'] = time_value.hour
    # data['weekdat'] = time_value.weekday

    # 把时间戳特征删除

    # axis=1 1等于列 0等于行(pands)
    # sklearn 1等于行 0等于列(sklearn)
    # 表示删除time特征列
    print(data)
    data = data.drop(['time'], axis=1)
    # data = data.drop(['row_id'],axis=1)

    # print(data)

    # 把签到数量小于N个目标位置删除
    # 根据plcae_id 分组
    place_count = data.groupby('place_id').count()
    print("分组成功")
    # print(place_count)
    # 删除数量小于3的分组 并重置列表
    tf = place_count[place_count.row_id > 200].reset_index()
    print("重置列表成功")

    # 取出data和tf具有相同数据的样本
    data = data[data['place_id'].isin(tf.place_id)]

    print("取出相同数据样本成功")

    # 取出数据中的特征值和目标值
    # x 特征值 y目标值
    y = data['place_id']
    x = data.drop(['place_id'], axis=1)

    print("取出特征值和目标值成功")

    # 进行数据的分割训练集合测试集(这一句话是老师写的)
    # 分割训练集和测试集(这句话才是我理解的)
    # train 训练  test 测试
    x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.25)

    print("分为训练集和测试集成功")

    # 特征工程 (标准化)
    std = StandardScaler()

    # 对测试集和训练集的特征值进行标准化
    x_train = std.fit_transform(x_train)
    x_test = std.transform(x_test)

    print("标准化成功")

    # 进行算法流程
    knn = KNeighborsClassifier()

    # 构造一些参数的值进行搜索
    param = {"n_neighbors": [1,3,5,7,9,11]}

    # cv=2 2折交叉验证
    gc = GridSearchCV(knn, param_grid=param, cv=10)

    gc.fit(x_train, y_train)

    # fit, predict, score  喂数据
    # knn.fit(x_train, y_train)

    # print("x_train",x_train)
    # print("y_train", y_train)

    # a = [[0], [1], [2], [3], [4], [5], [6], [7], [8]]
    # b = [0, 0, 0, 1, 1, 1, 2, 2, 2]

    # print("a",a)
    # print("b",b)

    # knn.fit(a, b)

    print("喂入数据成功")

    # 得出预测结果
    # y_predict = knn.predict(x_test)

    # print("预测的目标签到位置为:", y_predict)

    # 得出准确率
    print("预测的准确率为:", gc.score(x_test, y_test))

    print("在交叉验证中最好的验证结果:", gc.best_score_)

    print("最好的模型是:", gc.best_estimator_)

    print("超参数每次交叉验证的结果:", gc.cv_results_)

    return None

if __name__ == "__main__":
    knncls()

运行结果

data.size 106260
E:/workspace/PycharmProjects/first/ai/sklearn/train.py:35: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead
See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
  data['day'] = time_value.day
            row_id       x       y  accuracy    time    place_id  day
600            600  1.2214  2.7023        17   65380  6683426742    1
957            957  1.1832  2.6891        58  785470  6683426742   10
4345          4345  1.1935  2.6550        11  400082  6889790653    5
4735          4735  1.1452  2.6074        49  514983  6822359752    6
5580          5580  1.0089  2.7287        19  732410  1527921905    9
6090          6090  1.1140  2.6262        11  145507  4000153867    2
6234          6234  1.1449  2.5003        34  316377  3741484405    4
6350          6350  1.0844  2.7436        65   36816  5963693798    1
7468          7468  1.0058  2.5096        66  746766  9076695703    9
8478          8478  1.2015  2.5187        72  690722  3992589015    8
9357          9357  1.1916  2.7323       170  319999  5163401947    4
12125        12125  1.1388  2.5029        69  532507  7536975002    7
14937        14937  1.1426  2.7441        11  445598  6780386626    6
20660        20660  1.2387  2.5959        65  616095  3683087833    8
20930        20930  1.0519  2.5208        67  163908  6399991653    2
21731        21731  1.2171  2.7263        99  550339  8048985799    7
26584        26584  1.1235  2.6282        63  316089  5606572086    4
27937        27937  1.1287  2.6332       588  618714  5606572086    8
30798        30798  1.0422  2.6474        49   75510  1435128522    1
33184        33184  1.0128  2.5865        75  487899  1913341282    6
33877        33877  1.1437  2.6972       972  140281  6683426742    2
34340        34340  1.1513  2.5824       176  309820  2355236719    4
37405        37405  1.2122  2.7106        10  315301  2946102544    4
38968        38968  1.1496  2.6298       166  636960  9598377925    8
41861        41861  1.0886  2.6840        10   11616  3312463746    1
42135        42135  1.0498  2.6840         5   95801  3312463746    2
42729        42729  1.0694  2.5829        10   57817  1812226671    1
44283        44283  1.2384  2.7398        60  629289  8048985799    8
44549        44549  1.2077  2.5370        76  522601  3992589015    7
44694        44694  1.0380  2.5315       152  657007  5035268417    8
            ...     ...     ...       ...     ...         ...  ...
29070221  29070221  1.1678  2.5605        66  528907  2355236719    7
29070322  29070322  1.0493  2.7010        74   65604  3312463746    1
29070934  29070934  1.1899  2.5176        28  186248  2199223958    3
29071712  29071712  1.2260  2.7367         4  620357  2946102544    8
29072165  29072165  1.0175  2.6220        42  304927  5283227804    4
29073572  29073572  1.2467  2.7316        64  592178  8048985799    7
29074121  29074121  1.2071  2.6646       161  613821  5270522918    8
29077579  29077579  1.2479  2.6474        42  670110  2006503124    8
29077716  29077716  1.1898  2.7013         5  733871  6683426742    9
29079070  29079070  1.1882  2.5476        28  520404  1731306153    7
29079416  29079416  1.2335  2.5903        72  384495  6766324666    5
29079931  29079931  1.0213  2.6554       167  106545  5270522918    2
29083241  29083241  1.0600  2.6722        71   88510  9632980559    2
29083789  29083789  1.0674  2.6184        88  380389  1097200869    5
29084739  29084739  1.2319  2.6767        63  475457  2327054745    6
29085497  29085497  1.0550  2.5997       175  214293  1097200869    3
29086167  29086167  1.0515  2.6758        57  608677  6237569496    8
29087094  29087094  1.0088  2.5978        71  339901  1097200869    4
29089004  29089004  1.1860  2.6926       153   84384  2215268322    1
29090443  29090443  1.0568  2.6959        58  205222  2460093296    3
29093677  29093677  1.0016  2.5252        16  554616  9013153173    7
29094547  29094547  1.1101  2.6530        24  733474  5270522918    9
29096155  29096155  1.0122  2.6450        65  288464  8178619377    4
29099420  29099420  1.1675  2.5556         9  316067  2355236719    4
29099686  29099686  1.0405  2.6723        13  609851  3312463746    8
29100203  29100203  1.0129  2.6775        12   38036  3312463746    1
29108443  29108443  1.1474  2.6840        36  602524  3533177779    7
29109993  29109993  1.0240  2.7238        62  658994  6424972551    8
29111539  29111539  1.2032  2.6796        87  262421  3533177779    4
29112154  29112154  1.1070  2.5419       178  687667  4932578245    8
[17710 rows x 7 columns]
分组成功
重置列表成功
取出相同数据样本成功
取出特征值和目标值成功
分为训练集和测试集成功
标准化成功
喂入数据成功
预测的准确率为: 0.7897052353717554
在交叉验证中最好的验证结果: 0.7943376851987678
最好的模型是: KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',
           metric_params=None, n_jobs=1, n_neighbors=5, p=2,
           weights='uniform')
超参数每次交叉验证的结果: {'mean_fit_time': array([0.00540481, 0.00567071, 0.00516579, 0.00522406, 0.00502703,
       0.00583143]), 'std_fit_time': array([4.98834803e-04, 7.77823242e-04, 3.20233846e-04, 3.89286539e-04,
       1.84226608e-05, 1.14072332e-03]), 'mean_score_time': array([0.00691903, 0.00857737, 0.00983589, 0.01078453, 0.01147721,
       0.01263049]), 'std_score_time': array([0.00083247, 0.00047778, 0.00084054, 0.00045022, 0.00047045,
       0.0006643 ]), 'param_n_neighbors': masked_array(data=[1, 3, 5, 7, 9, 11],
             mask=[False, False, False, False, False, False],
       fill_value='?',
            dtype=object), 'params': [{'n_neighbors': 1}, {'n_neighbors': 3}, {'n_neighbors': 5}, {'n_neighbors': 7}, {'n_neighbors': 9}, {'n_neighbors': 11}], 'split0_test_score': array([0.77101449, 0.77826087, 0.78115942, 0.7826087 , 0.7884058 ,
       0.7884058 ]), 'split1_test_score': array([0.74854651, 0.78633721, 0.79069767, 0.79069767, 0.78197674,
       0.78052326]), 'split2_test_score': array([0.7630814 , 0.79069767, 0.78343023, 0.78924419, 0.79215116,
       0.78343023]), 'split3_test_score': array([0.76093294, 0.78862974, 0.79737609, 0.79883382, 0.79154519,
       0.78571429]), 'split4_test_score': array([0.74670571, 0.77891654, 0.7920937 , 0.77598829, 0.76427526,
       0.76720351]), 'split5_test_score': array([0.775     , 0.78823529, 0.80588235, 0.79852941, 0.80294118,
       0.80294118]), 'split6_test_score': array([0.77172312, 0.79381443, 0.79234168, 0.79086892, 0.78645066,
       0.78645066]), 'split7_test_score': array([0.73816568, 0.76775148, 0.77662722, 0.76775148, 0.75591716,
       0.76775148]), 'split8_test_score': array([0.7611276 , 0.78041543, 0.79525223, 0.79970326, 0.78783383,
       0.78486647]), 'split9_test_score': array([0.76820208, 0.79197623, 0.82912333, 0.81129272, 0.80980684,
       0.80089153]), 'mean_test_score': array([0.76045181, 0.78450931, 0.79433769, 0.79052369, 0.78612293,
       0.7848027 ]), 'std_test_score': array([0.01158634, 0.00761673, 0.01402476, 0.01197379, 0.0152013 ,
       0.01104281]), 'rank_test_score': array([6, 5, 1, 2, 3, 4]), 'split0_train_score': array([1.        , 0.87334748, 0.8573527 , 0.83956259, 0.8274849 ,
       0.81867145]), 'split1_train_score': array([1.        , 0.87371512, 0.85658346, 0.83700441, 0.8267254 ,
       0.81595693]), 'split2_train_score': array([1.        , 0.87110458, 0.85250449, 0.83602545, 0.82313591,
       0.81693588]), 'split3_train_score': array([1.        , 0.87228837, 0.85483608, 0.83819931, 0.82531398,
       0.81683249]), 'split4_train_score': array([1.        , 0.87626345, 0.85686338, 0.84170199, 0.82898598,
       0.82132377]), 'split5_train_score': array([1.        , 0.87241323, 0.85383738, 0.83591331, 0.8259736 ,
       0.81766335]), 'split6_train_score': array([1.        , 0.87455197, 0.85386119, 0.83659172, 0.82600196,
       0.81867058]), 'split7_train_score': array([1.        , 0.87510177, 0.85881778, 0.84139391, 0.8262498 ,
       0.81778212]), 'split8_train_score': array([1.        , 0.87156113, 0.85495686, 0.8370503 , 0.82679473,
       0.81458571]), 'split9_train_score': array([1.        , 0.87320964, 0.85432943, 0.83723958, 0.82519531,
       0.81640625]), 'mean_train_score': array([1.        , 0.87335567, 0.85539428, 0.83806826, 0.82618616,
       0.81748285]), 'std_train_score': array([0.        , 0.00152974, 0.00184093, 0.00200913, 0.00146002,
       0.00174073])}