Pandas笔记--表格利器

# DataFrame数据
df = DataFrame({'a': 1,'b' :2}, index=['1', '2'])
df.columns   # 获取列索引名称
df.values   # 获取数值(ndarray)
df.iloc[:, 1:3]   # 取某列series元素
df.columns.values[0]   # 设置列名称
y = train_data.pop('30').values   # pop将训练数据中的标签值y取出来
col = train_data.columns   
x = train_data[col].values

# Series数据(单列数据)
sr.index   # 获取行索引
sr.name   # 设置列名称
sr.values   # 获取数值(ndarray)

# MinMax标准化
df = df.iloc[:, 1:]
dat = (df - df.min()) / (df.max() - df.min()) 

## classification
kf = RepeatedStratifiedKFold(n_splits=kf)
i = 0
scores = np.zeros([kf*10, 2])
for train_idx, test_idx in kf.split(label_X, label_Y):
    y_train, y_test = label_Y[train_idx], label_Y[test_idx]

    # select features
    model = ExtraTreesClassifier(n_estimators=100)
    model.fit(label_X.iloc[train_idx, :], y_train)
    fim = model.feature_importances_
    b = np.array(sorted(enumerate(fim), key=lambda x: x[1]))
    idx = b[:80, 0].astype(int)

    x_train, x_test = label_X.iloc[train_idx, idx], label_X.iloc[test_idx, idx]

    # KNN classification
    neigh = KNeighborsClassifier(n_neighbors=3).fit(x_train, y_train)
    y_pred = neigh.predict(x_test)
    cm = confusion_matrix(y_test, y_pred)
    scores[i, 0] = np.mean(np.diag(cm) / np.sum(cm, axis=1))
    scores[i, 1] = f1_score(y_test, y_pred, average='weighted')

    i = i + 1

print(np.mean(scores, axis=0), np.std(scores, axis=0))

## regression
scores = np.zeros([10, 2])
for i in range(10):
    x_train, x_test, y_train, y_test = train_test_split(label_X, label_Y, test_size=0.3)

    # SVR regression
    svr_reg = SVR(kernel='linear')
    svr_reg.fit(x_train, y_train)
    y_pred = svr_reg.predict(x_test)
    scores[i, 0] = mean_squared_error(y_test, y_pred)
    scores[i, 1] = mean_absolute_error(y_test, y_pred)

print(np.mean(scores, axis=0), np.std(scores, axis=0))