# GBDT-特征重要性 
================================================================================

:download:`参考JSON<json/gbdtfeature.json>`

说明：GBDT目前仅支持二分类，LabelCol必须非0即1

## 参数

+------------------------+--------+----------+-------------------------------------------+
| 名称                   | 类型   | 是否必传 | 备注                                      |
+------------------------+--------+----------+-------------------------------------------+
| inputarg               | array  | Y        | 输入集合                                  |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.tbname1     | string | Y        | 表名                                      |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.labelcol    | string | Y        | 目标特征列                                |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.maxdepth    | int    | Y        | 最大树深度                                |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.mininfogain | double | Y        | 最小信息增益                              |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.impurity    | string | Y        | 计算信息增益的准则，entropy,gini,variance |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.iter        | int    | Y        | 最大迭代次数                              |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.rate        | double | Y        | 设置采样率                                |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.maxbins     | int    | Y        | 最大分桶数                                |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.stepsize    | double | Y        | 设置学习率                                |
+------------------------+--------+----------+-------------------------------------------+
| inputarg.#.columnArray | array  | Y        | 字段名称集合                              |
+------------------------+--------+----------+-------------------------------------------+


## 返回

+--------+--------+----------+--------------------------------+
| 名称   | 类型   | 是否必传 | 备注                           |
+--------+--------+----------+--------------------------------+
| result | String | Y        | 所有输出的JSON，按节点类型解析 |
+--------+--------+----------+--------------------------------+

## 算法入口及参数

+----------+------------------------------------------------------------------+--------------------+
| 名称     | 示例                                                             | 备注               |
+----------+------------------------------------------------------------------+--------------------+
| args[0]  | tmptb2019062511473300088201                                      | 输入表名           |
+----------+------------------------------------------------------------------+--------------------+
| args[1]  | sepal_length,sepal_width,petal_length,petal_width                | 字段集合           |
+----------+------------------------------------------------------------------+--------------------+
| args[2]  | result                                                           | 目标列             |
+----------+------------------------------------------------------------------+--------------------+
| args[3]  | 3                                                                | 最大树深度         |
+----------+------------------------------------------------------------------+--------------------+
| args[4]  | 0.1                                                              | 最小信息增益       |
+----------+------------------------------------------------------------------+--------------------+
| args[5]  | gini                                                             | 计算信息增益的准则 |
+----------+------------------------------------------------------------------+--------------------+
| args[6]  | 10                                                               | 最大迭代次数       |
+----------+------------------------------------------------------------------+--------------------+
| args[7]  | 1                                                                | 设置采样率         |
+----------+------------------------------------------------------------------+--------------------+
| args[8]  | 2                                                                | 最大分桶数         |
+----------+------------------------------------------------------------------+--------------------+
| args[9]  | 0.0001                                                           | 学习率             |
+----------+------------------------------------------------------------------+--------------------+
| args[10] | hdfs://namenode:9000/ispatial/sysfile/tmpv2019062605545200088312 | 结果输出地址       |
+----------+------------------------------------------------------------------+--------------------+