加权平均聚合
编辑加权平均聚合
编辑一个 单值 指标聚合，计算从聚合文档中提取的数值的加权平均值。这些值可以从文档中的特定数值字段中提取。
在计算常规平均值时，每个数据点都有相同的“权重”……它对最终值的贡献是相同的。另一方面，加权平均值对每个数据点赋予不同的权重。每个数据点对最终值的贡献量是从文档中提取的。
用公式表示，加权平均值是 ∑(值 * 权重) / ∑(权重)
可以把常规平均值看作是每个值都具有隐式权重 1 的加权平均值。
表 53. weighted_avg 参数
| 参数名称 | 描述 | 必需 | 默认值 |
|---|---|---|---|
|
用于提供值的字段或脚本的配置 |
必需 |
|
|
用于提供权重的字段或脚本的配置 |
必需 |
|
|
数值响应格式化程序 |
可选 |
value 和 weight 对象具有每个字段特定的配置
示例
编辑如果我们的文档有一个保存 0-100 数值分数的 "grade" 字段和一个保存任意数值权重的 "weight" 字段，我们可以使用以下方法计算加权平均值
resp = client.search(
index="exams",
size=0,
aggs={
"weighted_grade": {
"weighted_avg": {
"value": {
"field": "grade"
},
"weight": {
"field": "weight"
}
}
}
},
)
print(resp)
response = client.search(
index: 'exams',
body: {
size: 0,
aggregations: {
weighted_grade: {
weighted_avg: {
value: {
field: 'grade'
},
weight: {
field: 'weight'
}
}
}
}
}
)
puts response
const response = await client.search({
index: "exams",
size: 0,
aggs: {
weighted_grade: {
weighted_avg: {
value: {
field: "grade",
},
weight: {
field: "weight",
},
},
},
},
});
console.log(response);
POST /exams/_search
{
"size": 0,
"aggs": {
"weighted_grade": {
"weighted_avg": {
"value": {
"field": "grade"
},
"weight": {
"field": "weight"
}
}
}
}
}
这将产生如下响应
{
...
"aggregations": {
"weighted_grade": {
"value": 70.0
}
}
}
虽然允许每个字段有多个值，但只允许有一个权重。如果聚合遇到一个具有多个权重的文档（例如，权重字段是一个多值字段），它将中止搜索。如果遇到这种情况，应该构建一个 运行时字段 将这些值组合成一个权重。
这个单一权重将独立应用于从 value 字段中提取的每个值。
此示例展示了如何对具有多个值的单个文档使用单一权重进行平均
resp = client.index(
index="exams",
refresh=True,
document={
"grade": [
1,
2,
3
],
"weight": 2
},
)
print(resp)
resp1 = client.search(
index="exams",
size=0,
aggs={
"weighted_grade": {
"weighted_avg": {
"value": {
"field": "grade"
},
"weight": {
"field": "weight"
}
}
}
},
)
print(resp1)
response = client.index(
index: 'exams',
refresh: true,
body: {
grade: [
1,
2,
3
],
weight: 2
}
)
puts response
response = client.search(
index: 'exams',
body: {
size: 0,
aggregations: {
weighted_grade: {
weighted_avg: {
value: {
field: 'grade'
},
weight: {
field: 'weight'
}
}
}
}
}
)
puts response
const response = await client.index({
index: "exams",
refresh: "true",
document: {
grade: [1, 2, 3],
weight: 2,
},
});
console.log(response);
const response1 = await client.search({
index: "exams",
size: 0,
aggs: {
weighted_grade: {
weighted_avg: {
value: {
field: "grade",
},
weight: {
field: "weight",
},
},
},
},
});
console.log(response1);
POST /exams/_doc?refresh
{
"grade": [1, 2, 3],
"weight": 2
}
POST /exams/_search
{
"size": 0,
"aggs": {
"weighted_grade": {
"weighted_avg": {
"value": {
"field": "grade"
},
"weight": {
"field": "weight"
}
}
}
}
}
三个值（1、2 和 3）将被作为独立值包括在内，所有值的权重均为 2
{
...
"aggregations": {
"weighted_grade": {
"value": 2.0
}
}
}
聚合返回 2.0 作为结果，这与我们手动计算的预期结果一致：((1*2) + (2*2) + (3*2)) / (2+2+2) == 2
运行时字段
编辑如果必须对与索引值不太匹配的值进行求和或加权，请在 运行时字段 上运行聚合。
resp = client.index(
index="exams",
refresh=True,
document={
"grade": 100,
"weight": [
2,
3
]
},
)
print(resp)
resp1 = client.index(
index="exams",
refresh=True,
document={
"grade": 80,
"weight": 3
},
)
print(resp1)
resp2 = client.search(
index="exams",
filter_path="aggregations",
size=0,
runtime_mappings={
"weight.combined": {
"type": "double",
"script": "\n double s = 0;\n for (double w : doc['weight']) {\n s += w;\n }\n emit(s);\n "
}
},
aggs={
"weighted_grade": {
"weighted_avg": {
"value": {
"script": "doc.grade.value + 1"
},
"weight": {
"field": "weight.combined"
}
}
}
},
)
print(resp2)
response = client.index(
index: 'exams',
refresh: true,
body: {
grade: 100,
weight: [
2,
3
]
}
)
puts response
response = client.index(
index: 'exams',
refresh: true,
body: {
grade: 80,
weight: 3
}
)
puts response
response = client.search(
index: 'exams',
filter_path: 'aggregations',
body: {
size: 0,
runtime_mappings: {
'weight.combined' => {
type: 'double',
script: "\n double s = 0;\n for (double w : doc['weight']) {\n s += w;\n }\n emit(s);\n "
}
},
aggregations: {
weighted_grade: {
weighted_avg: {
value: {
script: 'doc.grade.value + 1'
},
weight: {
field: 'weight.combined'
}
}
}
}
}
)
puts response
const response = await client.index({
index: "exams",
refresh: "true",
document: {
grade: 100,
weight: [2, 3],
},
});
console.log(response);
const response1 = await client.index({
index: "exams",
refresh: "true",
document: {
grade: 80,
weight: 3,
},
});
console.log(response1);
const response2 = await client.search({
index: "exams",
filter_path: "aggregations",
size: 0,
runtime_mappings: {
"weight.combined": {
type: "double",
script:
"\n double s = 0;\n for (double w : doc['weight']) {\n s += w;\n }\n emit(s);\n ",
},
},
aggs: {
weighted_grade: {
weighted_avg: {
value: {
script: "doc.grade.value + 1",
},
weight: {
field: "weight.combined",
},
},
},
},
});
console.log(response2);
POST /exams/_doc?refresh
{
"grade": 100,
"weight": [2, 3]
}
POST /exams/_doc?refresh
{
"grade": 80,
"weight": 3
}
POST /exams/_search?filter_path=aggregations
{
"size": 0,
"runtime_mappings": {
"weight.combined": {
"type": "double",
"script": """
double s = 0;
for (double w : doc['weight']) {
s += w;
}
emit(s);
"""
}
},
"aggs": {
"weighted_grade": {
"weighted_avg": {
"value": {
"script": "doc.grade.value + 1"
},
"weight": {
"field": "weight.combined"
}
}
}
}
}
应该如下所示
{
"aggregations": {
"weighted_grade": {
"value": 93.5
}
}
}
缺失值
编辑默认情况下，聚合会排除 value 或 weight 字段的值缺失或为 null 的文档。可以使用 missing 参数来为这些文档指定默认值。
resp = client.search(
index="exams",
size=0,
aggs={
"weighted_grade": {
"weighted_avg": {
"value": {
"field": "grade",
"missing": 2
},
"weight": {
"field": "weight",
"missing": 3
}
}
}
},
)
print(resp)
const response = await client.search({
index: "exams",
size: 0,
aggs: {
weighted_grade: {
weighted_avg: {
value: {
field: "grade",
missing: 2,
},
weight: {
field: "weight",
missing: 3,
},
},
},
},
});
console.log(response);
POST /exams/_search
{
"size": 0,
"aggs": {
"weighted_grade": {
"weighted_avg": {
"value": {
"field": "grade",
"missing": 2
},
"weight": {
"field": "weight",
"missing": 3
}
}
}
}
}