OpenSearch/docs/reference/ml/anomaly-detection/categories.asciidoc

302 lines
13 KiB
Plaintext

[role="xpack"]
[[ml-configuring-categories]]
=== Categorizing data
Categorization is a {ml} process that considers a tokenization of a field,
clusters similar data together, and classifies them into categories. However,
categorization doesn't work equally well on different data types. It works
best on machine-written messages and application outputs, typically on data that
consists of repeated elements, for example log messages for the purpose of
system troubleshooting. Log categorization groups unstructured log messages into
categories, then you can use {anomaly-detect} to model and identify rare or
unusual counts of log message categories.
Categorization is tuned to work best on data like log messages by taking token
order into account, not considering synonyms, and including stop words in its
analysis. Complete sentences in human communication or literary text (for
example emails, wiki pages, prose, or other human generated content) can be
extremely diverse in structure. Since categorization is tuned for machine data
it will give poor results on such human generated data. For example, the
categorization job would create so many categories that couldn't be handled
effectively. Categorization is _not_ natural language processing (NLP).
[float]
[[ml-categorization-log-messages]]
==== Categorizing log messages
Application log events are often unstructured and contain variable data. For
example:
//Obtained from it_ops_new_app_logs.json
[source,js]
----------------------------------
{"time":1454516381000,"message":"org.jdbi.v2.exceptions.UnableToExecuteStatementException: com.mysql.jdbc.exceptions.MySQLTimeoutException: Statement cancelled due to timeout or client request [statement:\"SELECT id, customer_id, name, force_disabled, enabled FROM customers\"]","type":"logs"}
----------------------------------
//NOTCONSOLE
You can use {ml} to observe the static parts of the message, cluster similar
messages together, and classify them into message categories.
The {ml} model learns what volume and pattern is normal for each category over
time. You can then detect anomalies and surface rare events or unusual types of
messages by using count or rare functions. For example:
//Obtained from it_ops_new_app_logs.sh
[source,console]
----------------------------------
PUT _ml/anomaly_detectors/it_ops_new_logs
{
"description" : "IT Ops Application Logs",
"analysis_config" : {
"categorization_field_name": "message", <1>
"bucket_span":"30m",
"detectors" :[{
"function":"count",
"by_field_name": "mlcategory", <2>
"detector_description": "Unusual message counts"
}],
"categorization_filters":[ "\\[statement:.*\\]"]
},
"analysis_limits":{
"categorization_examples_limit": 5
},
"data_description" : {
"time_field":"time",
"time_format": "epoch_ms"
}
}
----------------------------------
// TEST[skip:needs-licence]
<1> The `categorization_field_name` property indicates which field will be
categorized.
<2> The resulting categories are used in a detector by setting `by_field_name`,
`over_field_name`, or `partition_field_name` to the keyword `mlcategory`. If you
do not specify this keyword in one of those properties, the API request fails.
The optional `categorization_examples_limit` property specifies the
maximum number of examples that are stored in memory and in the results data
store for each category. The default value is `4`. Note that this setting does
not affect the categorization; it just affects the list of visible examples. If
you increase this value, more examples are available, but you must have more
storage available. If you set this value to `0`, no examples are stored.
The optional `categorization_filters` property can contain an array of regular
expressions. If a categorization field value matches the regular expression, the
portion of the field that is matched is not taken into consideration when
defining categories. The categorization filters are applied in the order they
are listed in the job configuration, which allows you to disregard multiple
sections of the categorization field value. In this example, we have decided that
we do not want the detailed SQL to be considered in the message categorization.
This particular categorization filter removes the SQL statement from the
categorization algorithm.
If your data is stored in {es}, you can create an advanced {anomaly-job} with
these same properties:
[role="screenshot"]
image::images/ml-category-advanced.jpg["Advanced job configuration options related to categorization"]
NOTE: To add the `categorization_examples_limit` property, you must use the
**Edit JSON** tab and copy the `analysis_limits` object from the API example.
[float]
[[ml-configuring-analyzer]]
===== Customizing the categorization analyzer
Categorization uses English dictionary words to identify log message categories.
By default, it also uses English tokenization rules. For this reason, if you use
the default categorization analyzer, only English language log messages are
supported, as described in the <<ml-limitations>>.
You can, however, change the tokenization rules by customizing the way the
categorization field values are interpreted. For example:
[source,console]
----------------------------------
PUT _ml/anomaly_detectors/it_ops_new_logs2
{
"description" : "IT Ops Application Logs",
"analysis_config" : {
"categorization_field_name": "message",
"bucket_span":"30m",
"detectors" :[{
"function":"count",
"by_field_name": "mlcategory",
"detector_description": "Unusual message counts"
}],
"categorization_analyzer":{
"char_filter": [
{ "type": "pattern_replace", "pattern": "\\[statement:.*\\]" } <1>
],
"tokenizer": "ml_classic", <2>
"filter": [
{ "type" : "stop", "stopwords": [
"Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday",
"Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun",
"January", "February", "March", "April", "May", "June", "July", "August", "September", "October", "November", "December",
"Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec",
"GMT", "UTC"
] } <3>
]
}
},
"analysis_limits":{
"categorization_examples_limit": 5
},
"data_description" : {
"time_field":"time",
"time_format": "epoch_ms"
}
}
----------------------------------
// TEST[skip:needs-licence]
<1> The
{ref}/analysis-pattern-replace-charfilter.html[`pattern_replace` character filter]
here achieves exactly the same as the `categorization_filters` in the first
example.
<2> The `ml_classic` tokenizer works like the non-customizable tokenization
that was used for categorization in older versions of machine learning. If you
want the same categorization behavior as older versions, use this property
value.
<3> By default, English day or month words are filtered from log messages before
categorization. If your logs are in a different language and contain
dates, you might get better results by filtering the day or month words in your
language.
The optional `categorization_analyzer` property allows even greater customization
of how categorization interprets the categorization field value. It can refer to
a built-in {es} analyzer or a combination of zero or more character filters,
a tokenizer, and zero or more token filters. If you omit the
`categorization_analyzer`, the following default values are used:
[source,console]
--------------------------------------------------
POST _ml/anomaly_detectors/_validate
{
"analysis_config" : {
"categorization_analyzer" : {
"tokenizer" : "ml_classic",
"filter" : [
{ "type" : "stop", "stopwords": [
"Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday",
"Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun",
"January", "February", "March", "April", "May", "June", "July", "August", "September", "October", "November", "December",
"Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec",
"GMT", "UTC"
] }
]
},
"categorization_field_name": "message",
"detectors" :[{
"function":"count",
"by_field_name": "mlcategory"
}]
},
"data_description" : {
}
}
--------------------------------------------------
If you specify any part of the `categorization_analyzer`, however, any omitted
sub-properties are _not_ set to default values.
The `ml_classic` tokenizer and the day and month stopword filter are more or
less equivalent to the following analyzer, which is defined using only built-in
{es} {ref}/analysis-tokenizers.html[tokenizers] and
{ref}/analysis-tokenfilters.html[token filters]:
[source,console]
----------------------------------
PUT _ml/anomaly_detectors/it_ops_new_logs3
{
"description" : "IT Ops Application Logs",
"analysis_config" : {
"categorization_field_name": "message",
"bucket_span":"30m",
"detectors" :[{
"function":"count",
"by_field_name": "mlcategory",
"detector_description": "Unusual message counts"
}],
"categorization_analyzer":{
"tokenizer": {
"type" : "simple_pattern_split",
"pattern" : "[^-0-9A-Za-z_.]+" <1>
},
"filter": [
{ "type" : "pattern_replace", "pattern": "^[0-9].*" }, <2>
{ "type" : "pattern_replace", "pattern": "^[-0-9A-Fa-f.]+$" }, <3>
{ "type" : "pattern_replace", "pattern": "^[^0-9A-Za-z]+" }, <4>
{ "type" : "pattern_replace", "pattern": "[^0-9A-Za-z]+$" }, <5>
{ "type" : "stop", "stopwords": [
"",
"Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday",
"Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun",
"January", "February", "March", "April", "May", "June", "July", "August", "September", "October", "November", "December",
"Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec",
"GMT", "UTC"
] }
]
}
},
"analysis_limits":{
"categorization_examples_limit": 5
},
"data_description" : {
"time_field":"time",
"time_format": "epoch_ms"
}
}
----------------------------------
// TEST[skip:needs-licence]
<1> Tokens basically consist of hyphens, digits, letters, underscores and dots.
<2> By default, categorization ignores tokens that begin with a digit.
<3> By default, categorization also ignores tokens that are hexadecimal numbers.
<4> Underscores, hyphens, and dots are removed from the beginning of tokens.
<5> Underscores, hyphens, and dots are also removed from the end of tokens.
The key difference between the default `categorization_analyzer` and this
example analyzer is that using the `ml_classic` tokenizer is several times
faster. The difference in behavior is that this custom analyzer does not include
accented letters in tokens whereas the `ml_classic` tokenizer does, although
that could be fixed by using more complex regular expressions.
If you are categorizing non-English messages in a language where words are
separated by spaces, you might get better results if you change the day or month
words in the stop token filter to the appropriate words in your language. If you
are categorizing messages in a language where words are not separated by spaces,
you must use a different tokenizer as well in order to get sensible
categorization results.
It is important to be aware that analyzing for categorization of machine
generated log messages is a little different from tokenizing for search.
Features that work well for search, such as stemming, synonym substitution, and
lowercasing are likely to make the results of categorization worse. However, in
order for drill down from {ml} results to work correctly, the tokens that the
categorization analyzer produces must be similar to those produced by the search
analyzer. If they are sufficiently similar, when you search for the tokens that
the categorization analyzer produces then you find the original document that
the categorization field value came from.
NOTE: To add the `categorization_analyzer` property in {kib}, you must use the
**Edit JSON** tab and copy the `categorization_analyzer` object from one of the
API examples above.
[float]
[[ml-viewing-categories]]
===== Viewing categorization results
After you open the job and start the {dfeed} or supply data to the job, you can
view the categorization results in {kib}. For example:
[role="screenshot"]
image::images/ml-category-anomalies.jpg["Categorization example in the Anomaly Explorer"]
For this type of job, the **Anomaly Explorer** contains extra information for
each anomaly: the name of the category (for example, `mlcategory 11`) and
examples of the messages in that category. In this case, you can use these
details to investigate occurrences of unusually high message counts for specific
message categories.