Machine Learning Lab 3: I am getting slightly different results

I ran the machine learning job with all of the detectors configured and correct bucket_span selected but I am getting different results from what's in the answers. For question 10, they are quite a bit different. Below is what I see. Any suggestions of what can be the issue?

Could you provide the job you're inputing for this exercise?

Here is JSON for the machine learning job from question 10. Is that what you were looking for?

{
"job_id": "second-advanced-job",
"job_type": "anomaly_detector",
"job_version": "5.6.4",
"description": "",
"create_time": 1510693262890,
"finished_time": 1510693295263,
"analysis_config": {
"bucket_span": "1h",
"detectors": [
{
"detector_description": "mean(CPUUtilization) by "instance.keyword"",
"function": "mean",
"field_name": "CPUUtilization",
"by_field_name": "instance.keyword",
"detector_rules": ,
"detector_index": 0
},
{
"detector_description": "Awesome description",
"function": "mean",
"field_name": "DiskReadBytes",
"by_field_name": "instance.keyword",
"detector_rules": ,
"detector_index": 1
},
{
"detector_description": "mean(DiskReadOps) by "instance.keyword"",
"function": "mean",
"field_name": "DiskReadOps",
"by_field_name": "instance.keyword",
"detector_rules": ,
"detector_index": 2
},
{
"detector_description": "mean(DiskWriteBytes) by "instance.keyword"",
"function": "mean",
"field_name": "DiskWriteBytes",
"by_field_name": "instance.keyword",
"detector_rules": ,
"detector_index": 3
},
{
"detector_description": "mean(DiskWriteOps) by "instance.keyword"",
"function": "mean",
"field_name": "DiskWriteOps",
"by_field_name": "instance.keyword",
"detector_rules": ,
"detector_index": 4
},
{
"detector_description": "mean(NetworkOut) by "instance.keyword"",
"function": "mean",
"field_name": "NetworkOut",
"by_field_name": "instance.keyword",
"detector_rules": ,
"detector_index": 5
},
{
"detector_description": "mean(NetworkIn) by "instance.keyword"",
"function": "mean",
"field_name": "NetworkIn",
"by_field_name": "instance.keyword",
"detector_rules": ,
"detector_index": 6
}
],
"influencers": [
"instance.keyword",
"region.keyword"
]
},
"data_description": {
"time_field": "@timestamp",
"time_format": "epoch_ms"
},
"model_snapshot_retention_days": 1,
"model_snapshot_id": "1510693293",
"results_index_name": "custom-second-advanced-job",
"data_counts": {
"job_id": "second-advanced-job",
"processed_record_count": 19846,
"processed_field_count": 177967,
"input_bytes": 5264594,
"input_field_count": 177967,
"invalid_date_count": 0,
"missing_field_count": 647,
"out_of_order_timestamp_count": 0,
"empty_bucket_count": 0,
"sparse_bucket_count": 0,
"bucket_count": 349,
"earliest_record_timestamp": 1445990400000,
"latest_record_timestamp": 1447248300000,
"last_data_time": 1510693292277,
"input_record_count": 19846
},
"model_size_stats": {
"job_id": "second-advanced-job",
"result_type": "model_size_stats",
"model_bytes": 1132730,
"total_by_field_count": 37,
"total_over_field_count": 0,
"total_partition_field_count": 8,
"bucket_allocation_failures_count": 0,
"memory_status": "ok",
"log_time": 1510693292000,
"timestamp": 1447243200000
},
"datafeed_config": {
"datafeed_id": "datafeed-second-advanced-job",
"job_id": "second-advanced-job",
"query_delay": "60s",
"frequency": "600s",
"indices": [
"cloudwatch-*"
],
"types": [
"usage"
],
"query": {
"term": {
"region.keyword": {
"value": "eu-central-1",
"boost": 1
}
}
},
"scroll_size": 1000,
"chunking_config": {
"mode": "auto"
},
"state": "stopped"
},
"state": "closed"
}

I've been looking into this and can't quite figure it out yet. I'm going to continue to look. Double check everything is configured exactaly as the instructions say - I can't seem to reproduce.

It seems like all of my setting are correct. Could it be that if my computer is not powerful enough, machine learning uses some more simplified versions of calculations which can cause different results?

What locale + time zone are you running? Perhaps there is an offset

I am in USA, Eastern Time.

That isn't the issue then, thanks

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.