BMC Helix ITSM Insights 21.05

Proactive Problem Management flow starts with job configuration that specifies lookback period of historic incident data to use for the clustering algorithm, number of clusters and fields.
During the job execution, incident data is extracted from BMC Helix ITSM and processed through a data pipeline where tokenization, cleaning, stemming, lemmatization, stop-word removal and vectorization takes place. Then a machine learning algorithm (k-means) is used to group the incident data into clusters that groups most similar incidents together using distance-based
metrics and these cluster results are written back to the Elasticsearch storage data lake. Each such cluster of incidents also is given a caption based on word frequency.

Clustering algorithm is a machine learning algorithm that groups incidents together based on the similarity of incidents on the Description or Summary field even if their text does not exactly match. For instance, if you have two incidents with similar description such as “I cannot connect to Skype” and “Skype does not start”, they will be clustered together.

After running the algorithm, you can visualize hundreds of clusters. For example, you might have clusters named “Skype-connect-issue”, “VPN-connectivity-fail”, “VPN-password-reset” and so on with each cluster having a set of incidents that are closely related to each other.

For proactive problem management in BMC Helix ITSM Insights, we have applied the clustering algorithm to closed incidents to identify interesting clusters which are candidate for ‘problems’.

We use an industry standard open source k-means algorithm to do clustering. We also use state-of-the-art techniques for NLP as a part of the algorithm design.

We use a machine learning algorithm to automatically determine the name of the cluster. The algorithm checks for the most important and frequently-used words in the text of all incidents in each cluster and creates a name for each cluster as a three-part name.

Examples: “platform-restart-container", “VPN-password-reset"

To begin with, you can look at the biggest cluster with maximum number of incidents. BMC Helix ITSM Insights also provides filtering ability to slice and dice the data. The various factors to visualize the data can be service (if you have defined service in incidents), trend, and the total time for resolution.

Trend is useful to find those clusters that have increasing trends on number of incidents. Service allows you to focus on specific services. Time spent on incidents can help visualize the amount of time service agents spend on resolving similar incidents.

BMC Helix ITSM Insights caters to both data scientists and problem managers who may not be data scientists. For data scientists, there is an advanced configuration where you can set parameters to finetune machine learning.

The clustering algorithm can be fine-tuned by proper selection of parameters. The most important parameter is the number of clusters, k. The higher the k value you choose, the more granular the clusters become, while a smaller k value leads to more coarse grained clusters.

You can start with coarse grained clusters and then fine tune it by increasing the k. If you don’t specify k, the system will automatically choose it.

The other parameters that also can impact the clustering quality are the fields selected for text clustering. Summary, Description, Notes are a few fields that can be used as the primary field for clustering. In some cases, you can also use the Resolution field for clustering. The system allows you to select the fields for clustering.

Clustering algorithm can be fine tuned by proper selection of parameters. The most important parameters are the fields used for clustering as well as the similarity threshold set in the configuration page.

In addition to text based clustering, if proper bucketing of data is needed for separation of clusters by services or other categories, level 1 group-by fields can also be added. Adding these fields segregates the clusters better by the field value.

If you know your stop words that you would like to ignore while clustering, you can upload stop words for configuring the job. This will ensure that your stop words are not used for clustering.

You can specify a group-by level 1 field while configuring the job. For example, if you use Service, Product Name, Product Categorization, or Operational Categorization, or a custom field in the incident record, you can perform a first level grouping using any of these fields and then apply clustering using machine learning (text) within each service, product name, or category. You can also group by any other fields that have menus or are enumerated fields or even custom fields.

BMC Helix ITSM Insights supports other languages besides English. From the configuration options, you can switch to a new language such as French, German, etc.

Yes, if you have enabled Proactive Service Resolution (PSR) integration, Helix Operations Management events can automatically be converted to incidents that are then used in clustering.

The BMC Helix ITSM permission model is maintained when running and visualizing AI algorithm results both at the UI and data access levels.

User access level: Proactive problem management is only available to those users in BMC Helix ITSM who have the appropriate problem management permissions assigned to them. Real-time incident correlation is only available to BMC Helix ITSM users who have the appropriate incident management permissions.
Data row level access level: In Proactive problem management, a job is configured in the context of the user who creates the job and only those incidents to which the the user has access are processed. This ensures that row level security for each of the incident records is enforced. For real-time incident correlation, the company is used to restrict the incidents in the cluster that a user can see.

No. Right now, the data needs to be in BMC Helix ITSM before it can be consumed by the AI services.

No. Right now, the algorithms provided by AI services can only be developed and deployed by BMC.

The PDF for BMC Helix ITSM Insights is located here. The online documentation is updated regularly, while the PDF contains a snapshot of the content at a particular point in time.

The BMC Documentation portal gives you the ability to generate PDF exports of multiple pages in a space.  

Creating PDF exports

You can create a PDF of a page or a set of pages. (Non-English page exports are not supported.) You can also create a Word document of the current page.

To export to PDF

1. Click the Export icon in the upper-right. 
2. From the Export menu, select PDF.
3. Follow the prompts for page selection and other options.
4. Click Export.

Depending on the number of topics included in the export, it might take several minutes to create the PDF. Once the export is complete, you can download the PDF.