Knowledge Hub

There is a very thin line between ModelOps and MLOps. ModelOps is the next stage of MLOps where focus is more towards automating, retraining, and maturing models on its own.

MLOps mainly focuses on model management, model monitoring and CI/ CD whereas ModelOps mainly focuses on addressing and sharing AI performance to the stakeholders without the need of Data Science and Machine Learning Team members.

Few differences between ModelOps and MLOps.

• MLOps focuses on ML models operationalization whereas ModelOps focuses on decision models and AI operationalization.
• MLOps focuses on continuous cycle of development, implementation, and monitoring of models whereas ModelOps focuses on life cycle management and governance of models.
• MLOps focuses on creating models for AI enabled applications whereas ModelOps focuses on AI transparency and usage of KPIs by business users.

In nutshell, ModelOps supports AI capabilities where involvement of Data Science become less and less as Models run on their own, retrain on their own, deploy on their own, extract business KPIs on their own etc.

AIOps was initiated by Gartner in 2016. It supports DevOps, DataOps, MLOps, DLOps and ITOps together.

AIOps came to surface after the inception of Big Data era where huge data started to flow in, where DevOps, DataOps, MLOps, DLOps rather every kind of operational tasks had tripled. And as we know everyone has recently realized the importance of Data so now everyone wants to progress in data ignoring the fact that automation requires processes and governance etc. That is why AIOps came into picture where everything is handled by the system itself.

AIOps is about assisting DevOps, DataOps, MLOps and ITOps etc., to work seamlessly. AIOps uses algorithms to make things happen like preventing outages, maintaining uptime and assure continuous services. Unlike DevOps, MLOps and ITOps where human interactions are still required, here AIOps makes sure everything runs on its own.

As mentioned, AIOps handles end to end operations on it owns, below are few stages.

To understand Data Algorithms, let's quickly know what an algorithm is. It is a procedure consisting of multiple steps with a set of instructions for a desired result. It is independent of any language, in other words, an algorithm can be implemented in any language.

Algorithm Characteristics

• Unambiguous: It should be clear and unambiguous. Each of its steps, and its inputs/outputs should be clear and must lead to only one meaning.
• Input: It should have 0 or more well-defined inputs.
• Output: It should have 1 or more well-defined outputs and should match the desired output.
• Finiteness: It must terminate after a finite number of steps.
• Feasibility: It should be feasible with the available resources.
• Independent: It should have step-by-step directions, which should be independent of any programming code.

Now from a data point-of-view, how the algorithm behaves.

• Search: Algorithm to search an item in a data structure.
• Sort: Algorithm to sort items in a certain order.
• Insert: Algorithm to insert an item in a data structure.
• Update: Algorithm to update an existing item in a data structure.
• Delete: Algorithm to delete an existing item from a data structure.

As the name says itself, Data Splitting is about dividing data into multiple datasets which can be for multiple departments, for multiple regions, for multiple time periods etc. but here we will discuss how it’s used for data science?

Data Splitting for data science activities is an important aspect of creating, processing, validating, and testing machine learning models.

In Data Science, mostly 2 approaches are adopted to split datasets.

• Approach#1: Dataset are split in 2 subsets i.e., 1) Training 2) Testing
• Approach#2: Dataset are split in 3 subsets i.e., 1) Training 2) Validating 3) Testing

Let’s understand with an example, consider a dataset contains 100,000 bank customers, now a created model’s aim is to tag each customer under three categories e.g., Silver, Gold or Premium based on their credit history, buying behavior, transaction values etc.

Now let’s see how both approaches on with above example.

This Feature is not that traditional feature 😊. Here, we will be discussing Feature which is an integral part of Data Science activity. I am sure non-data scientists or non-data citizens might not be aware of this term.

We know what Data Science is, it is about running algorithms and machine learning models (ML models) on structured, semi and unstructured data sets for analytics.

For any ML model there is a basic requirement of data, right? And we all know Data Scientists love to play with Raw Data. Now, Raw Data means irrelevant data as well, so Data Scientists need Raw BUT relevant Raw Data. In Big Data, we know data sizes are huge and to train models, one doesn’t need all Raw Data. So, what data scientists do, they themselves or ideally data engineers clean and filter the data, in other words do Data Wrangling, before feeding into ML models.

For example, there is a table with 1 million rows containing data around the globe with all kinds of customers, but ML models is supposed to predict customers buying behavior for a particular country and for a particular set of products. Means, we don’t need to feed 1 million rows to our ML model so now we will create another table (FEATURE) where we will only keep data for that country and for required set of products and now make that table as source for our ML models.

All these terms like Data Integrity, Data Quality, Data Scrubbing, Data Cleansing and Data Cleaning are very interrelated.

• Cleaning: You eat something, and you get something on your face, you clean your face. You are painting something, and you get splashes of paint on your face, you clean your face. You are walking and you feel the dust, you wash and clean your face.
  SAME WAY, data cleaning can be done at a very high level by Business Users themselves as it requires little effort e.g., export data in excel, clean it, and send it downstream.
• Cleansing: You went on safari or played a football match on muddy ground, now you must use some face cream to cleanse your face plus you have to spend 10-15 minutes. Then it is better to ask someone or ideally go to the parlor for good face cleansing.
  SAME WAY, data cleansing can be done by business users themselves but ideally IT SME(s) can write SQL queries at the database level to cleanse data so next time when you extract, you get the right cleansed data.
• Scrubbing: This is the best level to achieve the highest Data Quality. Coming back to the same example, you need to attend a wedding or birthday party or any kind of valuable event, now you will go to a proper parlor, you will find the best worker, you will select the best brand of creams (based on your pocket vs the event you are attending) to scrub your face skin. You will make sure your face has been properly cleaned, cleansed, and scrubbed, and you will make sure you look ready for the event.
  SAME WAY, data scrubbing is to make sure your data is near to perfect, you will find tools that can do this for you, you will find good resources for this task, you will make sure the findings are implemented from Source to Downstream systems.

Data Wrangling or Data Mangling or Data Munging are used interchangeably and do the same function.

Data Wrangling is something that every ETL developer, Data Engineer, Data Analyst etc. do daily. It is something that is done before Data Mining, Data Science, Business Intelligence or Business Analytics.

In simple words, Data Wrangling is to prepare data for processing. Note, prepare data for processing, not process data. Yes, you heard it correctly, sometimes Data Wrangling techniques are used for permanent data fixes and sometimes to prepare data for specific use cases.

Let's take an example, refer to the tables above, fixing and streamlining data in fields i.e., Name, Phone, Birth Date and State will clean the data permanently which will assist in fast job processing, save analysts time for investigation etc. Secondly, you can see row#5 has been removed as, if the analysis is on Date then Year is missing from row#5 so this row will not assist in modelling.

Normally Data Cleansing and Data Enrichment are confused with each other.

‘Data Enrichment is about enhancing the data whereas Data Cleansing is about correcting the data.’

Data Cleansing has been covered in one of the previous topics, so we will focus only on Data Enrichment. As the definition explains itself, Data Enrichment is about adding more data into your first-hand data set from other internal or third-party data sources to add additional context. In this era of social media and unstructured sources, there is always additional data available which can enhance the context of your data generated from internal systems.

‘Data Enrichment must depend on Business Goals.’

For example, if a brand wants to open a new branch, it can buy geographical data from third parties. If a bank wants to offer Credit Cards deals, it can buy crediting rating data from third parties.

Please note, there can be many sources of data sets for Data Enrichment and of course not all can be Enriched in your data stores. Just like any other implementation, Data Enrichment must follow business requirements.

Data Liberalization and Data Democratization are very confusing terms as both work together and most of the time mixed with each other.

‘Data Liberalization ideally is a prerequisite for Data Democratization.’

The aim for Data Democratization is to remove the fear factor about data. Allow everyone to ask analytical questions. To have data access availability for everyone. Allow everyone to raise questions about data and then suggest resolutions, if possible.

Data Liberalization is about providing data in such a way that it should be reusable by users. It can be in any formal i.e., via API, via BI tools, via DaaS etc.

When we say, ‘Aim is to have data access availability for everyone.’ doesn’t mean show everything, of course with access control. Data Liberalization completes the cycle to make sure whatever user is accessing, is USABLE for them rather a pile of junk data.

Data Driven Organization concept can’t be completed just by Data Democratization, Data Liberalization is also a very crucial component to achieve.

Data Entitlement is normally mixed with Data Authorization which is not 100% accurate. There is no doubt, there are few places where both can be used interchangeably but mostly, they should not.

Entitlement is something which you inherit, or which comes as a default or which you get along-with something else whereas Authorization comes because of an action, or it expects an action in future. For example, you are ENTITLED to something and at the same time you are AUTHORIZED to do X, Y and Z things.

Before we understand the difference between Authorization and Entitlement, it’s very importance to understand what Prescriptive and Descriptive behaviors are.

• Prescriptive means, how it should be, and an action is expected.
• Descriptive means, how it is.

‘Authorization has a Prescriptive behavior whereas Entitlement has Descriptive behavior.’

Let’s try to understand with an example, at work you must have heard, you are ENTITLED to 20 annual leaves and maybe 12 medical leaves etc. But when you apply for a leave, your manager has the AUTHORITY to accept or reject your leave.