• 28 January 2021
  • Marcin Miazga
  • Big Data

There is no one definition of data processing that completely and thoroughly explains its purpose and construction. While browsing through the web you can find different approaches to the subject, depending on the point of view. The description will be different if it’s done by a company that sells software for this job, or when it is done by a company that works with data security. It’s often a case that someone is actually already doing data processing, but doesn’t call it that yet.

Of course, there are also parts where definitions meet because some of the stages and purposes of this process are the same independently from the environment in which they are used. But overall and apart from details, data processing appears when data is taken from various raw sources and turned into a readable and analyzable form. Sometimes data processing is also understood as a part of information processing.

Concluding the above considerations with a quote from Data Processing and Information Technology by Carl French: Data Processing is “the collection and manipulation of items of data to produce meaningful information.”

 

Why is data processing becoming so popular?

Nowadays data is something that surrounds us constantly. We produce huge amounts of it ourselves, will it be with our phones, watches, home appliances, or any other device that has a processor unit in it.

All this data, or part of it, may be gathered and stored for future use. This may be for machine learning, analyzing, systems improvement, safety, or any other use we might think of. Because by its nature, this information can be highly sensitive, like personal or medical, processing it raises many challenges on law regulations level. This is why it’s important to understand different ways of data processing when it comes to the kind of information that is being worked on.

We have to consider a wide variety of data sources, which differ by size, change frequency, access type, data format/schema, how it is processed.

To have an overall view of how often we meet data processing ourselves it’s worth mentioning that practically every website that is visited do some kind of data processing, will it be by its own algorithms or by a third party, like for example Google.

It’s indisputable that data processing is needed and performed widely, but why so? What do we gain from it apart from loads of data being processed and stored?

 

Data processing tools

Data processing technological stack can be built on many tools working on different stages of processing, or it can be only one application taking care of all levels from raw to final. Probably one of the most known tools in regards to end-to-end capability and end-user ease of use for data processing would be Power BI from Microsoft. It has many connectors to raw data sources, good processing capabilities, and very intuitive presentation modules.

But, when it comes to choosing data processing software that fulfills business needs it’s important to get good market research, as currently, the choice for such a solution is overwhelming. To do that it’s inevitable to firstly define needs and expectations for processing results, as well as input parameters previously mentioned.

 

Stages of data processing

Data processing is a procedure that may consist of several stages, of which some are always executed, while others might be omitted.

Stages of data processing:

 

Collection

The first most basic step on the path to process data is its gathering from various sources. Those can be flat files, relational databases, IoT devices, cloud storages, and so on. This data will often be unstructured, redundant, badly formed, incomplete, or damaged thus very hard to use. This stage is very important due to the fact that all other steps depend on it. If data that is collected gets any additional damage during this part of the process, then it may be impossible to fix it in the following steps.

 

Preparation

Once data is collected, it usually needs cleaning, specifically some deduplication and data quality checks. This part is responsible for getting most of the rubbish and errors out of what was imported. After that, more advanced processes can start working on it.

 

Input storage

Data prepared in the previous step can be now transferred into its initial storage, where it can be further analyzed and processed. This step can often be omitted, while it is a good practice to store cleaned up information in case some processes need to be rerun. That way we do not need to clean it again.

 

Processing

Most advanced and important step. Data that has been prepared in previous steps, can be now ingested by various tools and processes. This is where algorithms and machine learning can reveal their full potential. This part can be done also more traditionally, by data scientists, who describe procedures for data handling. Both ways need to end up with an organized result that is ready for analysis.

 

Analyzing

After data is processed, it is now available in more readable form than in the beginning, and it can be further analyzed by for example data scientists, and after that presented in eye pleasant and informative form, like for example graphs or reports. In this step full value of the whole process is shown. Based on those results companies can make game-changing decisions or improve their processes.

 

Storage

This highly functional final data needs to be safely stored and at the same time has to be easy to get. Currently, the most popular storage is of course cloud-based.

One can say: “Why implement such a costly process just to show some graphs?”. This question could be valid several years ago, but currently, having access to all this information and being able to analyze it gives unbounded possibilities for companies and governments to gain an advantage in many fields. Especially given that most of our lives and businesses are going online, and that seems to be the current direction for a long time.

 

Author