Discover the potential of synthetic HR data with these 3 online tools
06.01.2023What if you could have every dataset you wanted, whether to anonymise private data or train AI models? What if you could generate synthetic data with the same distribution as real data, without losing much time? And what if this process were repeatable, so that even your most sensitive and critical datasets remain safe?
One reasonable thought, as we ourselves have emphasised in the past, is that as long as we do not have sufficient volumes of high-quality data, we cannot get to work with the data. To help change this, here are three tools that help both HR data analysts and HR software developers unlock the potential of synthetic data.
Written by Lotte Van der Sijpt
But let us start at the beginning. What are synthetic data? Synthetic data are data created by algorithms, not by people. These algorithms do create the data in a way that is similar to how humans would create them. In other words, synthetic data are a computer-generated version of real data.
The most important thing to understand is that synthetic data and real data are not interchangeable. You cannot simply make HR decisions based on figures that were never actually collected within your organisation. However, synthetic data can be very useful in certain situations, for example when you want to create a mock-up dashboard, share sensitive datasets or even test an AI model without using the real data. It is especially useful in domains where access to good, relevant data is limited, as can sometimes be the case in the world of HR analytics.
1. Starting from scratch with Mockaroo
Mockaroo is an online tool that allows you to create synthetic datasets completely from scratch. It is a great option if you want to start generating test data immediately, without any knowledge of programming languages. One of Mockaroo’s greatest strengths is its ease of use, making it highly suitable for interested HR professionals without first having to obtain a data engineering degree.
Through the Mockaroo platform, you can get started immediately for free. You do not even need a profile. Here, you can create columns and choose from a wide range of data types. From fake app names and marketing slogans to numbers and lists that you can fill in as if they were possible answers to a survey.
Mockaroo, LLC. (2022). Screenshot of Mockaroo.com. Mockaroo - Random Data Generator and API mocking tool. https://www.mockaroo.com/
Mockaroo, LLC. (2022). Screenshot of Mockaroo.com. Mockaroo - Random Data Generator and API mocking tool. https://www.mockaroo.com/
What particularly distinguishes Mockaroo from other free generators is the ability to add your own rules to the data using formulas (similar to Excel). This allows you to create new columns that are the sum of several previous ones, set correlations between variables, choose the distribution of values, as well as the amount of repetition and randomness. In no time, you will have a thousand rows of test data that meet your specific requirements. Once you have created your dataset, you can finally download it in various file formats (including Excel format).
2. Building on existing data with Tonic.ai
Unlike Mockaroo, Tonic.ai is a tool that can help you generate synthetic data from existing data. The advantage of this is that you can create synthetic data that consistently follows the same rules and is just as reliable as your original data, without having to explicitly define those rules yourself.
Tonic.ai is best suited for situations where you have a limited amount of data but would like to get more out of it. This may be because collecting a large amount of data is too expensive or too time-consuming. For example, if you have a database with training scores from different employees or CVs from different applicants, and you want to build a model that can automatically predict who will be the best candidate, then Tonic.ai can help you with this. The disadvantage of Tonic.ai is that a basic knowledge of a programming language is recommended, and it may therefore seem more overwhelming at first sight for beginning data analysts. Fortunately, they do provide clear documentation and videos on their YouTube channel.
For those who would still like to take on the challenge, you can get started with a free demo at djinn.tonic.ai. Here, you import a dataset, which is then used to train a machine learning model. Afterwards, you can continue working with the trained model and the new dataset. One final interesting aspect is the comparison report you receive about the original and new dataset. As illustrated below, you can easily check whether the datasets have similar distributions and even whether the relationships between different variables have remained equally strong!
Kamor, A.;Tonic.ai. (2022). Screenshot from Djinn.tonic.ai. Djinn. http://djinn.tonic.ai/
Kamor, A.; Tonic.ai. (2022). Screenshot from Djinn.tonic.ai. Djinn. http://djinn.tonic.ai/
3. Enjoy the best of both worlds with Mostly.ai
We have just seen how you can manually create synthetic data from scratch and how you can generate it based on existing data through machine learning. The final tool we would therefore like to introduce is Mostly.ai. This online tool can address both of these challenges, making it even more useful for a wide range of use cases. On top of that, Mostly.ai is also very user-friendly and allows you to try a free demo, complete with an onboarding tour. Definitely worth trying!
Mostly.ai. (2022). Screenshot of mostly.ai. MOSTLY AI. https://synthetic.mostly.ai/jo...
Want to know more?
Would you like to learn more about the potential of synthetic data? Then you can certainly start your search with this webinar featuring the founders of Mockaroo and Tonic.ai.
Would you like even more inspiration on HR analytics and AI in HR? Then be sure to subscribe to our bi-monthly newsletter for tips and trends, register for our next free webinar, or contact us.