Big Data & AI Paris article – Dataiku

Big Data & AI Paris: Focus on Dataiku

We were also able to learn more about Dataiku, a unified Data Science platform and the event’s platinum sponsor, notably through the feedback of a French energy player. The French electricity giant uses Dataiku in addition to Power BI to democratise data internally and deliver high value-added customer use cases.

Feedback from a French energy player

The company’s Head of Innovation Labs explains that these two solutions were chosen for their accessibility and ease of use, as both are no-code/low-code platforms. Both pieces of software are fed by several data capture solutions and complement each other to meet users’ needs: Dataiku is used for data blending (Data Science), while Power BI is used to create dashboards (Data Visualisation).

One advantage cited for Dataiku is the platform’s interface, which is presented as a visual pipeline. The different stages of this pipeline are clearly displayed and serve as documentation when an employee takes over the existing project, making it a sustainable solution.

Let us now see how this company uses Dataiku in practice to create value through two use cases:

Use case No. 1: Anticipating reception workload => predictive analysis

• Finding: 8 million customer contacts on the platforms with significant variations
• Objective: Model reception activity workload by taking into account influencing factors and historical figures

Use case No. 2: Steering connection actions => descriptive analysis

• Finding: Very strong growth in connections while working to reduce lead times
• Objective: Compare and identify over time whether improvement actions are truly delivering results.

What’s new in Dataiku DSS 12

In addition, we were able to speak with Dataikers who presented the main new features of the latest version of Dataiku DSS (version 12):

Generative AI 

An OpenAI ChatGPT plugin is available to build natural language models accessible to non-coders: text generation, classification and summarisation, as well as drafting answers to one or more question(s).

Automatic feature generation 

The “Generate features” recipe automatically generates new features from your datasets while avoiding prediction loss. Feature generation can include transformations such as, for example, numerical aggregations, extracting parts of dates, and calculating values within certain time windows. This recipe requires a primary dataset as well as an enrichment dataset.

Universal feature explainability 

Feature importance visualisations provide consistent and comparable explanations for all model types.

Uplift modeling

This machine learning capability available in the Lab makes it possible to measure cause-and-effect relationships and estimate the impact of an intervention on outcomes.

New governance views 

You can now get an overview of the deployment stage of all your Dataiku projects in a Kanban board. This view makes it easier to track all governed projects.

Our experts can support you in implementing these new features.

If you would like to learn more about the solution and/or its new features, you can contact us at the following email address ybusidan@meltone.com