Self-Service Data Science

Dashboards are essential tools for data visualization and informed decision-making. Yet, their potential often remains untapped, especially when it comes to harnessing the power of machine learning (ML). Traditionally, integrating ML into dashboards has been challenging due to varying user expertise and the need for seamless, context-driven solutions. Business leaders and data scientists alike yearn for ways to derive more sophisticated insights from their dashboards without investing extensive time and resources.

We’re introducing machine learning features to our dashboards, ensuring they are accessible to everyone – from business executives to data scientists. With our intuitive ‘One Click’ solution, users can instantly access ML insights. For those seeking a more in-depth experience, we offer integrated Jupyter Notebooks. This balanced approach allows business users to gain insights swiftly, while tech enthusiasts have the flexibility to delve deeper and fine-tune data and algorithms as they see fit.

Enhance Dashboards with a Single Click

Imagine being able to predict future trends or detect anomalies with just a click. The One Click feature is designed with business users in mind. With a simple interface and minimal adjustments, users can instantly generate forecasts and insights.

However, while the One Click approach offers quick insights, it’s essential to understand its limitations. It serves as an initial peek into the data and isn’t a replacement for a comprehensive ML tool. Data quality remains paramount—if you input inconsistent data, the results may not be reliable.

Detect Outliers

Steps:

  1. Open the dashboard in Edit mode.

  2. Open the visualization’s context menu and select Detect outliers.

    Detect Outliers
  3. Set a sensitivity and click Apply.

    Detect Outliers

The algorithm detects and highlight data point outliers:

Detect Outliers

Highlight Data Clusters

Steps:

  1. Open the dashboard in Edit mode.

  2. Open the visualization’s context menu and select Cluster.

    Highlight Data Clusters
  3. Set how many different clusters you want your data to be categorized into and click Apply.

    Highlight Data Clusters

The algorithm detects and highlights the data clusters:

Highlight Data Clusters

Forecast

Steps:

  1. Open the dashboard in Edit mode.

  2. Open the visualization’s context menu and select Forecast.

    Forecast
  3. Set how many periods into the future you want to create a forecast for and click Apply.

    Forecast

The algorithm extrapolates a forecast with estimated error bands:

Forecast

Dive Deeper with Jupyter Notebooks

For those looking for more control and sophistication, the integrated Jupyter Notebook offers an immersive experience. Users can fetch data directly from their dashboard, modify algorithms, preprocess data, or even visualize results—all within the same ecosystem. This integration blurs the lines between data visualization and data science, fostering a cohesive environment for enhanced decision-making.

GoodData ensures that these Jupyter notebooks are accessible to both novices and experts. With guided walkthroughs and clear explanations at every step, users can confidently harness the power of ML to optimize their dashboards.

Usage in the Demo Environment

To create your own visualization that support machine learning in the GoodData Labs demo environment, you need to create either a:

  • line chart or bar chart that includes the string #forecast in its name
  • scatter plot that includes the string #cluster in its name
  • line chart that includes the string #anomalies in its name

when you are creating it in the Analytical Designer. Only one # string per visualization is allowed. You can later rename it in the Dashboard edit view.