Qlik AutoML (automated machine learning) brings AI-generated machine learning models and predictive analytics directly to your organisation’s larger community of analytics users and teams, in a simple user experience focused on augmenting their intuition through machine intelligence. With AutoML, you can easily generate machine learning models, make predictions, and plan decisions – all within an intuitive, code-free user interface.
Qlik AutoML easily profiles data, identifies key drivers in the dataset, and generates models. You can then make future predictions, complete with prediction influencer data (Shapley values) at the record-level, allowing you understand why predictions were made – which is critical to making the best decisions and taking the right actions.
Predictive data can be easily published into Qlik Sense® and other cloud platforms, and models can be integrated using advanced analytics integration for real-time exploratory analysis and what-if scenario planning. With Qlik AutoML, your analytics teams can go beyond descriptive analysis to predictive and prescriptive analytics, with detailed insight that’s uniquely powerful when combined with our best-in-class, associative exploration.
What is AutoML
Machine learning (ML) is a branch of artificial intelligence (AI) focused on the process of recognizing patterns in historical data to predict outcomes in the future. ML uses historically observed data as an input, applies a mathematical process against that data, and creates an output called a machine learning model based on patterns in historical data.
This model can then be used to make future predictions and test scenarios. AutoML, or automated machine learning, is an approach to machine learning that automates the data pipeline, data processing steps, feature selection, algorithm selection, model training, model hosting, and deployment processes through the use of AI.
Designed specifically for data analytics teams, Qlik AutoML empowers analysts and analytics teams to make predictions about business outcomes, understand why those predictions were made, and take the most effective action to influence those outcomes.
This capability is known as prescriptive analytics. Industries and business functions worldwide use machine learning (and now AutoML) for a variety of predictive analytics needs – use cases ranging from sales forecasting to churn reduction, customer acquisition, inventory optimisation, spend analysis, and more. As you can see, the breadth and depth of ML is extensive.
How Does Qlik AutoML Work?
Prepare your data
Qlik Cloud® includes a broad set of connectors that allow you to connect to data warehouses, business intelligence tools, and cloud applications where your data resides. Use your existing tools or Qlik Data Integration to prepare a training dataset, and Qlik AutoML will load the data and establish the ML pipeline.
As data is loaded, Qlik AutoML applies a variety of data science techniques to prepare data for training and testing, such as null handling, cardinality, encoding, feature scaling, cross-validation, and holdout.
Generate a machine learning model
Once your data is loaded and prepared, Qlik AutoML builds the predictive model. Qlik AutoML automatically tests best-of-breed algorithms for the dataset to determine the best fit. With the click of a button, you’ll see model results from hundreds of approaches and information about how they perform. Qlik AutoML will automatically select the best performing model, but your users can still choose any other model explored along the way – allowing them to contribute business context and intuition the machine learning may have missed.
Explore key drivers, refine your model
Qlik AutoML provides model metrics that expose all the relevant data science and statistical metrics to show you exactly how a model performs. For example, if you’re familiar with AUC, F1, Recall, Precision, Accuracy, R-squared, and RMSE, you can evaluate these statistics. There’s no black box here.
​
Feature Importance is a measure of how critical various key drivers are to the predictions in a model. Qlik AutoML uses two approaches for this: Permutation Importance and SHAP Importance. With Feature Importance, you’ll know exactly what key drivers are influencing the outcomes in your data. Ever wonder what the most important factors are for lead conversion? With Qlik AutoML, you get your answer.
Predict future outcomes
Next, load forward-looking data into Qlik AutoML and it will process the data using the model you built, then automatically add predictions and explainability values to your dataset. Qlik AutoML will include the predicted outcome as well as the probability of the prediction for each row of your dataset. In addition, it also adds a new column adjacent to every column in your dataset with Prediction Influencers, or Shapley values, communicating the positive or negative influence each column has on the prediction made. In other words, you’ll know the key drivers for each prediction made, and subsequently, what action you can take to influence the outcome.
You can download predicted datasets or securely deploy them into Qlik Sense or your analytics solution of choice. Or, you can integrate the model itself through Advanced Analytics Integration APIs to explore real-time predictive calculations and test what-if scenarios in Qlik Sense.
Test scenarios and plan decisions
You can leverage Qlik Sense to develop predictive analytics-oriented applications that tell the story of what’s going to happen and allow your users to interact through powerful Associative exploration. These apps include a full set of Prediction Influencers (Shapley values), allowing your users to evaluate possible actions without needing the expertise to use Qlik AutoML directly. Decision makers can understand predictions and the steps they can take to affect the outcomes at the record level – by cohorts, geographic regions, or any other selections they make. And through Advanced Analytics Integration, as your can perform what-if scenario planning, making selections or changing variables and getting new calculations directly from Qlik AutoML – in real time – so they can easily ask new questions and pivot their thinking in new directions. You’ve now achieved true prescriptive analytics.
Take Action
Qlik offers capabilities that compel immediate action based on your data. Traditional BI was built to inform users, and by extension to inform action, but not to trigger action. Capabilities such as intelligent alerting and application automation are meant to drive action when specific conditions arise. Combining Qlik Sense and Qlik AutoML, you can now drive human action and orchestrate events or workflows based not only on historical data but also on predictive analytics. Have a customer that would be more likely to renew if they were on a different plan? You could trigger an automated email to deliver an offer the same moment that insight is uncovered.
