No-Code Machine Learning with Qlik AutoML

Discover how no-code machine learning via Qlik AutoML empowers business analysts to generate automated models, gain insight into key drivers, and forecast outcomes for sales, churn, and optimization.

Explore the features of Qlik AutoML, including automated machine learning for analytics teams, predictive analytics, what-if scenarios, and publishable, explainable models for interactive dashboards.

Explore how automated machine learning powers use cases across healthcare, finance, retail, manufacturing, and public sector, including forecasting, risk analysis, churn reduction, and next best actions.

Explore the algorithms used by Qlik AutoML, including binary and multiclass classification and regression models such as random forest, logistic regression, extreme boost, and nearest neighbors.

Learn how the f-score, including the F1 score, measures model accuracy by harmonically combining precision and recall in binary classification, using true positives and false positives.

Learn to measure feature importance using permutation importance: shuffle each column, observe performance deterioration in model predictions, and quantify each feature's impact.

Explore how SHAP values break down an individual prediction using tree SHAP and linear SHAP, and apply local and global feature importance to improve model understanding.

Learn how model fitting measures a model's generalization by comparing predictions to real labels, adjust hyperparameters to reduce error, and uncover patterns for practical business insights.

Explain what a correlation matrix is and how it lists correlation coefficients for all variable pairs. Interpret positive, negative, and near-zero correlations for feature selection in machine learning.

Explore the receiver operating characteristic curve, mapping true positive rate to false positive rate across thresholds; learn how auc indicates performance and is threshold- and scale-invariant.

Explore how the confusion matrix evaluates binary classification by comparing actual and predicted targets (positive and negative), highlighting true positives, true negatives, false positives, and false negatives, and informing accuracy.

Set hyperparameters before training, external to the model, guiding learning and not part of the final model; they influence training speed and quality.

Assess real-world model performance with cross validation and automatic holdout. AutoML randomly stratifies data and reserves 20% as unseen holdout for final evaluation.

Sign up for a no-code machine learning free trial with Qlik AutoML, verify your business email, and start creating projects.

Start your no-code machine learning journey with Qlik AutoML by signing up, exploring the project interface, and starting a new project after managing data and seeking help when needed.

Add data sets to Qlik AutoML by uploading files or connecting to data providers, then manage datasets and split data into train and test sets for model training and predictions.

Set up a project and analysis in Qlik AutoML with your dataset. Configure a target, inspect data, build a pipeline, and run analyze with AutoML.

The lecture shows training a model, selects logistic regression with an F1 score of 0.973, and uses it for predictions; it analyzes feature importance, correlations, the confusion matrix, and accuracy.

Explore how Qlik AutoML cards reveal permutation importance and feature importance, interpret F1 scores, and refine and download analyses to improve no-code machine learning.

Review the training and test split, see AutoML pick logistic regression as the best model by accuracy and AOC F scores, then generate and download test data predictions.

Compare predicted versus actual values to assess model accuracy, review prediction probabilities and outliers, and note logistic regression achieving 98% correctness on the test data.

Explore the scenario editor to perform what-if analyses on predictions by adjusting parameters, running scenarios, and comparing current and scenario outcomes to understand model behavior.

Discover how to create multiple analyses in the same project by refining train data, excluding unused features, and updating pipelines and algorithms to improve predictions and F1 scores.

Discover how to download automated predictions with api keys and tokens, set authorization headers, and access results across apps in a no-code machine learning workflow.

Deploy a production version of your model via an API, enabling integration into workflows with post requests. Rename and manage multiple analyses and versions, and select the dataset for prediction.

Collaborate with teammates by inviting them via email, sharing project links, and setting permissions to view or edit your no-code machine learning projects.

Explore the no-code machine learning help resources in Qlik AutoML, including the help pane, tooltips, API docs, contact support forms, and demos to troubleshoot issues and learn features.

Explore no-code stock value prediction with Qlik AutoML using time series analysis on real stock data, select relevant features, set the high as target, and generate forecasts with visualizations.

Launch a no-code churn prediction project using Qlik AutoML. Set a binary target, train the model, and analyze feature importance and scenarios with the random forest predictor.

Build a weather forecast model in Qlik AutoML using Melbourne's historical daily temperatures to create a univariate time-series forecast, a no-code machine learning approach, and download or retrain as needed.

Perform multiclass text classification on the BBC news dataset with a no-code Qlik AutoML workflow. See automated preprocessing and predictions across business, tech, sport, and entertainment without coding.

Improve your model score by ensuring clean data, increasing sample size, removing outliers, and adapting training data after a significant change, while testing non-linear models and metrics to capture signal.

Explore diverse sources for practicing machine learning datasets, including open data portals, Google dataset search, Microsoft Research Open Data, and the UCI repository to build models with no-code tools.

Celebrate completing the no-code machine learning course and see how automated machine learning puts capabilities into the hands of business analysts for sales forecasting, churn reduction, and spend analysis.